Ohh I love the uniformed idiots that just mindlessly repeat things without ever bothering to check for facts.
Fact: Heat is directly releated to two things and two things along. How much power a chip uses, and how good your cooling system is at getting rid of that heat.
Now, check out the power consumption of AMD and Intel processors, it's quite clearly documented in their respective tech docs (or at least it is in AMD's tech docs, Intel uses the rather ambigious "thermal design power" number to describe how much power the chip uses, with the TDP being a little bit less than the actual maximum power the chip can use).
Here's a sample for ya:
AMD AthlonXP 2000+ (180nm fab process): 70.0W Max, 62.5W typical Intel P4 2.0GHz (180nm fab process): 75.3W TDP
AMD AthlonXP 2600+ (130nm fab process): 68.3W Max, 62.0W typical Intel P4 2.60GHz (130nm fab process): 62.6W TDP
Long story short, these chips are all in the same basic range, all within about 10% of one another except for the old 2.0GHz (Willamette) P4s.
So, if the power used is roughly the same, than the only real differences comes down to how good your cooling system is, and this is why the P4 seems better. See, almost all P4 heatsinks are at least 70x70mm at their base, and most are around 80x80mm. By contrast, a lot of Athlon heatsinks are much smaller 60x60mm. However, if you use one of the 80x80mm heatsinks for an Athlon, guess what? The chip is nice and cool, even with a slow-spining fan that hardly makes any noise.
* Data Starved Processors - and this is all about latency
That's mostly a function of main memory. With 512K of cache, your hit rate is typically in the 95-98% range. Throwing tons of cache on the motherboard rarely helps much, since it usually only bumps your hit rate up to around 96-98%. Generally speaking, if your data set doesn't fit into 512KB of cache, it usually won't fit into any amount of cache, no matter how big. So what we really need is a type of memory that offers very low latency but is cheap enough that it can be used as main memory. Some technologies like prefetching can help hide this latency, but sooner or later, all those break down.
Ohh, and to make L3 cache really effective, it would probably have to be hanging off a backside bus of the processor anyway, not off the chipset. I know that Intel talked about doing this with their current line of Xeons, but I don't know what ever came of those plans.
# Improved hyperthreading support
Hyperthreading performance is about 49.5% CPU, 49.5% software, and about 1% chipset. There is virtually nothing that can be done on the chipset to specifically improve hyperthreading performance. All the chipset manufacturers could do here are fairly generalized improvements that would end up helping out chips both with and without hyperthreading.
where are nice stuff like block data copy between video and RAM (like the SGI chipsets for the Indy/O2 had)
Uggg.. copying data back and forth between main memory and the video controller? That's a sure way to hurt your performance! The SGI solution only made sense because it was cheaper/easier for them to have a single high-bandwidth bus with and a single chunk of a GB of memory or so. However if they could have had a GB of video memory and a GB of main memory, with each having tons of bandwidth, they would have been better off. PCs are in that situation. These days, having a video card with 10-20GB/s of memory bandwidth and 128MB+ of memory is cheap, which essentially eliminates the need to read/write between (slow) main memory.
There's really nothing wrong with AGP other than the fact that it's original design idea has become obsolete by the fact that memory is dirt-cheap now. Otherwise it offers over a GB/s of bandwidth for what amounts to essentially 1-time read/writes. After that, all the magic happens on the video card itself.
# Standard interfaces for custom silicon
Umm.. ok.. whatever. The market for this is approximately 2 people. Still, believe it or not, you're actually going to see just such a thing in about 6-months time with AMD's Hammer processors. These chips/chipsets will have Hypertransport links, which will offer a high-bandwidth connection directly to the chipset if you so desire. Of course, if you want to make use of it you're going to have to design your own motherboard from the ground up, because the market for what you're looking for is TINY, and no motherboard manufacturer is going to waste their money on such a thing.
You're basing your entire opinion on integrated graphics on a PC-Chips motherboard?! PC-CHIPS?!?!
That's the same company that put fake chips on a board that pretended to be cache but didn't actually do anything at all! This is the same company that refuses to put their own name on any board that they sell for fear that someone might be able to track the things back to them!
And you were surprised that the board was crap?! It doesn't matter how good the chipset is, a sufficiently crappy job on motherboard design will fuck it up. PC-Chips made some TERRIBLE boards based on the Inel 440BX chipset (arguably the best ever PC motherboard chipset). They are quite capable of screwing up anything they touch.
That's it!? Geez, how can you survive, your minimum frame rate MUST drop bellow 100 f/s on that setup! Hell, it might even get bellow the 85Hz refresh rate of the monitor, or if you're really unlucky, bellow the 60 f/s which is the maximum the human eye can perceive!
It's no good unless you're frame rate NEVER drops bellow 200 f/s! </sarcasm>
Well, since the article is/was slashdotted, I guess I can't fault you TOO much for not having read it:>
ALL of the boards tested in this article have an external AGP slot. In fact, it's quite rare these days for a board with integrated graphics to NOT have an external AGP slot. Yes, they do exist, and you could probably find one if you bought a HPaq or a Dell, but in the retail or white box market such a beast is VERY rare.
