Slashdot Mirror

I would argue just the opposite. L2 cache is less important in games than other applications. Do you remember the first Celerons with no L2 cache on die or anywhere else?

They benchmarked very poorly in the typical ZDnet benchmarks for Windows. They also performed pretty well in games such as Quake. Here are some benchmarks from Tom's Hardware:

Office Performance Windows 98
3D Gaming Performance
Notice the difference between a 300 Mhz Celeron with 128k L2 cache and one without.

in Quake2:
with L2 cache: 50 frames per second
without L2 cache: 45 frames per second
pentium 233mmx: 32 frames per second

in Business Winstone:
with L2 cache: 25
without L2 cache: 19.3
pentium 233 mmx: 19.5

Granted this is only applicable for Quake2 (and the other quake engines more than likely), it still indicates at least one game doesn't depend very heavily on L2 cache. I think the generalization to be made is that L2 cache boosts integer performance moreso than floating point performance.

Check out the Coppermine vs. Athlon benchmarks you were referring to. The ones I've seen indicate that the on die L2 cache helps the Coppermine catch up to the Athlon in integer performance, but not in floating point.

I would argue just the opposite. L2 cache is less important in games than other applications. Do you remember the first Celerons with no L2 cache on die or anywhere else?

They benchmarked very poorly in the typical ZDnet benchmarks for Windows. They also performed pretty well in games such as Quake. Here are some benchmarks from Tom's Hardware:

Office Performance Windows 98
3D Gaming Performance
Notice the difference between a 300 Mhz Celeron with 128k L2 cache and one without.

in Quake2:
with L2 cache: 50 frames per second
without L2 cache: 45 frames per second
pentium 233mmx: 32 frames per second

in Business Winstone:
with L2 cache: 25
without L2 cache: 19.3
pentium 233 mmx: 19.5

Granted this is only applicable for Quake2 (and the other quake engines more than likely), it still indicates at least one game doesn't depend very heavily on L2 cache. I think the generalization to be made is that L2 cache boosts integer performance moreso than floating point performance.

Check out the Coppermine vs. Athlon benchmarks you were referring to. The ones I've seen indicate that the on die L2 cache helps the Coppermine catch up to the Athlon in integer performance, but not in floating point.

Ummm, you seem to be mis-informed. AMD's Dresden MFG facility is more than capable of producing all the CPUs they need now and for next few years. VIA should have no trouble making their chipsets for them. I don't know about AMDs past production, but I know right now they aren't having any problems. Intel is, plain and simple. Check Tom's Hardware for some more info on the problems Intel is having, not only with CPU production but with thier i820 chipsets too.

The reason is that Intel is scared shitless of AMD. Intel knows that AMD can best them in the MHz (GHz?) race at any time, and they know that AMD has 1GHz chips. They also know that the Athlon beats the pants off the old 686 core of the Pentium III (which is really only a Pentium Pro that went uptown).

The fact of the matter is that Intel is scrambling to keep its mindshare, so it makes big news about things that will happen six months from now. People that trade stocks and make PCs now have it in their heads that Intel actually has a 1GHz system, and that they were the first ones to break the GHz barrier. Those people forget about AMD and the K7. That's the really Intel's strategy: keep announcing things that aren't here yet so the spotlight never strays too far, even though the PIII is inferior. Make people forget about that "other" chip company.

But don't take my word for it, no. Go try and buy an 800 or even 750 MHz PIII system. Then go shop around for an 850 MHz Athlon system. AMD announces things when they happen, like a company should. Intel is the hardware equivalent of Microsoft and I hope their subterfuge and bully tactics (look through Tom's Hardware for articles about Intel and K7 motherboard manufacturers for a little info about friendly old Intel) come back to bite them in the ass.

-B

I've always been one to buy Intel, but I'm thinking my next box will probably have an AMD chip. I expect AMD will have matching or faster chips within a month or two.

You all might want to have a look at The Register and Tom's Hardware to get better information about this story. Both sites have been publishing some interesting stories (specifically The Register) concerning this chip race that we all love watching. They seem to do a good job of getting the dirt on both AMD and Intel, and although they do seem to play favorites with AMD (as I'm sure most of you do as well), they have posted relatively unbiased articles about both companies. When I say unbiased, you have to look at all of their storys, as both sites have taken to Intel bashing as of late, and for good reason.

Intel is still having production problems with their 800Mhz PIII, whether they will confirm that is another matter. The Register contacted a German distributor and he said that he isn't getting the 800Mhz chips, let alone something running at 1Ghz. Also, a story on The Register a couple of days ago said that Intel had told manufacturers to expect 866Mhz's by the end of February. That obviously hasn't happened.

