Unlocking the GeForce 6800

Timmus writes "Firingsquad is running a story on how to unlock all 16 pipelines in nVidia's GeForce 6800. By default the card only ships with 12 pixel pipelines enabled, but with a tool and a few mouse clicks, the card can be unlocked to run with all 16 pipes. Performance improvements are seen everywhere, so it's a pretty nice free upgrade. These cards are currently selling for $200 online, so a 16-pipe GeForce 6800 delivers great bang for the buck."

Comments

Nice Work!

[coop0030]

I'm surprised, this actually showed a significant increase in performance in their charts. This is one of the best mods I have ever seen on a Video Card.

I wonder if this would actually hurt, or help Nvidia's sales, or have no effect?

I currently have an ATI card, and am very happy with ATI, but would be willing to switch to Nvidia since the price/performance on this card is so high now.

Re:Nice Work!

[Anonymous Coward]

I'm sure NVidia will be happy to jack up the prices of the 6800's once this mod becomes widely known.

Re:Nice Work!

[Anonymous Coward]

price/performance high = price too high for the performance.

Re:Nice Work!

[Adrilla]

I'm sure NVidia will be happy to jack up the prices of the 6800's once this mod becomes widely known.

or when they start shipping it with the mod unlocked themselves.

Re:Nice Work!

[ergo98]

I wonder if this would actually hurt, or help Nvidia's sales, or have no effect?

Well regardless of nvidia sales, this would be a perfect opportunity for the publisher (in this case FiringSquad) to have some good contextual ads - if there was an ad on the page to buy a 6800 I'd very likely have followed it.

Going back to nvidia, we're talking a 7-10% speed increase - this lets the hackers rush to get this card rather than considering some of the potent ATI competitors. I doubt it's an error that nvidia made the pipelines capable of being enabled.

Re:Nice Work!

[phasm42]

They do note that that some games had weird graphical glitches that came and went after turning the additional pipelines on, so the increase in performance may come at the cost of quality, depending on the card you got.

Re:Nice Work!

[darc]

This is unlikely. In the article itself, they note that graphical artifacts appear when you turn on this mod. This is likely because they didn't QA that section of the chip, and it's probably defective-ish. Now, that doesn't mean you'll definitely get a buggy chip, but NVidia won't unlock this because the part isn't tested, and assumed broken.

In fact, it may be reject parts from the Ultra series that makes it 6800 standard. From what it looks like, they deactivate the broken pipelines and then sell it as a lower model, much like CPUs do with clockrates.

Re:Nice Work!

[stratjakt]

Yeah, these types of mods are old hat. Go look for the Radeon 9500 to 9700 mods, etc. Unlocking pipelines is as old news as overclocking, and it results in a whole bunch of 'l337 g4m0r' kids with dead video cards.

What the hell is this "confirm im not a script" crap, anyways?

Re:Nice Work!

[Anonymous Coward]

RTFA. They do. It's called the 6800 Ultra and GT. NV40 cores that test best go into Ultras, cores that barely fail go into GTs and for cores that really suck they created the plain 6800. Basically you're enabling the defective part of your card.

Re:Nice Work!

[NetNifty]

Probably won't - IIRC you can do pretty much the same thing on some models of the NVIDIA 5900 to make it a 5950 by flashing it, and I don't think they increased the price to compensate.

Re:Nice Work!

[Mac Degger]

This kind of thing has been going on for a very long time. ATI does it too. What happens isn't that the card isn't tested, it's that it /is/.

After the card is made, they run a series of tests on it. If all parts work perfectly, you have the "6855,5 UltraDuper"; if all parts work, but instability occurs at higher clockspeeds, they call it a "6849 Ultra"; if certain parts (ie a few pixel/vertex shader units) don't work, they lock these off and call the card a plain "6800"; if more than a certain number don't work, they just trash the card.

Thing is, it's nothing new; ATI has been doing it (and softmods [software based] and hardware based modshve been available) since at least the 8xxx series. So whilst this is news, this isn't as hot as the blurb or /. would have you believe...it was actually innevitable. Shit, soon /. will have a press release out on how to mod your x800/6800 into a fireGL/quadro :) Whooppie...what news...

BTW, the same kind of thing goes on with cpu's, where it's called 'binning'.

Re:Nice Work!

[Cipster]

Yep. Just did this with a 6200 card. My 3D Mark scores went way up but I am sometimes getting artifacts. They are not enough to make me switch back but they are definitely noticeable at times.

Re:Nice Work!

[KarmaMB84]

They may also mark down fully functional parts to lower model numbers to meet demand for "low end" parts.

Re:Nice Work!

[dotgain]

Slashdot love doing things to their site, not announcing it, and then waiting around the corner with a sackful of "Offtopic" mods for those that discuss it, but anyway:

Since they got crapflooded a couple of weeks back, they have an whaddyacallit on the comment post page for everyone except those with excellent karma. So AC's and ratbags have to do it every time they post.

While I'm at it, M1 is a joke, and to actually M2 properly requires more time than folk are willing to spend if you ask me. How come you don't have to read at -1 if you've got mod points? What a joke.

Re:Nice Work!

[scotch]

What do you mean "confirm I'm not a script?"

Re:Nice Work!

[Decker-Mage]

He's probably seeing the same thing I am: "To confirm you are not a script, please type the text shown in this image: wtfwrxy". Stupid really, since there are two problems. (1) It ain't section 508 compliant (biggie to me as I'm also disabled, and these are really visually challenging to moi!). (2) I'm logged in as me and last time I looked, I ain't a script, although some might have a different opinion.

If they are really having problems with scripted spam from logged in users, block the fraggin' accounts! Or better yet, after you prove you ain't a script, remember it in the stupid cookie! How many times do you have to prove you are you?

Re:Nice Work!

[ameoba]

This is somewhat odd since manufacturers are known for making sure that hardware review sites get "golden sample" cards that are known to be defect free and overclock better than normal. On the flipside, many people in one of the forums I frequent not only unlock their cards but get monster overclocks without artifacts - generally not quite to 6800GT levels, but 25% boosts in performance aren't that uncommon.

I won't say this is the best card unlocking ever - I think that goes to the 9500->9700 trick but, if you already have the card, a pure-software tweak that can (relatively) safely boost performance this much is worth the small ammount of time it takes to perform & do some stabilty tests.

Re:Nice Work!

[Kazymyr]

This sort of thing has indeed been going on for a long time, and was invented by "granddaddy" Intel. The oldest example of a chip manufacturer turning off some capabilities on their chips and selling them as a lower end product that I'm aware of happened in the 486 era. Intel released the 486DX as a high end, expensive product with integrated FPU for the first time. Then they released the 486SX as a low end budget priced chip. The 486SX did not have a FPU, but what Intel didn't tell (though it became known rather quickly) was that the SX chips were identical to the DX except that the FPU was turned off in the fab. Thus the same assembly line produced both DX and SX chips.

And Intel even went the next step. They later marketed an "upgrade" for the 486SX which "added" a FPU to those systems. That was the 487DX, which was supposed to be installed in a separate socket on the motherboard and work in tandem with the existing 486SX. Again what Intel didn't say was that the 487DX was in fact a 486DX, and when it was installed on the motherboard it would simply turn off the SX altogether and take over. You could remove the SX chip and throw it away and it wouldn't make a difference. The 487DX was priced below the 486DX, but you ended up paying more for the 486SX+487DX than for a regular 486DX. And to prevent people from buying 487DX chips and use them instead of 486DX, they made the socket pinouts incompatible.

Re:Nice Work!

[Anonymous Coward]

You know there are bots that can read those things for you. =)

Re:Nice Work!

[sud_crow]

They may...

Actually they _do_ that.
And i mean 'they' 'coz its two of them, ATI and nVidia, both do it.
For instance, i know a guy who used to sell 9500 modded to 9500 Pro (announcing the mod, more or less like Companies now do, and sell a "secure overclocked" card), that is activating the 4 extra pipes on the board and overclocking a bit. Most of the times he would convert them without a glitch, but once in a while they would have some "artifacts" (wierd textures and such) display on some benchmarks or games.
I guess the saving for the companies lays in not having to test (QA assurance) those extra pipes or the clock rates, not in actually placing them in the board.

Re:Nice Work!

[chrisnewbie]

It's only going to make nvidia look like a cheap bastards that sells overpriced card for minor modification.

I hope it hurts them and that they lower their price.In canada their prices are crazy the top card goes for 800$

Re:Nice Work!

[ian mills]

Makes you wonder if this is what Sony is doing with Cell and how they have 8 SPE's and one for "redundancy" so only 7 are available.

Re:Nice Work!

[alexo]

> They may also mark down fully functional parts to lower model numbers to meet demand for "low end" parts.

A lot of "they" do it routinely.

If your manufacturing capacity is N units and the demand is for, say, 10% high end and 90% low end, it is cheaper to only test until you get the 10% (or slightly above) working at the high speeds and market the rest as low speed, untested.

I hate to trolll but....

[packeteer]

Thsi is kinda old news. People have been doing this for about a year now ever since the card came out. Either way its a good guide to getting some extra bang for your buck although everyone needs to remember that if the card worked 100% fine with 16 pipelines they would have sold it that way.

Re:I hate to trolll but....

[grasshoppa]

although everyone needs to remember that if the card worked 100% fine with 16 pipelines they would have sold it that way.

Not necessarily. If they need to fill a price point, chip companies will sell the higher grade stuff at a lower price point and intentionally cripple it.

Intel and AMD have been doing it for years, and they are hardly the only ones.

Re:I hate to trolll but....

[iminplaya]

...if the card worked 100% fine with 16 pipelines they would have sold it that way.

A lot of equipment is sold with diminished capacity to cover up possible defects.