Hmm.. that's interesting. I use an nForce board at home with it's integrated graphics and find it quite crisp and clear at the 1024x768 I usually use it at, and it also seemed fine at 1152x9??. It may not be a top-end, super-high-quality 2D card, but in my experience it's been better than the vast majority of systems I've used at various jobs (which mostly had el-cheapo video cards installed).
And just how hard is it to stick a video card in the AGP slot included with these boards?
Come on people! Take a look at the boards before sprouting off mindless opinions! Every single board in the review had a fully functional AGP slot alongside their integrated video!
I bought a nForce board about a 8 months ago. I play games from time to time, but I'm not worried about getting 200+ f/s on all my games, so the integrated graphics is fine for me for now. In 6 months to a years time, the integrated graphics will probably be getting a touch on the slow side, so I'll probably pick up a GeForce4 Ti 4200 than (which should be nice an cheap after the GF5 comes out) and it should serve me well for at least another year, possibly two.
Well the OEMs dont' seem to agree with you. The nForce chipset is currently being used by ALL of the major OEMs that sell Athlon based PCs. Most also include a low-end SiS or VIA based solution as well, but the nForce seems to take up a lot of the mid-range with integrated graphics and the high-end without integrated graphics.
I know that people tend to really hate it when I post facts instead of blathering out mindless opinions.. however I'm going to go with the facts anyway.
Here's a link to some real and true benchmarks of the nForce 220 and 420 with their integrated graphics as compared to a GeForce2MX in Unreal Tournament and Max Payne.
For UT, the nForce 420 is 9.5% slower than the GeForce2 MX (at 1024x768), and for Max Payne the nForce is just under 15% slower. That's a far cry from "1/3rd the frame rate".
Ohh, and the original nForce was available without integrated graphics as well (the nForce 415 chipset).
Long story short, the nForce can and does play games just fine. It's not getting 200+ f/s at 1600x1200, but as long as you can get by with only 30-60f/s at 800x600, the nForce is up to the task. The nForce 2 should be about twice as fast since it has somewhat higher memory bandwidth and some new stuff which reduces it's dependance on memory bandwidth.
Transmeta had something interesting in their code morphing software, but they didn't make use of it. Instead all they ever managed to produced is a butt-slow x86 processor that used an ass-backwards way of getting x86 compatability.
Their only saving grace was that, for a while, they had lower powered chips then the other companies out there, however even that has been eliminated with the ULV PIIIs and the VIA C3 chips. Once Intel's Banias chip is available, that'll be the final stake in Transmeta's coffin if someone hasn't bought them up by that time.
Ohh, and yes, I am aware that I have been speaking of Transmeta in the past-tense. They're a dead company. They'll probably be bought out by someone who has some potential uses for their code morphing software. My guess is IBM, but there are other possibilities.
2. Allocate your engineering resource to make the processor/system run cooler instead, to the point that it no longer need the active cooling measure(fan) on processor and (hopefully) whole board/system.
You mean like the Banias chip that Intel is developing?
3. Make a new small, low power, quiet PC form factor standard
4. Make this combo your main production, push other heatmaker to the niche.
Now that probably isn't something that Intel is after. There isn't the money to be made in that market as their is in the high-end CPU market... at least not at this point in time. Things may change.
Anyway, long story short, there is some thought to doing what you suggest, but initially it will be all targeted towards laptop users. Maybe it'll start to bubble down to the desktop as well. I certainly wouldn't mind a TiVo/DVD/DivX type system sitting in my living room, and a full sized PC with a loud fan is not exactly what I had in mind.
The original timeframe was samples in Q2 of '02 and product shipping in the ever elusive "2nd half of '02". It was already pushed back to volume shipments in Q4 '02 or Q1 of '03. Now it's being pushed back again until (probably) Q2 of '03.
On the upside, samples have been shipping for a little while now. This suggests to me that the problem is related more to the manufacturing of the chip then the chip design itself, though that's obviously just a guess at this point in time (and it doesn't exactly matter what the cause of the delay is if the end result is the same).
Re:This shouldn't be much of a suprize....
on
AMD Delays Hammer
·
· Score: 1
Where do you think AMD has originations from?
AMD traces it's origins back to Fairchild, same as Intel. Actually AMD and Intel were formed only 6 months apart under relatively similar circumstances. They even setup shop just down the road from one another.
I suppose you could say that AMD was just copying Intel then though, after all, Intel set up in early '69, while AMD didn't setup until mid/late '69!:>
intel puts out something that is more intuitive
More intuitive?! It's a processor! How intuitive or unintuitive can it get? You take the chip, you stick it in a socket!
Take a look at SPEC CPU2000 sometime. Ok, maybe it's not the end-all-be-all of processor benchmarks, but it's the best that we've got right now. In terms of raw processing power, we get the following rankings:
As you can see, AMD and Intel are doing quite well competitng with the rest of the world here, and even beating them. Add an extra 20% to the performance here, and AMD would have a VERY fast little chip on their hands.