I find it hard to believe that Chipzilla is going to be able to jump 200Mhz in one month. Maybe they'll take a prototype, tape an ice-cube to it and ship it to Dell so they can say they did. You may see these chips in volume by June, but I sure wouldn't believe this information coming out of Chipzilla now...

Here's an article at The Register that explains it nicely.
And there's a stable 1.1GHz Athlon that was shown at CeBIT. Tom's Hardware has more.

IMHO, the question of when the consumer can buy this chip and what quantities they'll be available in will determine who really was the first to release a 1GHz chip.

My personal choice: I got a board capable of dual Intel processors, but only put one Celeron 333 in it. I figure if I get really antsy for CPU speed, which hasn't been a problem yet, I'll either get a much faster single chip or buy a pair of moderately faster matched chips. The board was $235 about a year ago and was one of the cheapest dual-capable boards that supported 1 GB RAM.

Good luck with whatever you decide, and don't forget to make a web page describing your achievements!

While I don't think that this is the death of Intel, for they have too many fingers in too many lucrative pots, it suggests that they have misstepped badly with RAMBUS are are going to lose their market dominance in CPUs if AMD keeps it's act together... which other than Irongate and a scarcity of Athlon MBs, they have been doing fairly well.

As far as Irongate's AGP issues, there isn't a whole lot of difference performance-wise at this point between AGP 1x and AGP 2x. Maybe in the next iteration of video cards we'll see a more significant difference, but I'd rather have a CPU that is 15%-20% faster than sweating about a 4% hit on the AGP bus. Some people feel that Athlon is not being entirely honest & ethical with the issue, which may taint their reputation in the long run.

My next machine will be an Athlon based system. I've suffered extreme technolust since they were released, and they just get better and better.

It will be quite some time before Intel has anything in market to compete with Athlon, and by that time it might be too little too late. Their most recent efforts have yielded uncertain results in comparison to the Athlon.

For once, the US Gov't's own stupid laws can work for us. If they'd intelligently removed the arbitrary 56-bit limit, then we'd have a much tougher beast to deal with. However, consider this:

The keyspace is only 2^56 in size - the same size as RC5-56. Remember, that algorithm that distributed.net killed a year or so ago? Now, Moore's Law (and Tom'sHardwareGuide) say that our collective computing power has increased by a few hundred percent since the start of that contest.

So, let's launch a new contest, then, except this time we'll have:

1. More willing participants (you directly benefit from the results!)
2. Much, much faster equipment
3. A keyspace that is only 2^56/n, where n is the number of monitor vendors who've been issued unique keys.

In any case, it should only be a few months until we could have the decryption keyspace entirely mapped.

Now, is that sweet irony, or what? God bless our Congress!

The Tom's Hardware Guide article is a must read. I don't agree with all the points he tries to make, but there is a wealth of insight into all the things AMD and Intel have done right and wrong to get us to this point.

And this CPU demo seems to fit right in with what he was saying.

Over the next few months, other rumors (all undoubtedly from "reliable sources") will be published suggesting that Intel's next generation "Athlon killing" processors are only a few days away, yet until the Willamette is released no sooner than October, Intel will have nothing new to offer.

... the eigth page. Then the FUD hit the fan. After all they're talk about Intel vs. Athlon, (which I particularly enjoyed, being an Athlon owner), the Author had to go on and prove how little he actually knew about the future of computers. Page 8

According to the author, the Microsoft X-Box will not only be more powerful then the PSX2, but will also be cheaper! That's absolutely amazing! According to every other article I've read, the X-Box is supposedly less powerful and over twice as expensive. He mentions that the X-Box is due to use an advanced "chipset from Nvidia" but last I checked (yesterday) nVidia didn't have any processors that could come close to the PS2s raw computing power. In fact, the next generation of 3-D Graphics chips on the PC market are set to still run approximately 1/10th the polys/second of the Emotion Chip, and lets not even get into the Gigapixel ratings between the two.

He then goes on to talk about how the Dreamcast is an "X-Box Lite". Apparently because it runs CE the Dreamcast is super-sweet. But not only that, he suggests that because the DC runs CE and contains DirectX explains why there's so many titles already available for the DC.

Could you please show me where they are? I can't seem to find any good titles except for three or four that shipped back in September. When is the DC going to be releasing the 100's of titles they promise and push back every month?

Its this kind of idiocy in reporting that makes me absolutley HATE the media.
--
"A mind is a horrible thing to waste. But a mime...
It feels wonderful wasting those fsckers."