Re:I hate to trolll but....

[Anonymous Coward]

Thus, the point of the grandparent poster...

Re:I hate to trolll but....

[jericho4.0]

It's easier and cheaper to design one fast chip, and cripple half of the chips to sell to a lower market. Any company who fabs chips does this.

Re:I hate to trolll but....

[michaelbgrant]

No silly, this is like the flashing the bios from 6800GT to Ultra. They are the same card with different bios.

Re:I hate to trolll but....

[damiam]

Yeah, but probably binned differently. So the cards on which all sixteen pipes work are thrown in the Ultra bin and sold as such, while those where the extra pipes show problems have them deactivated and are sold as vanilla 6800s.

Re:I hate to trolll but....

[Breakfast Pants]

They don't always cripple it intentionally. They design for instance cache so that if there is a defect in one half of it they can disable it and still use the other half (ever bought a celeron?). So the majority of it isn't just straight out intentional crippling, but I do admit that has happened a good bit in the past.

Re:I hate to trolll but....

[msim]

because of the italics i read that as ...if the card worked 100% fine with 16 pipelines they would have sold it that way.

I then thought what the??? i need another coffee i think.

Re:I hate to trolll but....

[Anonymous Coward]

In many cases they do deliberately cripple it even though it works perfectly. The Celeron 366 which can almost always be safely overclocked to 500MHz or higher is an example of this. Unfortunately it is hard to tell whether it was crippled because part of it failed testing or if it was crippled only because they could make more money by selling it as an inferior product unless you test it yourself.

Re:I hate to trolll but....

[masklinn]

You also often find that with processors: good steppings are often "perfect" (more or less) chips that got crippled to fill low end/mid end demand of the sales market

Re:I hate to trolll but....

[Anonymous Coward]

The old apple II 6502s had lots of undocumented or unofficial features that were on the edge of the die. At the time, the technology was such that these things would only work about half the time; hence, they weren't part of the official feature set, but typically you could get some of these things to do something.

Re:I hate to trolll but....

[Cat_Byte]

Not necessarily. If they need to fill a price point, chip companies will sell the higher grade stuff at a lower price point and intentionally cripple it.

Intel and AMD have been doing it for years, and they are hardly the only ones.

486 SX and 486 DX come to mind. SX was usually the ones that failed the match co-processor test so they just disabled it and shipped it as an SX.

Re:I hate to trolll but....

[bhtooefr]

Nitpick: The old Celeron 366 was its own core (Mendocino) - read: not a crippled Pentium II core. In fact, it had MORE functionality than a Deschutes P2 core, as it had on-die L2 cache. Intel didn't start disabling broken cache until the Coppermine Celerons - the Covington chips were Klamath (IIRC) cores, but without the cache chips attached (remember, Pentium IIs didn't have on-die L2, just on-cartridge).

Now, clock speed was crippled - most Celerons could get to at least 525MHz, if not higher (except for the 300A and 333 - they couldn't get that high because they (or their chipsets, but last I checked, a BX could handle an OC to 133 or higher) couldn't handle the high FSBs required for those speeds). Intel even sold 533MHz Mendocino Celerons (they were really bad buys, as a 366 could get to 550 with not much more cooling).

I turned

[Jozer99]

I turned my 6800 into a Radeon 9700 Pro with my 1337 sk1lz!

Re:I turned

[01000011011101000111]

I hope you H4xor'd it Nekkid - otherwise your box is toast ;)

Re:I turned

[justforaday]

but are you 1337 enough to turn it into an AIW while yer at it?

Re:I turned

[indy_Muad'Dib]

i was l33t enough to turn a Hard Drive into a smoldering mass of plastic and aluminum once...

Mental note, connecting 2 12v plugs into a drive does not turn a 7200RPM drive into a 14400RPM one.

Re:I turned

[msim]

Ah a fellow kindred spirit, i have managed to nuke a scsi card by accidentally peircing a wire on the scsi cable with the backs of one of the pins elsewhere on the card. but 24v, wow, i can't beat that!

Re:I turned

[rhennigan]

I turned my 6800 into a heap of melted plastic with my wannabe sk1lz!

Re:I turned

[Anonymous Coward]

Wow, your 1337 skilz gave you an increase of 2900! Amazing! Plz post pics!

Re:I turned

[buysse]

Plugged a floppy cable in to one of the expansion ports on an old DPT SCSI card (EISA!), back in the day. The ports were meant for RAID processors, cache, battery, alarms...

It had the right number of pins. However, the floppy cable melted and ignited when I turned the machine on, and some of the burning plastic hit the motherboard.

My boss was not pleased.

Still not as fun as making an Apple Disk II controller explode (the old external ribbon cable 5.25" floppys for Apple IIs, if you offset the cable right, it fed power back to the controller chip which would then explode.)

Re:I turned

[hawk]

The original TRS-80 used the same connector for power, video, and cassette.

I learned this from my fuming boss who had reached behind the unit to connect cords he couldn't see. Found sockets that fit everything, and *ZAP!!!*.

He was even more dismayed that anyone would design such an idiotic system than he was angry at losing the system.

hawk

Re:I turned

[Jozer99]

I can turn a 3Com chipset 10/100 Ethernet card into a TV Wonder HD with some tin foil and a pair of earbuds. Beat that!

The next 9500pro

[MasterJeff]

I had a 9500pro and it was a great deal. Hopefull this can fill that roll and I may just pick one up!

Re:The next 9500pro

[WhatAmIDoingHere]

Did you already do the softmod to turn that into a 9700?

Re:The next 9500pro

[Bachus9000]

The softmod was for the vanilla 9500, not the pro model. I know because I tried to soft mod the 9500. It worked, but I got lots of fun graphical artifacts all over, so I quickly went back to the regular drivers. Now that I've said that, though, there could have been a mod for the pro chips, but AFAIK the 9500 Pro already used all eight pipelines.

Re:The next 9500pro

[mabhatter654]

The pro was already a slower 9700. It was the first to hit the price point so they literally used "underclocked" 9700s... there's not much you can push a 9500pro because it's still limited by the cheaper memory they used... When the next 9500 & 9600 came out they used an entirely different core stripped out of the extra piplines...but it could go faster.. so depending on the game, a 9500pro is still a "competive" card.. where the 9600 gets a bit dated

Disabled Hardware??

[Khyber]

In the first place, why would you deliberately lock down 4 of the pipelines to begin with? Wouldn't it make more sense to just go ahead and have all 16 pipelines pumping out the frames in the first place, to give a TRUE impression of what the card can actually do, instead of crippling the card?

Re:Disabled Hardware??

[Siergen]

MBAs learned long ago that in many businesses you can make more money selling both high-end and low-end products in the same market than you would by selling just high-end. Disabling 4 of the pipelines allows them to do this with one main production line and one product development effort.

Re:Disabled Hardware??

[edwdig]

Two reasons to turn off some of the pipelines:

a) They were defective, and this allows you to salvage the part.

b) People with too much money will gladly pay significantly more money for a slightly higher end version of the same card.

Re:Disabled Hardware??

[Anonymous Coward]

You could, you know, maybe read the article and learn the answer to that.

Re:Disabled Hardware??

[Azreal]

Usually the 4 pipes are locked for couple of reasons. Early on, when the card first came out, demand was fairly high across the board for 6800-6800ultra's and production was fairly low. As a result, some 6800's were basically 6800gt's with pipes locked, or in the rare instance of like the evga 6800limited edition (not to be confused with 6800LE), it was a 6800ultra clocked down or pipe locked...I forget. Now, production has basically met demand and the pipes are usually locked because of defects in the manufacturing process to said pipes. If a single pipe is "broken" it can't be sold as a 6800gt so 3 other pipes are locked to make it a 6800 vanilla. You can unlock all 4 pipes but you'll usually see artifacting. If you can figure out exactly which pipe or pipes are broken, you can re-lock those ones and have a better performing 6800..but not quite a 6800gt.

Re:Disabled Hardware??

[Anonymous Coward]

Its because when one makes silicon chips, testing them is expensive, so any removal of features means less stuff to test, and hence is cheaper to make. Also, turning off one of the pipelines means defective chips, that might otherwise be rejected can be sold. Given that fab processes are far from perfect (good ones have ~40% yield), any extra chips you can sell from a wafer = more $$$.

Re:Disabled Hardware??

[Khyber]

I *DID* read the article. Still, the question makes sense to ask. And for the hell of it, let's add another question, you coward.

Why would the PCI-E version only ship with 12 pipelines, hrm? Especially when it's got double the bandwidth and two-way writing?

Re:Disabled Hardware??

[jimi the hippie]

But then they couldn't put the same (except not disabled) chip in another card and then show off the huge jump in performance you get for the extra $50.

Re:Disabled Hardware??

[jericho4.0]

You were modded "troll" because you called an Anonymous Coward a coward.

Welcome to /.

Re:Disabled Hardware??

[mikael]

For profit reasons, a graphics chip maker wants to cover all possible market segments. For the cheap end of the market, they provide a card with basic functionality ie. four pipelines. For the expensive end of the market, they provide a card with as many pipelines as possible (16, 32, 64). To fill in the space between, they have graphics card with different capabilities - card makers can't really nobble screen resolution or colour depth any more, so they are left with 3D performance. And so they provide cards that have things like overlays, hardware accelerated line drawing and multiple pipelines disabled.

Re:Disabled Hardware??

[angle_slam]

Didn't Intel do something like this? I forget the exact model number, but they used to sell a crippled version of the 386 with the co-processor disabled. If you wanted to buy the math co-processor, you bought a 387. But, in reality, all the 387 was is a 386 that disabled a crippled 386. (It might not have been the 386, it could have been the 286).

Re:Disabled Hardware??

[angle_slam]

I should have just Googled before I posted. It was actually the 486 that had this "feature". The 486DX was the first Intel '86 CPU with an integrated floating point processor. The 486SX is the crippled version. The 487SX was merely the 486DX that also disabled the 486SX. See this page.

Re:Disabled Hardware??

[Jozer99]

Yes, but then the car performs better for the price. Imagine this: A $249 6800 (I haven't actually check ed the prices today, but don't have a spaz, I am just trying to prove a point) performs about half as well as the $499 6800 Ultra. That makes sense, twice the price, twice the performance. Now, unlock the pipelines, and the 6800 is now the same as the 6800 Ultra, but with a slightly lower clock and slower RAM. Now you have 80% of the performance with 50% of the price, and sales of your top card drop. Now, the top of the line card always has a high profit margin, because people who need the extra 0.2 FPS in Doom 3 don't worry too much about price. Now, if you raise the price on the 6800 to compensate for the increased performance, 80% price for 80% performance, you are now stealing sales from the 6800GT. And then there is a gap in you lineup, the middle of the line card is missing.

Re:Disabled Hardware??

[Anonymous Coward]

If you'd read the article, you might have seen an explanation of what you're asking about:

"In order to decrease manufacturing costs, NVIDIA (like all semiconductor manufacturers) uses tactics such as tweaking clock speeds and disabling features on their chips in order to service multiple markets with what is essentially the same product.

In the case of the NV40 family, the chips that are verified and tested to run at the highest clock speeds are put into GeForce 6800 Ultra cards, while those that just miss the cut go into GeForce 6800 GTs. NVIDIA then played it safe and created a third NV40-based product, GeForce 6800. Typically whenever a brand new product is introduced, some chips just don't cut the mustard and aren't able to run at the required clock speeds (especially when its built on an untested manufacturing process). In NVIDIA's case for NV40, they were concerned that some chips may not be able to hit the clock speeds they'd envisioned with all sixteen pipelines running. Rather than throw these chips away as waste, they instead go into GeForce 6800 cards."

Re:Disabled Hardware??

[tokennrg]

Sounds kinda like the thing PC manufacturers used to do with the "Turbo" button. If you turned the turbo off it would actually slow the CPU down.

Re:Disabled Hardware??

[Waffle Iron]

Sounds kinda like the thing PC manufacturers used to do with the "Turbo" button. If you turned the turbo off it would actually slow the CPU down.

That was there for a very specific reason: the lame-assed DRM in the original Lotus 1-2-3 used a CPU delay loop to time hacks on the floppy drive that they used to prevent normal copies from working. The DRM scheme failed with CPUs that ran faster than the original 4.77MHz 8086.

Therefore, to load Lotus 1-2-3, you had to turn off the turbo button to slow your machine down to the original speed of a 4.77MHz PC. It was also useful to run a handful of early games that used CPU speed to time the action.

What was really stupid is that the DRM scheme drove millions of otherwise law-abiding people to use questionable cracked copies. The original DRM'd 1-2-3 floppies were so precious, and floppy disks were so unreliable and subject to wear, very few people would risk using their original disks for day-to-day use. Most everyone I knew, even in large corporations, used cracked disks instead. The original disks stayed safely on the bookshelf in those thick cardboard ring binder + carton combos that software always used to come in.

Re:Disabled Hardware??

[Kris_J]

c) Something else was wrong and the entire card was downgraded.

Re:Disabled Hardware??

[Kris_J]

The 486SX wasn't crippled on purpose, it was a 486DX with a faulty maths co-pro. But you are right that the 487, or Overdrive, or whatever it was called was in fact a complete CPU that took over from the original.

Re:Disabled Hardware??

[be-fan]

Its a market issue. Basically, the defect rate of the 300m transistor chips ensures that a significant number will have at least some non-operational pipelines. Instead of throwing away those cores (and increasing prices to compensate), they clock them lower, pair them with cheaper RAM, and sell them as cheaper cards. Now, if the manufacturing process gets better, they might not have enough defective cards to sell as cheaper parts. They can't just take all the perfect chips and sell them as expensive cards, because then there would be a shortage of the low-end product which would drive up their price. So they take the perfect chips and put them in budget cards.

Re:Disabled Hardware??

[be-fan]

No, because in the case with the Turbo button, there was nothing wrong with the chips. This is more like a retailer sending clothes with slight defects to an outlet store where its sold at a cheaper price.

Re:Disabled Hardware??

[Anonymous Coward]

Well more Economists than MBAs. It's called second degree price discrimination or product differentiation. The only real novel thing is that they can reduce costs and waste by having only one production line and then binning rather than multiple production lines.

Re:Disabled Hardware??

[Khyber]

Thanks for the honest input, if you were the one who modded me as a troll.

Re:Disabled Hardware??

[jericho4.0]

I was making a joke, and slashdot doesn't allow one to moderate and post in the same discussion. I really have no idea why you were modded troll, slashdot is a great reaffirmation of the "everyone is a stupid idiot" school of thought. I keep coming back though....

Re:Disabled Hardware??

[ameoba]

The 6800 chip was designed for AGP. nVidia saw a large enough market for PCIe cards that it would be more cost effective to sell a native PCIe chip than it would be to sell AGP parts with an AGP/PCIe bridge. It's a fairly minor change, but they saw fit to remove the extra 4 pipes from the fabrication process.

Re:Disabled Hardware??

[innocent_white_lamb]

The 486SX wasn't crippled on purpose, it was a 486DX with a faulty maths co-pro.

Early ones. There are actually two major versions/types of the 486sx CPU. Early 486sx's were 486dx's in "sheep's clothing", as you said. Later 486sx's left the FPU off of the die completely and really were FPU-less CPU's rather than defective 486dx's with a lock-out.

Re:Disabled Hardware??

[Mycroft_VIII]

IIRC the original 486sx were indeed units with bad math-co's that had been dissabled that intell sold to help recoup some of thier manufacturing costs.
But they sold enough of them and then the slightly reworked dx's as 487's that they took to deliberately dissabling the math-co's (even after yeilds got much much better) on perfectly good 486dx's to keep up 'low end' sales and of course sales of 487s. This allowing them to not only sell more of each waffer, but have higher total sales.

Mycroft

Re:Disabled Hardware??

[LarsG]

Why would the PCI-E version only ship with 12 pipelines, hrm?

It is a question of manufacturing yields and economics.

When the NV40 was introduced, they were manufactured using a brand new process. nVidia expected a low yield, leading to many parts with some vertex/pixel pipelines botched. So, instead of manufacturing two different cores - one with 12 pipelines for the mid-end and one with 16 pipelines for high-end, and then having to throw out all chips with defects - they instead only manufactured 16 pipeline cores, and chips with a certain number of defects could be down-graded to 12 pipelines. In effect increasing the number of chips that could be sold instead of thrown away.

As the manufacturing process matures, the number of chips with no defects increases. But since there are only so many people that want to buy a high-end chip, nVidia has to deliberately downgrade chips to fill the market demand for mid-end graphics cards.

Now, a 16 pipeline core is larger than a 12 pipeline core, meaning fewer chips per wafer. Once the manufacturing process is good enough, the required number of working high- and mid-end chips can be manufactured cheaper by using two different cores. Evidently, that line was reached by the time the PCI-E versions of the 6800 were introduced.

Note that this is quite common in the semiconductor industry.

Selling chips with certain faults as lower performance (or having redundant functional units on the core), decreasing the number of chips going to the trash. 80486DX with faulty math coprocessor were sold as 486SX. RAM chips often have a few redundant memory banks.

Deliberately down-clocking/-grading chips to fill market demand. Celerons based on a mature core / manufacturing process are usually very good overclockers.

Re:Disabled Hardware??

[koreaman]

If too many people want to buy the lower-quality cards, why doesn't NVidia just lower the price on the working ones? If they did that sufficiently, they could move demand from the lower-end chips to the higher-end ones and compensate properly.

Re:Disabled Hardware??

[Anonymous Coward]

Did you somehow miss his point "a"?

Re:Disabled Hardware??

[bhtooefr]

Wow. Didn't know they did that crap on the PC, too - just knew about it being that bad on the Apple II. *said with an old Copy II Plus disk in hand*

Anyway, I don't think most manufacturers did it for Lotus 123. Most manufacturers did it for the games that timed against the CPU, instead of a constant source (e.g., the RTC - although, that wasn't necessarily present in early 8088-based systems). Basically, they did some simple math, and figured that if the processor is going to run @ 4.77MHz, then if they put a delay of 4,770,000 clock cycles in (oversimplification - I don't know how many cycles a NOP takes on an 8088), it would wait 1 second. Problem was, when they went to faster chips, their 4,770,000 cycle wait took less time - a 33MHz chip would take ~0.15 seconds.

Re:Disabled Hardware??

[bhtooefr]

But if they artificially keep the price high on the higher-end chips, they can make more money on the higher-end chips. Observe:

1. Make high-end chips.
2. Realize that high-end chips have problems, create mid-end versions that don't use broken parts.
3. Improve process to the point that even the high-end chips don't have problems.
4. ??? (Hmm... maybe NOT lower the price on the high-end chips, and assume some are broken anyway?)
5. Profit!!!

They DO make money on the mid-end parts, but why price the high-end parts the same, when they can make more money?

So...

[FlyByPC]

Can I run this on my 6800GT, and end up with twenty pipelines? <g>

Re:So...

[Kazymyr]

No, but you can run it on two 6800GTs and end up with 32 pipelines.

I'll wait a while..

[Anonymous Coward]

till they unlock 64 pipelines.

question

[Ichigo Kurosaki]

how many pipe \lines will the 7800gtx ship with? how many will be enabled? hopefully it wont require tweaking to get max performance.