Uhhh... I'm getting so tired of this. No RAM technology is NOT becomming a bottleneck. Tom (www.tomshardware.com) tested p3-600MHz systems with standard pc100 SDRAM and 400MHz Rambus. Well, the Rambus systems were about 5% faster. But considering also that they had newer chipsets and, most importantly, on-die full-speed cache (as opposed to halh-speed cache of the old P3-600), one has to wonder what exactly caused the speed increase. Also, considering that Rambus is 6 *times* more expensive then SDRAM, the 5% speed increase does not look attractive at all.

___

Tom (from Tom's Hardware fame) just released a review of this card (DDR & SDR comparison), specs, and how it stacks up to the competition (G400 and MAXX). Worth a look here:

http://www6.tomshardwar e.com/graphic/00q1/000107/index.html

The gamers are brutally disloyal. They always hunger for the fastest biggest and baddest. They'll switch to an Athlon in a minute if they know the chipset will give them a monstrous speed advantage without an overly outrageous price tag. Which is to say if the coppermine is superior to the athlon by a maximum of 5fps in laboratory conditions, but the system costs $1000 more (an excellent real world analogy for this would be the RDRAM vs DDR SDRAM comparison), they'll go for the Athlon!

The Athlon only needs to stay close, or a breath ahead, of Intel, while keeping its prices down, down, down. The gamers will lead the exodus, and as the .18 micron chips roll out, DDR SDRAM comes into play, and AGP4x support becomes standard, all the vendors will go to bat for AMD.

When the dual boards come out, Intel has instantly lost the whole Linux community over the Chip ID issue. Linux people will have an economic and performance motivation for dumping Intel, plus the anti big brother aspect will weigh in.

A suggestion to AMD: advertise your chips with "No Big Brother Inside" ads and hit Intel down below the belt. It'll cause controversy galore, but you're going to score a lopsided public relations victory in the process.

peace!

Kryotech is selling the Super G, a 1000mhz Athlon running at a temperature of -40C.

However, your problem is not the chip speed, but the surrounding bus speed and RAM speed. The chip runs at 1000mhz, but the Front Side Bus is only 200mhz and the RAM runs at 100mhz. YUCK! Talk about your kinks in the hose!

The things you'll want to address before you improve your cpu speed are:
1) RAM speed. You'll theoretically wanna blast open that blockage in the pipeline with 800mhz RDRAM, although it has been shown that this does not deliver anywhere near the astronomical improvement that it is meant to deliver.
2) Cache speed. It runs at half the speed of the cpu - or, almost half. This needs to be rectified ASAP.
3) AGP support. This is very important in games. I predict AGP4x will really unleash the speed demon in every computer.

Is it any wonder that with all these bandwidth limitations, the Super Bypass only yields a 2-5% increase in performance over a non Super Bypass board?

RDRAM performance only gives you a small improvement over systems with PC133 RAM, according to the latest Windows general and game specific benchmark tests at Tom's Hardware. In the windows general test, the point spread between the i820/PC133 RAM and the i840/RDRAM is something like 25 points, with the lowest scoring 321 on the BAPco SYSMark98, and the highest at 343. In the game test, the i840 scores 122 in Q3 Arena (640x480x16), while the i820 scores 111.6. Hardly a mind boggling improvement.

You are not going to notice that difference in a game. But you are going to feel it quite long, hard and wide in your pocket book. A whole lotta spankin' for nothin'. You can get an Athlon 750 for less, and the performance in real time is either superior, or not far enough back behind the Coppermine to justify the cost of the Coppermine/RDRAM combo.

Likewise RDRAM + the Athlon won't mean anything, either. Athlon's support of AGP4x will mean a lot, however.

RDRAM performance only gives you a small improvement over systems with PC133 RAM, according to the latest Windows general and game specific benchmark tests at Tom's Hardware. In the windows general test, the point spread between the i820/PC133 RAM and the i840/RDRAM is something like 25 points, with the lowest scoring 321 on the BAPco SYSMark98, and the highest at 343. In the game test, the i840 scores 122 in Q3 Arena (640x480x16), while the i820 scores 111.6. Hardly a mind boggling improvement.

You are not going to notice that difference in a game. But you are going to feel it quite long, hard and wide in your pocket book. A whole lotta spankin' for nothin'. You can get an Athlon 750 for less, and the performance in real time is either superior, or not far enough back behind the Coppermine to justify the cost of the Coppermine/RDRAM combo.

Likewise RDRAM + the Athlon won't mean anything, either. Athlon's support of AGP4x will mean a lot, however.

RDRAM performance only gives you a small improvement over systems with PC133 RAM, according to the latest Windows general and game specific benchmark tests at Tom's Hardware. In the windows general test, the point spread between the i820/PC133 RAM and the i840/RDRAM is something like 25 points, with the lowest scoring 321 on the BAPco SYSMark98, and the highest at 343. In the game test, the i840 scores 122 in Q3 Arena (640x480x16), while the i820 scores 111.6. Hardly a mind boggling improvement.

You are not going to notice that difference in a game. But you are going to feel it quite long, hard and wide in your pocket book. A whole lotta spankin' for nothin'. You can get an Athlon 750 for less, and the performance in real time is either superior, or not far enough back behind the Coppermine to justify the cost of the Coppermine/RDRAM combo.

Likewise RDRAM + the Athlon won't mean anything, either. Athlon's support of AGP4x will mean a lot, however.

Also, does this mean we'll see a Kryo CoolAthlon announcement for 1.066 GHz? (upward multiplier is 133%)...drool drool....

Yup, and you can read about it here on Tom's Hardware.

numb

Me too. I've got an Athlon 500 at home, which, with the TNT2 Ultra I've got, feels so incredibly fast at every game I've tried...

And I'm buying two more. Doing my part to support them.

I don't understand why people would be buyin Intel chips - okay, Athlon motherboards are a bit more expensive and as far as I can tell, none of them support AGP 4X, but AGP 4X hasn't proved to be a big win anyway (and not many Intel boards support it). And after you pay for the Intel chip, the cheaper board doesn't win you anything.

I've got an Asus K7M, btw, which is a great board. Overclocking options right in the bios - woohoo!

(Oh, and check out this article, which basically says that you can add about 5% to all the benchmarks you've seen if you buy a newer motherboard or are lucky enough to have the newer version of the chipset that supports Super Bypass).

I would love to tell you the real story here but I don't think these hardware companies would appreciate it

Tom lost all my respect. And if you feel you need to read the surrounding text, look at the URL below. Wow. http://www6.tomshardwar e.com/graphic/99q4/991230/fury-14.html

Yeah, ASUS has released a K7 mobo, albeit somewhat surrepetitously, called the K7M. Tom's hardware does a review of it and 5 other K7 mobos HERE.

They're selling these mobos in, get this, plain white boxes with nothing identifying ASUS as the manufacturer. There's also NO info on their website on the K7M. They're officially releasing the motherboard under a dummy company called, I think, "Freeway". Their site search turns up nothing on the K7M, but Tom's has a link to a hidden place on ASUS's site that has K7M info.

I don't know who they're trying to fool with these antics, but I'm not impressed by either ASUS or Intel's strong-arm tactics. In fact, I'm calling shenannigans on this whole deal. SHENANIGANS!

With the K7 out, and multiprocessor K7 motherboards due out in 2000, there won't be any reason to buy Intel ever again! Death to Intel! Death to Microsoft!

Another thing to watch out for is that RAMBUS RAM is 5x or more to that of standard SDRAM/SGRAM prices. Tom's Hardware did a review of a new DDR board by Micron (called the Samurai) that shows equal promise with much less expensive RAM. unfortunately it is not available yet...but I have heard that the potential for US Gov. funding tends to make prjects "move faster" :). The review is here There is a discussion of what DDR means at the link. This still doesn't get you away from the Intel chips though. Hopefully with luck someone else will pick this up for the Athlon before long. (I drool at the thought of a DDR Kryotech K7...)

According to Tom ddram outperforms rambus anyway, due to rambus's inheritantly worse latency. Also if reliability is a concern :-) keep in mind that a slew of Rambus-related problems have dogged the new rambus motherboads/chipsets.

Kryotech's 1GHz Athlon system, "SuperG", is already shipping. Tom's hardware reviewed it more than a month ago, and it was very stable, and very, very fast. To quote Tom's:

Price and size aside, this unit will go down in the record books as the first consumer PC to hit the streets at the magical 1 GHz. Although 1 GHz doesn't really do anything special, it has been a number that most of us have thought to be a leap beyond. I can only imagine what we'll do when get our hands on a .18 Athlon to slap into the SuperG. If you've got the cash floating around and you want the best, Kryotech is undoubtedly your source one for the fastest PC in the world.

Sorry, it seems that my links didn't work correctly so I'm reposting
From everything I've heard (two places IIRC) Tom's Hardware and Gamespot a PIII with the 820 mobo and RDRAM actually runs faster than an equivalently clocked Athlon. Now, yes it is running a different memory, but as far as I know, Athlon doesn't support that memory (at least right now.) And the memory is VERY expensive (~$700-$1000 for 128MB), but if you are going for the FASTEST possible at this moment, I think it goes to INTEL. Another slight hit towards Athlon is that it isn't stable in all configurations. (according to gamespot)
Now I have to say that for some odd reason I feel biased towards Intel even though they are the giant. Maybe it's because AMD has underperformed in the past.
However, I will say that I'm happy AMD came out with their Athlon. Competition can only be good, and if they can take the crown, it will mean that I can buy an even faster processor. I don't have any problem with that.