Re:question

[be-fan]

Um, why shouldn't it? Nobody releases parts tweeked to the maximum they can, because if they did, there is a very high possibility the parts would fail. Would you prefer NVIDIA just release the parts with all pipelines unlocked, and when you bought one there would be a 50/50 chance some of them wouldn't work?

no,

[Run4yourlives]

because you want to charge an extra $50 for those cards.

It's cheaper to do things this way than it is to actually alter your production lines.

No one has mentioned this yet, but...

[JMan1865]

AGP only, it seems. No love for us PCI-E types.

Re:No one has mentioned this yet, but...

[alexmogil]

Yeah, another great move by me. But I have a faster bus, right?

Hello?

Re:No one has mentioned this yet, but...

[Anonymous Coward]

RTFA. The PCI Express card is built with a different core that's built with only 12 pipelines from the start.

Re:No one has mentioned this yet, but...

[Anonymous Coward]

He obviously read TFA, seeing as he lamented the fact which you so redundantly restated: The mod won't work on the PCI-E version.

And now I did it again! :)

Re:No one has mentioned this yet, but...

[Pahroza]

I've got a 6800GTO (in my cheapo Dell) that comes with PCI-E. I was able to make the change from 12 to 16 in mine without any problems (so far).

Actually the 6800 has 64 pipes

[Anonymous Coward]

And shipping processors only have 12 pipelines enabled. A member of the G70 team told me they sat around all this time smoking dope and doing nothing and they only have to short a small jumper block on the 6800 card to enable the new goodness and look like engineering champs.

cheapskate's advice

[tofucubes]

not to be jerk, but spending $200 on a video card ALREADY seems like overkill...the video card industry has coming out with faster and faster cards when alreadly the video cards were overkill.

People (expecially benchmarkers) have to turn on all the bells and whistles and crank up the resolution and everything just to make "full" use of thier video card. Nice to have and beautiful and all, but if you are willing to turn down resolution you can have a hundred dollars and run just as smooth.

Re:cheapskate's advice

[Anonymous Coward]

oh yeah just want to say that I appreciate the article and its cool especially since its free for those with the video card already ;-)

Re:cheapskate's advice

[Vertdang]

I'd say $200 is pretty decent.

I used to be the kind of guy that would spend $400 on a new videocard that you can get for $200 within 6 months. That's since changed

Last night, my mother in law bought me a new videocard (CRAZY!) because she missed my birthday a month ago... it's an ATI X700Pro 256meg that ran $179.00 after instant $20 rebate.
I installed it last night and it was VASTLY superior to the card I got 1&1/2 to 2 years ago for nearly $500. I turned everything up on Farcry, WoW, and BFVietnam. It's smooth as glass on my AMD3000 machine.

I wont be shelling out that much money again anytime soon.

Re:cheapskate's advice

[Locke2005]

You've obviously never played Doom III, have you?

Re:cheapskate's advice

[be-fan]

Yep. It's far better to get $200 ($150 retail) every year than to get $400 cards every couple of years. I recently build a machine, and I got a 6600GT instead of a 6800GT, even though my budget for the computer could have easily handled it. Why? Because next year, when the midrange G70-based cards come out, I can easily conscience spending another $150 to replace this $130 card, while if I had bought a $350 card, the decision would not have been so easy.

6800?

[Rorschach1]

Pff. I'll save my money for the Radeon 6502.

Re:6800?

[Anonymous Coward]

not a bad little chip that 6502. :)

Re:6800?

[LarsG]

But I'd much rather have a 68060. ;-)

I was all excited...

[Whoozit]

... until I noticed it's for AGP only cards. I got a 'cutting edge' PCI-Express 6800... which according to the article are made with only 12 pipelines. *sigh*.

Well, I guess I'll just buy another card and turn on the SLI on my motherboard... hehe... :D

Re:I was all excited...

[toddthefrog]

You can't SLI vanilla 6800's, only the 6600GT, 6800GT, and the 6800 Ultra.

You should have dipped into your wallet less (for the 6600GT) or more (for the 6800GT).

Re:I was all excited...

[be-fan]

Um, you should have known that before you bought the card. It's not like this information isn't in every GeForce 6800 review in existance...

Re:I was all excited...

[lordbeejee]

You can sli regular 6800's: http://usa.asus.com/products/vga/en6800/overview.h tm

DMCA

[a_greer2005]

How long befre nvidia sues

Re:DMCA

[Rude Turnip]

Probably right after the Q-Tip people sue me for breaking the end off a swab and picking my teeth. And definitely not before GM sues me for installing an after-market car alarm.

Re:DMCA

[zr-rifle]

Hmmm, lets say more or less a year, if nVIDIA sued tomorrow? This mod has been around for this amount of time now, and I'm suprised this has made it to the main page as "news".

More or less the same happened back in the Quadro/MX days. Using a soft-mod similiar to the one discussed in the article, you could transform your cheapo-but-great Geforce 2 MX into a much more costly Quadro card for graphic professionals.

The difference between now and then is that this mod isn't guaranteed to work for each and every 6800 card out there. So unless you get to test the hack before buying don't think you can get away with a 6800 Ultra at a lower price. If you buy it, take it home and the mod doesn't work because the extra pipelines are defective, you're stuck with a really pricey card (check the V9999 from ASUS) that will perfom worse per dollar than a 6800GT (or probably even a 6600GT).

Also, please read the DMCA again: this kind of mod is perfectly legal, unless you plan to put up a shop that sells them. If it weren't so, even overclocking would be illegal.

Re:DMCA

[Breakfast Pants]

Even if you put up a shop that sells them this is still legal under the DMCA. At least judging by the Supreme Court not taking up the Lexmark case.

Re:DMCA

[Anonymous Coward]

there is no encryption or circumvention actually happening here since its not like turning off the pipelines was done to protect the integrity of their intellectual property in any way.

Re:DMCA

[pipingguy]

So what exactly *is* the difference between the high-end consumer cards and the "professional level" cards aside from cost? I'm talking about the relative cost difference for a $2000 card vs. a $750 one. Does the percieved quality scale well in real-world use?

How can the professional level cards be specialized for performance when there are so many professional applications out there - is each professional card custom-tweaked to each application?

Please forgive me, I remember the old days when one needed a $4000 videocard and a $10,000 machine to see real-time, semi-photorealistic graphics.

Big Deal

[rsilvergun]

I can break a graphics card too, and I'm not even 1337.

is there a centralized list of these hacks?

[schweini]

does anybody know whether there's a website where a comprehensive list of these 'free update' hacks is maintained?
there seem to be an awful lot of them (Sony Clie 710->740, Siemens A55->C55, 720kb-->1,44MB Floppies, etc.) but usually they pop up in rather dubious threads on some weird forum, and having them in one nice place would certainly be nice.

Re:is there a centralized list of these hacks?

[cyberwiz01]

The overclockers.com forums has a sub-forum specifically for video cards and their mods.
http://www.ocforums.com/
check out the video card section here:
http://www.ocforums.com/forumdisplay.php?f=7

Reminiscent of Cannon 300D Hack

[CodeBuster]

This sounds rather like the incident which occurred with the Canon EOS 300D Digital Rebel where the entry level model had very similar features to the higher end "professional" model costing hundreds of dollars more. However, it was discovered by some enterprising users that a relatively simple hack, flashing the BIOS with a modified version, could "unlock" the hardware and enable most of the features that were found on the more expensive model. This type of hardware homogeneity protected by software locking is advantageous for the manufacturer because it reduces manufacturing costs, since only one version of the hardware need be produced, but it is also vulnerable to those users who are sophisticated enough to circumvent the software locks. Is it possible that NVIDIA was holding back these pipes as a stop-gap measure so that they could release a new "Ultra" version of an existing card on short notice to counter a new competitor release more quickly? Perhaps, but these two incidents, the camera hack and now the video card hack, may induce corporations to rethink their software locking strategies. In the meantime it appears that savvy consumers can reap the benefits of these companies' mistakes.

Re:Reminiscent of Cannon 300D Hack

[BiggerIsBetter]

I think it's more likely to push companies to rethink their legal strategies. I'd expect to see some trivial copy protection, and people getting sued (or at least threatened) for breaking it to upgrade their software. Even an EULA that specifically says, "Thou shalt not flash any model other than X" might be sufficient...

Re:Reminiscent of Cannon 300D Hack

[Anonymous Coward]

Perhaps, but these two incidents, the camera hack and now the video card hack, may induce corporations to rethink their software locking strategies

Although maybe new to yourself, hacking products in this manner is not new to manufactors or others. Remember the old SS/DS floppy disk debate? My Yamaha 200T 2x burner had a wire to clip and could be flashed to a 400T 4x back in 1997. I'm sure Nvidia is not happy but the fact reamins that for many electronic products, it is far cheaper to maintain one single assembly method and make one actual physical product and cripple it with a jumper or software then to support multiple product lines.

Re:Reminiscent of Cannon 300D Hack

[Anonymous Coward]

as smart as you sound it seems that you didn't know that people have been doing this to video cards for a long time.

Re:Reminiscent of Cannon 300D Hack

[Autobahn]

may induce corporations to rethink their software locking strategies

Or add weak encryption then sue under the DMCA

Re:Reminiscent of Cannon 300D Hack

[Kris_J]

It might make manufacturing cheaper, but it's like anti-marketing. If the market finds out you're screwing them over like this it does a lot of damage to your credibility.

Re:Reminiscent of Cannon 300D Hack

[fm6]

If the market finds out you're screwing them...

Who's being screwed? If you don't pay as much, you shouldn't expect to get as much.

Re:Reminiscent of Cannon 300D Hack

[be-fan]

How are they "screwing you over"? They're selling you the product that you paid for. Futhermore, the price you paid for it is quite a bit less because the "perfect" parts (the ones able to run with all 16 pipelines at higher clockspeeds), are sold at a premium, allowing other parts to be sold more cheaply.