Like the chipset, right?

Tom's Hardware recently ran benchmarks of Intel's chipsets. What I got from this (which may be wildly inaccurate :) is that the i820 (which is probably what you'll get if you buy a P3) is in many cases no better than the 440BX. In order to get a real performance boost, you'll need the i840, which Intel isn't yet releasing in mass quantities, and they're marketing it as a workstation (not desktop) chipset. hmm.

anyway, I've got a P2-350 and don't plan on upgrading anytime soon (and it'll probably be an AMD when I do. :)

When the Athlon was first released, Toms Hardware did a very detailed write up on the architecture of the Athlon, and how it relates to the PIII.

You can find that article here. To summarize, this advancement from Intel is basically irrelevant, as AMD could (probably) have 1GHz Athlons on the market already, the Athlon is designed to run at those kinds of speeds, and has a (in their opinion) dramatically superior architecture to the PIII.

Tom mentioned somewhere in the article that AMD would probably do to Intel what Intel had been doing to them for years, which was to one-up whatever speed they come out with. Watch for AMD to beat this by 50MHz or so pretty soon...

Anthony

When the Athlon was first released, Toms Hardware did a very detailed write up on the architecture of the Athlon, and how it relates to the PIII.

You can find that article here. To summarize, this advancement from Intel is basically irrelevant, as AMD could (probably) have 1GHz Athlons on the market already, the Athlon is designed to run at those kinds of speeds, and has a (in their opinion) dramatically superior architecture to the PIII.

Tom mentioned somewhere in the article that AMD would probably do to Intel what Intel had been doing to them for years, which was to one-up whatever speed they come out with. Watch for AMD to beat this by 50MHz or so pretty soon...

Anthony

But let's get back to the magic '256'. I could hardly believe my ears when I was finally told what the '256' stands for. NVIDIA adds the 32-bit deep color, the 24-bit deep Z-buffer and the 8-bit stencil buffer of each rendering pipeline and multiplies it with 4, for each pipeline, which indeed ads up to 256. So far about the fantasy of marketing people, they are a very special breed indeed.

Tom's Hardware did a Athlon vs Coppermine benchmark on Oct 23. Here are the results:

First of all the i830 chipset wasn't available at the time so Tom used a Apollo Pro 133+ mb.

The results of the first benchmark, SYSMark98 under Win98SE, showed that the Athlon had %4 better performence then the Coppermine at all clock speeds.

The results of the second benchmark, SYSMark98 under WinNT4 showed about equal performence.

Under 3DStudioMax under WinNT SP5 showed that the Athlong was %30 faster then the Coppermine.

Under Quake 3 Arena under Win98SE the Athlon was about %6 slower.

Under Descent3 under Win98SE using OpenGL both chips were about equal in speed.

With Descent3 on DirectX the Athlon was about %3 slower.

In all the Athlon almost always did a bit better then the Coppermine. With the exception of the 3DStudioMax benchmark where the Athlon badly beat the Coppermine.

As for pricing the Athlon is about $60 cheaper.

You can see the full benchmarks here.

Tom's Hardware did a Athlon vs Coppermine benchmark on Oct 23. Here are the results:

First of all the i830 chipset wasn't available at the time so Tom used a Apollo Pro 133+ mb.

The results of the first benchmark, SYSMark98 under Win98SE, showed that the Athlon had %4 better performence then the Coppermine at all clock speeds.

The results of the second benchmark, SYSMark98 under WinNT4 showed about equal performence.

Under 3DStudioMax under WinNT SP5 showed that the Athlong was %30 faster then the Coppermine.

Under Quake 3 Arena under Win98SE the Athlon was about %6 slower.

Under Descent3 under Win98SE using OpenGL both chips were about equal in speed.

With Descent3 on DirectX the Athlon was about %3 slower.

In all the Athlon almost always did a bit better then the Coppermine. With the exception of the 3DStudioMax benchmark where the Athlon badly beat the Coppermine.

As for pricing the Athlon is about $60 cheaper.

You can see the full benchmarks here.

If these comparisons largely depend on the graphics card (and possibly its drivers) instead of the processor, then you are not getting the whole picture.

If these comparisons depended largely on the graphics card, then they would show the same results at all processor speeds. Which they didn't. Ergo, the comparisons don't depend largely on the graphics card. I'm not sure why that was so difficult.

Now, it is true that in one specific case, out of seven benchmarks posted (note that it was *not* Q3:A, which scaled almost linearly with the CPU), the video card was the limiting factor. Of course, that situation is (obviously) liable to happen, and if Tom didn't show that, then he wouldn't be giving you the whole picture.