Re:Reminiscent of Cannon 300D Hack

[KarmaMB84]

The fact is, the manufacturer is doing the consumer a favor by simply disabling parts of the premium product and selling it for a lower price. If the guys shelling out the bucks for the full product weren't subsidizing the lower end models, the manufacturer would be charging more for a different design that still didn't have the features.

In the case of GPUs on video cards, hardware is usually disabled due to defects and marked down.

Re:Reminiscent of Cannon 300D Hack

[Kris_J]

But if I pay more I should expect to get more.

If a company produces two products; A and B, product A being product B with a couple of features technically available but disabled you have to ask why product A exists at all. Surely product B should just be sold at product A's price. If a company is selling crippled products at below-cost one can only assume they're trying to manipulate the market. If they're then selling the uncrippled version at a high markup, what exactly are they pulling?

Re:Reminiscent of Cannon 300D Hack

[dotgain]

Uh huh, software does the same too. Case in point, Microsoft Exchange. Available in Standard and Enterprise editions (maybe others too). Standard Edition, limited to one store, and that's limited to 32G. Enterprise Edition, no such limitations. It's not like they'd rewritten the software at all. Just open a header file, change a couple of constants, and recompile.

Probably exactly the same for XP Pro vs Home, SQLServer vs MSDE etc.

Re:Reminiscent of Cannon 300D Hack

[dotgain]

Actually I should say where you're actually removing whole pieces of functionality, as they to with XP Home, you'd probably not compile in those secions of code at all, but then it's Microsoft, XP Home may indeed have all the domain membership code lying around somewhere, just begging to be unlocked. Anybody know?

Re:Reminiscent of Cannon 300D Hack

[innocent_white_lamb]

hacking products in this manner is not new to manufactors or others.

It goes back a lot further than 1997.

I re-formatted a 10mb IBM hard drive to a whopping 15mb or so by putting it on a RLL controller card instead of a MFM controller card, and used that 15mb to hold FidoNet messages on my Opus BBS around 1990(?) or so.

I'm sure other folks can come up with much earlier examples.

Re:Reminiscent of Cannon 300D Hack

[Maljin Jolt]

I'm sure other folks can come up with much earlier examples.

1985(??), ZX Spectrum "Turbo" ROM, several nasty bugs in basic patched and 6 times faster cassette tape loader code, speeding up full 48k RAM loading from about 15 minutes down to less than 3.

Re:Reminiscent of Cannon 300D Hack

[Mycroft_VIII]

No they are not manipulating the market, but rather taking advantage of known properties thereof.
This has been going on for some time, since before modern pc's I'd bet, in other markets as well. And it's a fact most people learn about same time as they get old enough to vote or drink, so no one is being deliberately fooled.
It's simular to how you can find two brands of the same product in a store with as much as a 2:1 price difference and yet they are the EXACT same thing.
For example I once worked at a factory that packaged charcoal briquettes(sp?). There are exactly four types of them defined by two binary properties. All hardwood or mixed, with or without lighting fluids added. That's it, the packaging is just packaging and who's name on the outside has no bearing on the contents. When we had enough of brand x all we did was switch the bags we were filling, it all came from the same bin with absolutely nothing done between bag switches.
Yet go look in a store at the costs of the various brands. The same amount of the same thing can cost from $5 to $20 depending on who's name is on the bag.

Mycroft

Re:Reminiscent of Cannon 300D Hack

[Viceice]

What is wrong with you? I'd rather all this was hushed up and nobody but us geeks in the know knew about it. That this never hits mainstream news and that the companies know as little as possible about all this hacking.

If companies rethink this software lock out strategy, what exactly do you think is going to happen? I'll tell you what. Our cost of living is going to go up. So if they can't make product A and sell A-1 at $200 and A-2 at $100 then they will just sell product A at $250.

Companies exist to make a profit. Don't think for a moment that because you protest, they will sell product A at $100.

This is just like how companies gloss over the 2% credit card tax.

Re:Reminiscent of Cannon 300D Hack

[BVis]

I wonder how long it'll be before nVidia decides to pull a Lexmark and start suing people who distribute this information/software under the DMCA.

After all, if you can buy a cheaper card and turn it into the equivalent of the more expensive card, you're hurting nVidia financially. And while the clause in the DMCA I'm thinking of refers to defeating copy protection, one could argue (provided one threw enough lawyers/money at the situation) that unlocking the extra pipelines that nVidia locked constitutes defeating a software protection.

IANAL.

Re:Reminiscent of Cannon 300D Hack

[fLameDogg]

That's interesting WRT the charcoal. Around the beginning of spring, Albertsons had a good price on 20lb bags of Kingsford charcoal, so I grabbed one. The bag made a big deal out of the fact that all the briquettes were marked with a "K". Sure enough, it was kind of rough-looking, but there. I wonder if it's their way of saying "see, we make our own charcoal", or if they've just cut some kind of a deal with the big charcoal-making entity.

I use extra lighter fluid to overclock my charcoal. There, back on topic.

Re:Reminiscent of Cannon 300D Hack

[Mycroft_VIII]

Quite possibly, back when I worked in that factory Kingsford was one of people sending us bags to fill. All the charcoal was the same, no K's on any of it.

Mycroft

Canon Rebel

[Bifurcati]

It reminds me a little of when Canon released their sub-$1000 Digital SLR (300D). It had quite a few features disabled by software (such as flash power setting and I think some white balance settings). People promptly wrote a new firmware for it, and you ended up with a camera that wasn't so far off their more expensive model.

I understand why they do it, but if I'm paying for the hardware, I want to jolly well use it!

Be warned though

[Anonymous Coward]

This does not work on all GeForce 6800's. If you research this hack, as stated by another poster this has been a known hack for some time now, you will find information on why this is possible and why Nvidia "locks" some of the pipes. I seem to remember something about problems inherent in there manufacturing processes.

You can unlock all pixel piplines as will as additional vertex shaders. I bought a 6800 last year and tried this. I was able to unlock everything, but it resulted in artifacts and other issues that made games un playable.

Locking down makes sense... Business Sense

[Anonymous Coward]

Discluding the cards that might not be up to snuff to be sold to work consistently with 16 pipes on, it still makes sense to sell a "lower-end" card, if your aim is to make money, and not to help people that you don't know, and honestly are not that nice people anyways.

By selling a "low-end" and a "high-end" card, you can take the most money from everyone- Milk the guys that can afford it for all their worth, but still sell to the poor sods that still need to play Half-Life 2 at some overly-impressive benchmark.

This made sense before when the low and high end cards were different hardware, and it still makes sense now when the cost of manufacturing 2 different boards is higher than just making one and 'neutering it' to get two.

And I'm pretty sure it'll hurt sales. Not by any noticable amount, though since, come on, only an uber-nerd would really learn how to and then actually do this.

-Aylw

Re:Locking down makes sense... Business Sense

[Anonymous Coward]

Good point. As Max Horkeimer and Theodor Adorno wrote in their Dialectic of Enlightenment: "Marked differentiations such as those of A and B films, or of stories in magazines in different price ranges, depend not so much on subject matter as on classifying, organizing, and labeling consumers. Something is provided for all so that none may escape."

Re:Locking down makes sense... Business Sense

[VoidWraith]

It doesn't take an uber-nerd to unlock some things in RivaTuner.

alas

[atari2600]

Again, a lot of newbies posting on slashdot. VGA card modding is nothing new (ATI released moddable cards 9500-9700pro and other stuff) a while back. Just go google for them. Also, 200$ for a graphics card is not overkill - you get what you pay for.

Re:alas

[evilviper]

Also, 200$ for a graphics card is not overkill - you get what you pay for.

No, I get what you paid $200 for, for $40 a year from now...

Re:alas

[atari2600]

I paid for using the card for an year. I paid 300$ for a card an year ago which costs about 120$ now *shrug* - the card saw hell in the last one year thanks to Farcry, HL2 and other eyecandy games. If i like something enough, i am willing to pay for it.
An year from now, you will live one less year :) - doesn't mean the whole year didn't make a difference in your life.

Re:Two words...

[alexludd]

Are you sure of your facts?

this is a fault tolerence technique

[acidrain]

They manufacture the part with identical pixel pipelines, and if one of them is flawed they can just disble it. This is a common technique in silicon manufacturing. E.g. the celeron is a pentium with the flawed half it's cache disabled.

Flaws happen, and at say 20% rate per chip that is a lot of your profits. If you your design is redundant and can survive with parts disabled you can recover a lot of that 20%.

As another example the Cell processor has one SPU disabled in the PS3.

The flaws may not be visible in all games, or occur frequently. Thats why lots of people report the card working fine. The maker has better testing.

Of course it is possible that they also crippled a few that were just fine...

Re:this is a fault tolerence technique

[Anonymous Coward]

Less likely with nVidia. While Intel may indeed cripple a lot of perfectly P4s to meet Celeron quotas for OEMs, nVidia and ATI have absolutely no problem whatsoever selling 100% functional chips.

They simply don't have the pressures Intel/AMD do (ie; Dell saying "where are the million Celeron-D's we ordered!?!"). What they do have is a huge demand for the high end parts, stuff that is so tight they can overclock it 2% and sell it to complete boobs for $1000.

I've seen Radeon 9500's softmodded to 9700s, and they fail more often than a black kid at Princeton.

Note to scripters: the secret /. message is artxpmz!

Re:this is a fault tolerence technique

[wo1verin3]

>>Less likely with nVidia. While Intel may indeed
>>cripple a lot of perfectly P4s to meet Celeron
>>quotas for OEMs, nVidia and ATI have absolutely no
>>problem whatsoever selling 100% functional chips.

Not true in the case of ATI... I have firsthand knowledge regarding certain 3d functions in the radeon 8500 series (and several newer models) where we recieved very different benchmarks, and after pressing ATI it was explained that cards with defects had some functions disabled. Again, 99% of the public would never notice.

Re:this is a fault tolerence technique

[Anonymous Coward]

A big chunk of Nvidia and ATI sales go to the OEM market, so they are dealing with Dell the same way Intel is.

Consumer Sprinter

[Anonymous Coward]

"I understand why they do it, but if I'm paying for the hardware, I want to jolly well use it!"

Well you got MORE hardware for a given amount of money than those who spent more. Great! However all this means is that the manufacturers will now place their locks in inaccessable places, and still manage to hit their price points.

Basically the difference between a sprinter and a walker

Good Deal? You bet...

[Some_Llama]

This is actually one of the best found "upgrades" for video cards in recent past.

In order to enable the extra pipelines all you have to do is modify the Registry (in Windows) and if all of the pipelines are functional then it "just works". The great side to this is that if there are any problems witht he pipelines then you can just revert back to the original settings.

Previous mods like changing the Radeon 9800 pro into a 9800 XT required flashing the card with a different firmware to unlock the disabled features, or worse (like the old geforce4 to quatro mod) required soldering contact points on the card.

The first few batches of this card were pretty hit and miss ( and usually 75% miss) but as Nvidia refined their chipset manufacturing process more of these cards are actually using high quality chips that have fully functioning pipelines that have just been disabled to sell at the lower price point, so your chances of getting this "free upgrade" are pretty good (esp with certain models).

There is even a free tool http://downloads.guru3d.com/download.php?det=163
that gives a GUI interface that shows all of the pipelines, their status, and allows you to change them on the fly (you can change the settings back and forth but a reboot is required to take effect).

Test results...

[biode0]

Anyone else find their test results to be kind of odd? No ATi cards in the 3dmark05 benches. The lower end cards (9800 and 5900) achived results that are the complete opposite of every other test I've seen in Doom 3 and HL2. The test system was also fairly out of date, meaning the top end cards were probably somewhat limited. I'm not calling them liars, this stuff just seems kind of... iffy. (terribly sorry for the double post, I had problems logging in, not cowardace)

Does this work on all of them?

[B747SP]

A quick look at one online store here in Oz delivers about a dozen different AGP cards with this chipset. Does it work on all of 'em, or do I need to be picky?

Re:Does this work on all of them?

[Anonymous Coward]

This mod does NOT work with all cards. Anecdotal evidence sugggests that most cards will give you artifacts after modding. The submitter should have written "YMMV" in true Hot Deals fashion.

Remarkers Delight

[Hamfist]

How long before an enterprising individual buys several hundred of these, tests them to ensure that all of the pipelines work, and resell them as the higher end card?

Worst case scenario, you sell the 'defective' ones at face value.

Re:Remarkers Delight

[WhatAmIDoingHere]

Probably not that long. I hear nVidia already does that.

Re:Remarkers Delight

[N1ghtFalcon]

Can't - the higher models have 256MB of RAM :P

Re:Remarkers Delight

[be-fan]

No, NVIDIA does not do that. The GeForce 6800GT is a different card than the 6800, namely in that it uses more advanced (and more expensive) GDDR-3 memory while the regular 6800 does not.

Re:Remarkers Delight

[Anonymous Coward]

They can have 512 MB actually, but the 128 MB models are the only ones that have signifigant performance differences over lower memory cards..
More memory Isn't better, when wou have a problem keeping the memory on the card full in the first place..
Don't get me wrong, modern game engines can fill 512 MB of memory with textures etc, the problem is more with if you can retrieve it from system ram and have it on the card before the graphic card needs to process the texture, there is no actual performance gain from storing the texture in the card's video ram at all. 256MB and 512 MB cards don't actually 'earn' you any bonus performance for the extra money you spent on them.

Graphic card companies have been running a scam against the weak minded consumers ever since 3dfx went out of business... to make more profit, they invent ways to make certain cards seem better than others... and lately they've been 'crippling' the cheaper models to make the more expensive ones more attractive, because some people are starting to balk at paying $800 for a souped up toaster.

nowadays there is a $600 price difference in cards that realistically are only 1% faster, if that much.. in order to get people to pay the $800 so they don't go out of business they con them into believing they're actually much much faster.

Re:Remarkers Delight

[KarmaMB84]

Yes, they do, but they test the GPUs themselves for acceptable clock rate, defective parts, usable pipelines etc. Then they take a look at the chip stock they have, the demand for certain chips and go in and mark some down to meet demand (if necessary). Then they send them off to the manufacturers to get put on whatever cards with whatever memory they need them for. This is pretty standard procedure in the CPU and GPU sector.

Re:Remarkers Delight

[DrMrLordX]

There's also a different voltage regulator for the GPU on the 6800, 6800GT, and 6800 Ultra. I saw an article once comparing power consumption between the three cards, and the 6800 consistantly used less power than the other, more expensive cards, even when running the GPU and RAM at the same speeds(by way of OCing the 6800).

Re:Remarkers Delight

[be-fan]

I understand that, but if you actually read the thread, the "NVIDIA already does that" comment was directed at somebody asking "how long until dealers buy up 6800's, unlock the pipelines, and sell them as 6800GT's?" The thread was talking about complete cards, not individual chips. My point was that NVIDIA doesn't do that with complete cards --- the GT cards use different RAM as well.

Re:Remarkers Delight

[WhatAmIDoingHere]

You can't even READ... From the post I replied to:

"How long before an enterprising individual buys several hundred of these, tests them to ensure that all of the pipelines work, and resell them as the higher end card?"

Who mentioned "GT"? I sure didn't. He sure didn't. You're just

A) Wrong
and
B) Stupid

Thank you.

Re:Remarkers Delight

[be-fan]

Um, you didn't even understand the damn thing you quoted. It says right in the part you bolded: "resell them as the higher end card?"

There is no 16 pipeline version of the GeForce 6800. The 16 pipeline cards, the 6800 GT and the 6800 Ultra, are different cards that use not just a 16-pipeline chip, but also higher-end RAM. So the "higher end card" he referred to *is* the 6800 GT. There is nothing inbetween!

Re:Remarkers Delight

[WhatAmIDoingHere]

You're assuming that someone who resells something as something it isn't would call it something that already exists. What's keeping them from calling it a GS or a GR or a ST?

Fucktard.

Re:Remarkers Delight

[be-fan]

English is not your first language, huh? The OP said:

"the higher end card?"

The higher end card, not a higher end card. 'The' refers to things that already exist. If the poster meant the dealer could make up a GS or GR or ST, then he would have said "a higher end card".

you better start saving yesturday

[Neuroelectronic]

this card is projected to release at around $600

Did this

[Linthos]

I did this with my AGP GeForce 6800, and the extra piplines didn't work for me. They were damaged. Also you can unlock an extra vertex processor on it, which did work fine for me. I have read that it's about a 50-50 chance that the pipelines will work, as that is one of the reasons they are not sold as ultras. A reason they do work on some cards is that something else was wrong with the card that is also limited on the 6800 model, such as using less memory at slower speeds.

Re:Did this

[Kris_J]

Unless someone got the same info posted earlier, mod parent up as informative.

This makes a lot of sense. It makes both manufaturing and marketing sense, where crippling devices just to sell them cheaper doesn't make much sense at all.

Re:Did this

[KarmaMB84]

It does make sense if they can't sell them as the high end part but NEED the parts for the mid-range market.

Re:Did this

[Kris_J]

I guess you're right if we're talking about a part, not a completed product. But then why wouldn't you produce a "Special Edition" version with as many features available as possible? Is automated production so wasteful and company's opinions of their customers so low that they would rather cripple their products than produce the best they can produce?

(Of course it is, stupid question really.)

Re:Did this

[Mycroft_VIII]

Your talking about a LOT of special editions, not quite one per chip, but about one per hundred or so considering how many factors they test for and that not all of them are simple pass/fail (such as max stable speeds for parts).

Mycroft

Re:Did this

[Kris_J]

That's what makes them special!

It's not a mistake

[Neuroelectronic]

even after this information is released the hardware manufactures are still making out. If you mod your hardware in this way (assuming they notice) you void your warrenty. i'm guessing 90%+ of people will never do this hack anyway.

Sweet...

[N1ghtFalcon]

Went though with this using my Leadtek 6800 and it worked beautifully. Fired up X-Plane and don't see any artifacts at all.

Don't have 3DMark installed right now, so can't comment on the actual performance boost, but X-Plane did seem a little smoother to me (though that may be some physiological effects due to wishful thinking) :)

The only thing that's changed is the price

[Anonymous Coward]

This mod has been "out" for 6 months. There's also a voltage mod available, although it involves some soldering.

I'm losing faith in Slashdot. It used to be, like, current events and shit. Now, it's like, stuff that some MBA just figured out because it made directtv.com's homepage.

Come on. Get with it. wtf.

Go get your old PCI video card/couch leveler

[willisbueller]

checkerboarding? at least it's only a softmod, still though, I recommend keeping a PCI graphics card handy just in case.

Re:Test results...

[Bob of Dole]

Why would they benchmark an ATI card in order to show an improvement in nvidia performance?

Of course, however

[Sycraft-fu]

As time goes on the cards are getting better and better so it's more and more likely to work.

This trick is real common in the chip industry. If you get something that tends to have problems of partially, but not completely, failing, just disable part of it and sell it as a lower version. CPU makers do a similar thing with binning CPUs.

All of a given type of CPU come from the same assembly line. What they then do is test each core. Some don't work and get thrown out. The rest and rated as to their maximum speed and put in bins accordingly. You can then sell different models.