And, as noted before, the idea that anyone in their right minds would spend $2500 for an 80 pound supercooled computer and play games on it with anything less than the best video card is patently ludicrous.

I guess that's why I'd rather get my benchmarks from Tom than from you.

They must have plenty of bandwith if they can handle it! I'm sure Tom explained how he got to choose pair, but I can't seem to find the link (and Thomas the searchengive is dead...)

---

Toms Hardware Guide has a much less biased review comparing the MAXX and GeForce more clearly and openly.

Sharkey's review failed to point out that he was using the GeForce SDR, whilst (as THG points out) the GeForce DDR has 16% more memory bandwidth. The benchmarks on THG show how this is a big advantage with the MAXX not even coming close. I wonder where the discrepancy between THG and Sharkey come from with SDR though, with THG indicating that it is generally faster then the MAXX?

"the features of Fury MAXX in comparison with the other high-end 3D-solution, the GeForce from NVIDIA.

Fill Rate
ATI Rage128 Pro AFR:
2 x 250 Mpixels/s = 500 Mpixels/s

NVIDIA GeForce256:
480 Mpixels/s


In terms of fill rate, Fury MAXX has a slight edge over GeForce, but generally both solutions are close to identical. This will mean that both cards will score close to the same in games that don't use T&L, if fill rate is the only limiting factor.

Memory Bandwidth
ATI Rage128 Pro AFR:
2 x 2.288 GB/s = 4.576 GB/s

NVIDIA GeForce256:
2.656 GB/s as SDR-version
5.312 GB/s as DDR-version


You can see that Fury MAXX is way ahead of GeForce w/SDR in terms of memory performance. This will make Fury MAXX look a lot better than GeForce's SDR-version at high resolutions and high color-depths. GeForce's DDR-version however comes with a higher memory-bandwidth than Fury MAXX, something that might cost a lot more money though. "

Toms Hardware Guide has a much less biased review comparing the MAXX and GeForce more clearly and openly.

Sharkey's review failed to point out that he was using the GeForce SDR, whilst (as THG points out) the GeForce DDR has 16% more memory bandwidth. The benchmarks on THG show how this is a big advantage with the MAXX not even coming close. I wonder where the discrepancy between THG and Sharkey come from with SDR though, with THG indicating that it is generally faster then the MAXX?

"the features of Fury MAXX in comparison with the other high-end 3D-solution, the GeForce from NVIDIA.

Fill Rate
ATI Rage128 Pro AFR:
2 x 250 Mpixels/s = 500 Mpixels/s

NVIDIA GeForce256:
480 Mpixels/s


In terms of fill rate, Fury MAXX has a slight edge over GeForce, but generally both solutions are close to identical. This will mean that both cards will score close to the same in games that don't use T&L, if fill rate is the only limiting factor.

Memory Bandwidth
ATI Rage128 Pro AFR:
2 x 2.288 GB/s = 4.576 GB/s

NVIDIA GeForce256:
2.656 GB/s as SDR-version
5.312 GB/s as DDR-version


You can see that Fury MAXX is way ahead of GeForce w/SDR in terms of memory performance. This will make Fury MAXX look a lot better than GeForce's SDR-version at high resolutions and high color-depths. GeForce's DDR-version however comes with a higher memory-bandwidth than Fury MAXX, something that might cost a lot more money though. "

The simple facts (from Tom's Hardware)

Coppermine processors:
Rated Speed Bus Speed Multiplier
733 133 5.5
700 100 7
667 133 5
650 100 6.5
600 EB 133 4.5
600 E 100 6
533 EB 133 4
550 E (PPGA) 100 5.5
500 E (PPGA) 100 5

Now, since the multipliers are locked on Coppermine processors, there's no way that these people could get a 7.5 multipler. The highest is a "7" with a front-side bus of 100 mhz.

Note that they also make absolutely no mention of any method of cooling this thing. The only thing presented are the benchmarks, all of which are directly proportionate to the increase in mHz rating. Coincidence? I doubt it. There's no mention _at all_ of the hardware this was tested on, except for one screenshot that mentions an ABIT BX6 (Rev. 2) motherboard.

Before you go expounding on the wonders of 1ghz Coppermines, it'd be wise to check the facts first.

Check out these comparisons between the Coppermine and Athlon:

Ace's Hardware
Tom's Hardware

Read this and THIS and I'm sure you'll look at it a bit differently.

Read this and THIS and I'm sure you'll look at it a bit differently.

• X86.ORG
  
• Tom's Hardware
  

Why do I believe his side of this story in this? Mostly because of all the nasty things I've been reading about Intel's *apparent* strong arming of Taiwanese motherboard manufacturers, and the *apparent* cutting of a deal with Gateway so they'd stop using AMD CPUs in their machines. Not to mention Intel's failed attempt to *apparently* force RDRAM on OEMs and the market, and a number of other Micros1~ esque tactics.

The fact of the matter is, Tom is very thorough. He was the first to publish an overclocking guide for the Athlon for example. Another example of his dedication is his yearly trips to Taiwan to talk directly with manufacturers there. I mean who does this but someone into what he does? The guy doesn't get paid to do this you know, he's a medical doctor, not an employee of some corporation.

Regarding 3dfx. They've really been acting strange lately, and not a little bit cheesy, IMO.

For example, cutting off Creative and Diamond like they did, and making their chipset proprietary. Maybe they have a right to be paranoid, what with Intel in bed with S3 and S3 now owning Diamond... but this was a bad move. Look at the hell that Apple went through as a result of it's decision to go proprietary back in the days before the Return of Jobs(TM). 3dfx will, I suspect, go through similar troubles, as major manufacturers have little choice but to either create their own chipsets (which can be disastrous, look at #9's "Ticket To Ride") or use another company's chipset, like nVidia -- who is making better 3D chipsets right now.

Another thing that 3dfx did which has lost them quite a bit of market share, (as evidenced by contrasting nVidia's 1999 profit and loss statements with 3dfx's) was essentially ship an overclocked banshee card :) What else can you call the Voodoo3 with @ 16 bit? Then there's the glaring lack of a heat sync fan on the voodoo3s (fans come with nVidia based cards), and the vast, power supply crippling, 183 MHz of the Voodoo3 3500, which I've read has *apparently* been problematic for some users with lower wattage power supplies...

These things don't go unnoticed by consumers, or for that matter by honest hardware reviewers. The benchmarks Tomas Pabst used in the GeForce article are valid, and Voodoo3 scored at zero on the 32 bit true color tests, because it is a 16 bit card. Simple as that... if you'd read the whole article, you'd see that he did list GeForce's short commings, specifically the memory bandwidth problems with the SDRAM versions of the Card vs. the upcomming double data rate RAM versions of the card (to say nothing of the *expected* 64 MB GeForce that I'm waiting for).

I have not lost faith in 3dfx, and I hope that their troubles of late will cause some restructuring which will lead to an increased emphasis on design quality, rather than throwing MHz at the problem. I do have to applaud their efforts to support the Apple community with Mac drivers for voodoo cards, although I hear poor Microconversions *might* have been forced out of the mac/voodoo card business as a result :(

Despite all this, the 3D wars are far from over, and I suspect the Voodoo4 will be quite a sight to see, but I won't rush out and buy either a GeForce or a Voodoo4 until I know all the facts, and for that I'll probably read a number of reviews. I've also found that forums where actual users relay anecdotal accounts of their experiences with specific products are truly telling. If you don't believe me go check out the forums at Intel :)

Why do I believe his side of this story in this? Mostly because of all the nasty things I've been reading about Intel's *apparent* strong arming of Taiwanese motherboard manufacturers, and the *apparent* cutting of a deal with Gateway so they'd stop using AMD CPUs in their machines. Not to mention Intel's failed attempt to *apparently* force RDRAM on OEMs and the market, and a number of other Micros1~ esque tactics.

The fact of the matter is, Tom is very thorough. He was the first to publish an overclocking guide for the Athlon for example. Another example of his dedication is his yearly trips to Taiwan to talk directly with manufacturers there. I mean who does this but someone into what he does? The guy doesn't get paid to do this you know, he's a medical doctor, not an employee of some corporation.

Regarding 3dfx. They've really been acting strange lately, and not a little bit cheesy, IMO.

For example, cutting off Creative and Diamond like they did, and making their chipset proprietary. Maybe they have a right to be paranoid, what with Intel in bed with S3 and S3 now owning Diamond... but this was a bad move. Look at the hell that Apple went through as a result of it's decision to go proprietary back in the days before the Return of Jobs(TM). 3dfx will, I suspect, go through similar troubles, as major manufacturers have little choice but to either create their own chipsets (which can be disastrous, look at #9's "Ticket To Ride") or use another company's chipset, like nVidia -- who is making better 3D chipsets right now.

Another thing that 3dfx did which has lost them quite a bit of market share, (as evidenced by contrasting nVidia's 1999 profit and loss statements with 3dfx's) was essentially ship an overclocked banshee card :) What else can you call the Voodoo3 with @ 16 bit? Then there's the glaring lack of a heat sync fan on the voodoo3s (fans come with nVidia based cards), and the vast, power supply crippling, 183 MHz of the Voodoo3 3500, which I've read has *apparently* been problematic for some users with lower wattage power supplies...