Thing is, as a chip is made more and more, there are generally less and less failures. However people still want a more budget card/processor/whatever, yet others are willing to pay for the fill thing. Well rather than stop producing the low end ones and mark down the high end ones, they just start using fully functional units in low end cards and disabling the features. Doesn't cost them anything more.

So, I've no doubt that this hack was real hit and miss at first, I'm sure there were very few fully function chips being put in to those as there were plenty of broken ones ot choose from. However since the cards ahve been out for quite some time now, I imagine that's been largely rectified, and I bet you have a pretty high success rate. No gaurentee of course, but a good chance.

I'm not sure

[Mad Ogre]

I'm not sure of the point of this... I mean, I've got a puncy little FX 5200 and I can already run Doom3 at the highest graphics settings in my system. I'm not kidding. What is a spendy card like this going to give me? Is there an unlock tool for the 5200?

Re:I'm not sure

[KillShill]

running it at a playable speed.

my 256 hercules cga card also runs doom3 at max settings... in ascii rendering mode.

at a stupendous 1 frame per week rate.

you've obviously have no idea what you're missing and my advice would be not to find out. you'll only be disappointed with what you currently have.

stay ignorant and happy.

Re:I'm not sure

[tomstdenis]

I call bullshit. I have an FX5200 as well. While I can play Doom3 with it [on an AMD64 at 2.2Ghz with 1GB of DDR400] it's not at "max settings".

It's playable but it does lag in certain sequences even on modest settings. A 6800 would certainly make for a decent improvement.

Though for my case the 5200 is enough power.

Tom

Re:I'm not sure

[bongo69]

Either you have incredibly slow reactions, your retarded or you like playing games in slow motion. I have an FX5200 and while I can crank it upto the full settings the game is unplayable (on both Debian Sarge and WinXP ). The Doom 3 setup auto-configures the settings to the lowest possible resolution on my machine with slightly better frame rates on my Windows XP partition than the Debian one. Medium is playable, but I occasionally notice lag.

Re:I'm not sure

[Bassman59]

"I call bullshit. I have an FX5200 as well. While I can play Doom3 with it [on an AMD64 at 2.2Ghz with 1GB of DDR400] it's not at 'max settings'.

It's playable but it does lag in certain sequences even on modest settings. A 6800 would certainly make for a decent improvement.

I agree. I thought Doom3 would run fine on an Athlon XP 2000+ (1 GB SDR DRAM) box with a 5200, and it does, but not at anything other than "okay quality" video. I get stutters when it tries to deal with very complex scenes.

"Though for my case the 5200 is enough power."

Yeah, I know -- for everything else (including WORK, like PCB layout and microcontroller code compiling and FPGA simulation/fitting), the machine and graphics card are more than adequate. Now, the dual 19" (not flat) CRTs are buggin' me, so it looks like a pair of 20" Apple Cinema displays are in my future -- but my graphics card doesn't have DVI out, so ... the upgrade cycle continues.

OK, Honest question...

[JawzX]

I know there are many like me out there who have become so used to reading Slashdot, and only Slashdot, that we are horribly out of touch with whats what in nerd news aggregators, What if any other options are there that cover as much (or nearly as much) general nerd goodness in a (mostly) non-repetetive and up to date manner?

P.S. discussion of articles is one of the features I've come to appreciate here on slashdot.

3D Mark 03 Score

[lahuard]

I got a 8769 after doing this :)

I did this

[Erpo]

everyone needs to remember that if the card worked 100% fine with 16 pipelines they would have sold it that way.

This is not always true. Some cards work fine with all 16 pipelines enabled, and nVidia might be locking 4 pipes on 6800 Ultra GPUs to make a volume target in the 6800 non-Ultra market.

However, sometimes the cards really don't work. I tried this trick, and it unlocked the extra 4 pipes yielding a significant performance improvement, but it caused lots of visual artifacts. I promptly set it back to normal.

if you had read TFA...

[KillShill]

you would have noticed a small (how coincidental) portion of the article which states how the number of artifcats went up a lot.

personally, even a single artifact in a single frame is too much.

basically, if you don't care about that then yeah you can have 4 more pipelines...

like increasing the clock speed of your cpu and only getting errors once every 3 minutes or so... and running SETI/important distributed project...

nice work, but spend the extra dough, you'll be glad.

Re:if you had read TFA...

[DrMrLordX]

I unlocked my eVGA 6800 and overclocked it a bit(up to 350 mhz GPU speed and 800 mhz RAM speed). No artifacts.

As with most overclocking and other mods of this nature, it's "hit or miss".

Same thing with a new BMW 3-Series

[voxel]

Except it hasn't been "unlocked" yet.

The new 3-Series 325i has the same 3.0L engine as the 330i car.

The only so far-noticed mechanical difference is a larger air intake for the 330i vs the 325i. Otherwise mechanically they are the same vehicle engine/transmission wise.

The real difference is the software, which restricts the 255hp engine down to 215hp for the 325i model.

As soon as someone plugs in a 330i ECM into the 325 ECM and puts on a cheap larger air-intake, you should have a nice 330Ci thats badged as a 325i...

Re:Same thing with a new BMW 3-Series

[fwitness]

Yep, all the Beamer-owners I know are cost cautious. They love to save a buck. Hell, most of 'em take the "BMW" logo off, I mean who cares about the a silly brand name? I bought this for the *performance* man.

Re:Same thing with a new BMW 3-Series

[TheGuano]

Actually, that's "bimmer."

The motorcycles are "beemers" (and OK, sometimes "beamers") but given that, it still doesn't include the 3-series.

Re:Same thing with a new BMW 3-Series

[evilviper]

It's quite sad that you got an "Insightful" mod, when your post is clearly sarcastic... Idiotic moderators...

Re:Same thing with a new BMW 3-Series

[mollymoo]

Since when have sarcasm and insight been mutually exclusive?

Re:Same thing with a new BMW 3-Series

[evilviper]

They aren't necessarily mutually exclusive, his post is simply (obviously) not insightful.

Re:Test results...

[biode0]

Same reason they showed them in every other test, more examples give a better picture of what's going on. I point it out because every other test featured ATi, yet this didn't. Some think 3dmark is a very good means of benchmarking, some think it's crap. Overall, it's atleast very repeatable.

linux?

[the_greywolf]

they've been doing this for over a year now, but...

i'd jump on this card if a similar technique could be achieved in linux.

so has it been done?

Re:linux?

[florin]

The method that the Firingsquad article describes is for Windows only, because it relies on the RivaTuner utility.

RivaTuner is nice because it provides a convenient GUI for playing around with different configurations easily, and I would definitely recommend going that route first of all, to find out which pipelines and/or vertex shaders can be enabled on your particular card.

After any configuration change one should test it thoroughly with a bunch of shader intensive 3D apps (like HalfLife 2, Doom 3, 3DMark03/05 and the rthdribl demo). If you don't get any artifacts or stutters then you can use the info you got from the RivaTuner method for another one which works in Linux too.

See instead of using RivaTuner, you can also unmask pipelines and vertex shaders directly in the BIOS of your 6800(LE/NU/GT/Ultra) card. The advantage is that they will be enabled as soon as your card powers up. As such they'll work in any OS you boot, and they'll keep working if you move the card to another PC.

Unfortunately I don't really know about any nice walkthroughs for the BIOS method but have a look at this page (click on the link for 'iron cobra') for some pointers. If you're confident enough to feel that you should be trying this then that should be all you need. It works fine for me.. and in Linux too.

Re:linux?

[l_bratch]

There is indeed a softmod for this available for Linux, I found it through the Gentoo Forums: http://forums.gentoo.org/viewtopic-t-310871.html

Just tryed it

[Anonymous Coward]

I just finished modding my 6800 and it worked perfictly significant performance increase in HL2. Only took like 5 mouse clicks for like 15% increase.

Wow

[Mad Ogre]

Some of you are pretty hostile. Yeah, my box can run D3 at max settings and do pretty well. It can get a little laggy in some stages... never said it didn't. But then again I didn't think this was a pissing contest. I guess I'll have to get a better card and see for myself. I mostly build business machines, so I don't play many of the new games. Just thinking about GTA:SA coming out and wondering if I should upgrade for it... but I'm not seeing much reason to. Vice City runs just great....

right on

[Penguinoflight]

I also did a softmod on the vanilla 9500. It worked great, and I sold the system to a friend. The 9500 and 9500 PRO were actually different boards, however the 9700 and the 9500 were the same.

As I remember it, the mod was first tried when someone in europe (thinking Germany) spotted the one difference between a 9500 and a 9700, one solder point. They changed the solder point and their 9500 was a 9700.

Someone made a driver that ignored the signal the card sent to identify it's model, just assuming the model to be a 9700. A lot of the cards worked too, it sure kept the 9500 vanilla above the cost of the 9500 pro for a while.

Re:right on

[Anonymous Coward]

nice sig. Where is that dude today if he cheated death?

old news

[Anonymous Coward]

Man! This is old, OLD news. I was doing this with Rivatuner 15.1 RC2 in August of this year. Nothing to see here, move along.

typo

[Anonymous Coward]

typo in your sig.

Better off modding an X800Pro VIVO

[Anonymous Coward]

Can be modded to run as and sometimes faster then the X850XT-pe..

DUDE! THAT GIVES ME WOOD!

[Anonymous Coward]

I'm not sure why, but your post has given me wood. Can you elaborate, or something? Maybe just a little bit?

What all would be involved in hacking the car's BIOS? And swapping out in intake? Would the intake just involve a simple cut and weld, or would I need to modify the engine block?

How about the BIOS? I'm guessing I'd need to completely change the fuel/air ratio, along with the octane sensor. Maybe the EGT sensor.

I've totally still got wood, mate. Talk to me!

Apple

[ylon]

How can this be accomplished on Mac OS X on a Power Mac?

Hardware Hacks

[Anonymous Coward]

This reminds me of the hack for Promise Ultra66 controller cards a few years back. You could run a jumper and turn it into an Ultra66 RAID card. I have a 6800, but its PCI-E, so this hack did nuthin for me.

How is this NEWS?

[PaternityTest]

Seriously I did this 6 months Ago when I bought mine in november.

This fix is not for you!

[Anonymous Coward]

This causes major graphical glitches even in windows with many of the other GeForce 6800 cards. They probably shut down the 4 pipes because some quality control reason for all we know. I tested this on a Jaton 6800 just now myself.

Warranty...

[CreateWindowEx]

...messing with a $200 video card is one thing, but voiding the warranty on your brand new $30,000 car is another, especially a very high-tech new model. Also, I believe it's not just the intake, but also the intake manifold, MAF sensors, throttle body, and airbox.

It will be interesting to see what the various tuners do with the E90 325--presumably the big guys like Dinan who enjoy a close relationship with BMW might get some pressure not to release a cheap 255 hp upgrade. The VANOS systems are supposed to be very hard to modify--it may turn out to be non-trivial to make the change without inside information.

Yay Canon!

[Cryptnotic]

It didn't turn the plastic on the 300D into the magnesium alloy of the 10D. And it didn't change the color from that lame champagne to cool black. And it didn't move the status LCD from the back up to the top. And it didn't add a wheel for changing settings with your thumb instead of using the lame plastic buttons. The 10D was still a much more durable camera and was more comfortable to use. The 300D was just cheaper. The disabled features were things that the hardware should have been able to support (flash exposure compensation and using any auto-focus mode in any shooting mode). Those things were enabled using the firmware hack.

Re:Yay Canon!

[dlZ]

As someone who uses a 10D on a daily basis, I completely agree. It's a very sturdy camera, which goes almost everywhere with me. I would have smashed the 300D into little pieces by now, I'm sure. Not that I abuse my cameras on purpose, but I do use them in a lot of different situations, and have had them knocked into things by people rushing by, or going through crowded areas, etc.

Re:Yay Canon!

[Cryptnotic]

Good luck on your non-smoking. It gets easier and easier. Keep yourself from situations where you'll be smoking. Maybe go on a diet and exercise more too. Go all out, some of it will "stick", as they say.

Not LAME enough.

[Doc Ruby]

Someone's got to come up with Linux code for these cheap supercomputers that runs multiple simultaneous LAME MP3 encoder processes.

Now lets see them unlock

[orbitalia]

the broken VPU on the AGP 6800 cores. currently 2 million transistors on every AGP version ofr 6800 GT/Ultra are disabled due to a design failure. Quite annoying when they sold these cards with that feature and one of the reasons i bought it was the hardware decoding of WMVHD streams or so I thought. Now they expect you to buy a software kludge (pure video) to enable some parts of this hardware that you have already paid for.

NVidia really are testing my patience, I think it will be ATI next time for me.

Re:Good Deal? You bet...

[ameoba]

I got one back in november (day after thanksgiving sale) for $250, back when vanilla 6800s were still over $300. I managed to unlock (with rivatuner) and overclock it (using the overclocking features in every nVidia driver) without any problems. It only took me about 2hr to find the maximum stable overclock.

Unlocked and with a relatively modest overclock (5 vertex pipes/12 pixel pipes @ 325MHz core/700MHz memory to 6/16@380/820; some cards go over 400/900 on stock cooling) I managed to take myself from 9000 points in 3dmark03 to 11,000 points. You can say what you will about synthetic benchmarks but I am seeing about 20-25% better performance in "real world" tests.

Yes, it's a bit of a gamble, hoping you get a card that unlocks and/or overclocks, but the odds are pretty good - even without unlocking successfully, most of them get decent overclocks. With 6800s selling for just over $200 now, nothing else near price-point even comes close to the ammount of potential "free" performance you can get out of them.

Willy-waving

[Moraelin]

You have to take all overclocking claims with a bit of salt, because for some people it's like the size of their penis depends on it. They'll be... very creative and selective in what they tell you, and that's putting it very mildly.

I've briefly been into the overclocker willy-waving scene myself, so you can take that as an admission. Guilty as charged, guv'nor.

Anyway, I've played with it long enough to know that there very rarely is a hard point where the card works 100% flawlessly, and 1 MHz higher it just locks up. There's more of a gradient grey zone where the card sorta works enough to finish one particular benchmark, but glitches, is unstable, or eventually overheats. And where it might work at that frequency in one game or benchmark, but lock up hard in 20 others.

The big overclocking brag-fests you read are usually from this grey area, not from the 100% stable zone.

Yes, you see some screenshots of a mondo 3DMark number there or of some utility showing the card running at 4 gazillion megaherz, but what you don't see is that it runs stable only for the 10 minutes needed to finish the benchmark. After that it overheats and starts artefacting, or outright locking up.

Be even more suspicious of brag-fests where they only ran half of 3DMark, and hand-waved the other tests as "bah, they didn't make much of a difference on the score anyway." (Ever notice how the biggest overclocking claims fall in that category?) Usually it means it crashed or locked up in those tests.

So I wouldn't take those as a baseline or as "_all_ 6800 cards make it that high with no problems, and it's just the mean MBAs at Nvidia marking them down." Fully expect that any card you buy might not be quite stable that high.

Which brings me to another point. To paraphrase another saying "overclocking gives you something for 'free', if your time is worth nothing." Because in the end the price you'll pay is a lot of time tweaking and testing that overclock... for each new game you buy, time replaying 30 minutes worth of something _again_ because the card locked up just before the save point, etc. It can end up a passtime in and by itself.

Re:Willy-waving

[Anonymous Coward]

I've briefly been into the overclocker willy-waving scene myself, so you can take that as an admission...

Anyway, I've played with it long enough to know that there very rarely is a hard point ...

Whoa there.. are you talking about your willy or graphics card?

Re:Willy-waving

[Anonymous Coward]

Anyway, I've played with it long enough to know that there very rarely is a hard point ...

Whoa there.. are you talking about your willy or graphics card?

ROFL!

Both

[Moraelin]

Oh, I've played with both. The willy had more replay value, but I think people are more interested in the graphics card here ;) Or with CPUs for that matter. Had a 450 MHz K6-III which, when overclocked to about 500 MHz, would process about a SETI packet per minute. Lemme detail that: its FPU glitched hard, produced only garbage for any floating point operation, but zipped through SETI packets at surrealistic speeds. Again, producing just garbage as results. But, boy, did it churn garbage fast. Quite a useful glitch that, if I were dishonest enough to post that as proof of my l33t overclocked performance. Could have probably claimed I had some liquid-nitrogen rig cooling that rig :P

Yeah, and bombing Iraq is "spreading democracy"

[Run4yourlives]

Seriously, although it certainly makes sense that Manufacturers compensation for flaws, they've got one hell of a crap production line if their flawed products outnumber their good ones.

Think markets here: You'll sell a lot more mid range cards then high end ones because let's face it, not everyone is stupid.

Saying that, some of those mid-range cards are defective to the point where they can't be high end, but a large majority are perfectly fine.

So, you design you production line to make the high end cards, overprice them to cater to the high end market, and price disabled, mid-range cards so you still make a profit and volia: daddy get's a new ferrari!

(Actually, the profits from high end cards probably got back into R&D to cover the next gen of high end cards... the company survives day to day on the regular stuff I'd bet)

ATI has done it too

[Baikala]

My (now very old) Radeon 9600SE turnet out to be a 9600PRO without fan. The newer drivers even set the clock to 400 Mhz by default (the 9600SE chip is 325 Mhz). They probably test some samples and when QA fails they jut box the whole batch of cards in a lower model box.

Re:Good Deal? You bet...

[Some_Llama]

I also got one around that timeframe (the BFG model) but had artifacts when unlocking the pipelines, still even without the extra pipes it is a great card.

In fact I just bought a second card (got it for 200 on sale) and am excited to try unlocking this one since all of the original chips from first productions should have been used up by now ;)

Works for me

[dacaffinator]

I unlocked my MSI AGP 6800 last night and tested it by playing Counter Strike Source for a few hours. I didn't experience any visual artifacts whatsoever. I didn't do any before and after benchmarking so I'm not sure how much of a performance gain I got.

fuses

[hawk]

on-chip fuses that control which units are active. after the device is tested and assigned an identity, blow the fuses for the unused portions.

This could also have been done to fight chip counterfitting--fuses inside for the possible speeds, and blow the fuses for speeds that it didn't test as (or wasn't sold as). Alternatively, a bit on the case that is broken off to indicate the speed, with more broken off indicating lower speeds.

not even close

[hawk]

Just off the cuff, calculator chips in the 70's used to be used for several models, and articles ran in print magazines such as popular electronics explaining how to add buttons to the case to get the ectra functions.

in at least the 60's, IBM had field upgrades that amounted to removing a plug that disabled alternate clock cycles.

And I'm sure that there are even older examples.

Re:not even close

[Kazymyr]

Well in the '70s I was barely learning to read, and I wasn't around in the '60s so I have no experience with those. :)

But I did personally experience the 486 marketing ploy.

Re:not even close

[hawk]

True, I read the calculator articles as they printed.

However, IBM was doing this before I was born.

History--a wonderful thing :)

hawk

Re:not even close

[Kazymyr]

Don't get me wrong - I appreciate geek history as much as the next... geek. I just never happened to come across this particular bit before. Thanks for sharing!

Re:not even close

[hawk]

Ahh, but what other use has knowledge?

:)

hawk