These things don't go unnoticed by consumers, or for that matter by honest hardware reviewers. The benchmarks Tomas Pabst used in the GeForce article are valid, and Voodoo3 scored at zero on the 32 bit true color tests, because it is a 16 bit card. Simple as that... if you'd read the whole article, you'd see that he did list GeForce's short commings, specifically the memory bandwidth problems with the SDRAM versions of the Card vs. the upcomming double data rate RAM versions of the card (to say nothing of the *expected* 64 MB GeForce that I'm waiting for).

I have not lost faith in 3dfx, and I hope that their troubles of late will cause some restructuring which will lead to an increased emphasis on design quality, rather than throwing MHz at the problem. I do have to applaud their efforts to support the Apple community with Mac drivers for voodoo cards, although I hear poor Microconversions *might* have been forced out of the mac/voodoo card business as a result :(

Despite all this, the 3D wars are far from over, and I suspect the Voodoo4 will be quite a sight to see, but I won't rush out and buy either a GeForce or a Voodoo4 until I know all the facts, and for that I'll probably read a number of reviews. I've also found that forums where actual users relay anecdotal accounts of their experiences with specific products are truly telling. If you don't believe me go check out the forums at Intel :)

SDRAM DIMMS are 8 bytes wide (64 bits).

That article on Tom's Hardware Guide, " Performance Impact of Rambus " says that RDRAM's bus width had to be reduced by 75% to 16 bits (2 bytes) to run at 800MHz. Going backwards and you get 8 bytes for SDRAM (8 - 75% = 2).

Now if we divide all your SDRAM access times by 2 we get 20ns for 100MHz DDR SDRAM which is the same as 800MHz rambus.

There are some great articles regarding bandwith vs. latency in general and RamBus in particular at Tom's Hardware. To summarize the articles, even today's current SDRAM architecture provides more than enough bandwidth, especially with the current sophisticated cache systems that reduce memory accesses dramatically. However, what's tying up the CPU is latency, especially as CPU's get faster.

In other words, CPUs generally request small amounts of data with any given request, but it has to wait a long time for that request to get back. As CPU speed has increased, better cache systems have mitigated the resulting increased bandwidth demands but nothing has helped the resulting latency problems. So the way to speed up memory is to decrease latency and don't worry too much about bandwidth just yet. Unfortunately, RamBus goes in the exact opposite direction.

That said, I guess we should never underestimate the power of a behemoth like Intel to force acceptance of poor technologies :-/

There are some great articles regarding bandwith vs. latency in general and RamBus in particular at Tom's Hardware. To summarize the articles, even today's current SDRAM architecture provides more than enough bandwidth, especially with the current sophisticated cache systems that reduce memory accesses dramatically. However, what's tying up the CPU is latency, especially as CPU's get faster.

In other words, CPUs generally request small amounts of data with any given request, but it has to wait a long time for that request to get back. As CPU speed has increased, better cache systems have mitigated the resulting increased bandwidth demands but nothing has helped the resulting latency problems. So the way to speed up memory is to decrease latency and don't worry too much about bandwidth just yet. Unfortunately, RamBus goes in the exact opposite direction.

That said, I guess we should never underestimate the power of a behemoth like Intel to force acceptance of poor technologies :-/

There are some great articles regarding bandwith vs. latency in general and RamBus in particular at Tom's Hardware. To summarize the articles, even today's current SDRAM architecture provides more than enough bandwidth, especially with the current sophisticated cache systems that reduce memory accesses dramatically. However, what's tying up the CPU is latency, especially as CPU's get faster.

In other words, CPUs generally request small amounts of data with any given request, but it has to wait a long time for that request to get back. As CPU speed has increased, better cache systems have mitigated the resulting increased bandwidth demands but nothing has helped the resulting latency problems. So the way to speed up memory is to decrease latency and don't worry too much about bandwidth just yet. Unfortunately, RamBus goes in the exact opposite direction.

That said, I guess we should never underestimate the power of a behemoth like Intel to force acceptance of poor technologies :-/

Tom's Hardware has a full preview, although he's not allowed to print the performance results. This isn't SLI or PGC -- ATI is actually having the CPU's draw full alternate frames, so the image quality will be high yet the speed will be doubled. It's even buffered so if CPU0 is taking a long time drawing frame 0, CPU1 can keep drawing 1,2,3,... until CPU0 is done. What I like best about this is that the MAXX architecture will allow them to drop their latest chips in as they are developed, so even if their CPU architectures remain a little behind, they'll be able to keep competitive. I like ATI cards because of all of the MPEG and TV toys they build into them; the only other company that even comes close to offering those kind of toys is Matrox, but they're just too damn expensive for the full-featured cards.