Now, in CFP, things look somewhat worse for the P4 and esepcially for the Athlon. This is largely due to two things. First, CFP is a very bandwidth-intensive benchmark, and bandwidth is one of those things that can easily be bought by throwing money at it. IBM threw a LOT of money at bandwidth, and it performs very well.
The second thing with CFP is Sun's new compiler, which magically transformed the rather lackluster performance of the UltraSparc III into a real performer in CFP due almost entirely to a HUGE (8000%) gain in performance on a single sub-test. In other words, their results are not really accurately reflecting the performance of the chip relative to other chips at this time, because they will either perform quite a bit worse then their CFP numbers would suggest, or WAY better if your application happens to exactly match that one sub-test (99+ times out of 100 it would be the former).
Still even here AMD and Intel are reasonably competative. The P4, with a score of 929 is behind only the Power4, the Itanium2 and the Alpha EV68 in CFP. Even the AthlonXP's score of 710 isn't too far off the pace, albeit it is trailing the leading score of 1356 (Itanium2) by a reasonable margin.
As for the Hammer? It's integrated memory controller should REALLY help for CFP, particularly with regards to the Sledgehammer, with a 128-bit wide memory controller. CINT will benefit somewhat less, probably only the 20-30%, clock for clock, over what the AthlonXP (this is what AMD is saying would be typical).
Anyway, the point I'm trying to get at here is that the AthlonXP and the P4 are actually VERY powerful chips, and really there isn't anything that is "leaving them in the dust" at this time.
They'll get beaten until Sledgehammer is released.. not Claw hammer which will have no x86-64 desktop software support right off the bat, and will have to rely solely on it's pure x86 performance.
64-bit performance is unlikely to be significantly different from 32-bit performance. Actually, all else being equal, 64-bit software is SLOWER then 32-bit software (twice as much data to read/write from memory), however the ability to properly access more then 4GB of memory can and does offset that penalty for many applications. In the specific case of the AMD Hammer and x86-64, making use of the 64-bit mode is expected to boost performance by about 5%, due primarily to the fact that AMD doubled the number of registers for x86-64 over IA-32.
Microsoft shafted them on the X-box because Intel paid Microsoft 200 million to use the Pentium III
I don't know any details of the deal between Intel and Microsoft, and in fact, I highly doubt that ANYONE reading and posting here does, just rumors heard from rumor mills. It is clear that Intel gave MS a VERY good deal, though Intel is probably not making any money at all off this deal, especailly given that X-Box sales have been well bellow expectations and profit from this sort of chip would be driven purely by volume.
Nvidia was stuck with an unused AMD integrated chipset for X-box and Nforce was born.
The chipset market is a market that nVidia HAD to get into at one time or another. Integrated video chipsets have grown from being ~5% of the market to being ~60% of the market in just over 3 years time. With an integrated video chipset, nVidia's potential market was more then cut in half. That being said, the nForce seems like a good thing for both AMD and nVidia. For AMD it has given them a stable platform that OEMs trust (as indicated by HPaq using it in their business PCs, something that is VERY rare for non-Intel chipsets). For nVidia, they have a successful product that is likely close to if not the top selling chipset for AMD processors (VIA mostly owns the retail channel, but nVidia has taken almost all the OEM sales).
Intel will pay Microsoft to shaft them again. No x86-64 Windows XP for AMD despite AMD testifying on Microsoft's behalf in exchange for anti-trust testimony
You seem to think that somehow Intel and Microsoft LIKE one another, or that by beating down AMD it would somehow HELP Microsoft? Perhaps you even believe in the whole "Wintel" alliance?! Well I hate to break it to you, but Microsoft wants/needs a company like AMD to give Intel competition. This lowers the prices of Windows PCs as comapred to the Unix boxes, it helps increase their performance and push things forward. Helping AMD makes GOOD business sense for MS, particularly if it doesn't cost them much money. And the x86-64 port of WinXP apparently didn't cost too much money, and they have already demonstrated it on Clawhammer systems.
Probably about the same time as they support Hyper threading and SSE3 for Intel.
WinXP already supports Hyperthreading. Even Win2K will work with Hyperthreading, though it appears like it might have been a touch of a hack (ie the chips show up as two separate processors, not a single processor which can run two threads at once).
As for SSE3, I'm really not quite sure where Intel is going with that, I'll probably have to read up on it a bit more. It doesn't seem to make much sense to me though, they barely got people started on writing SSE code when they released SSE2 and told people to start again. Now they're planning on doing the same thing with SSE3? I'm not even sure that SSE3 will add all that much with is actually useful.. I guess we'll just wait and see.
The whole "Wintel" relationship is mostly just a huge myth. At best, Intel and Microsoft are forced-friends.
Both Intel and Microsoft have realized for quite some time that they need each other to survive, at least for the time being, but neither of them are happy about that fact, and both are striving to change it. Microsoft has been a fairly strong supporter of AMD for some time now, while Intel was one of the first major hardware companies to jump on the Linux bandwagon.
Think about it, Intel is pretty much the dominant force in PC hardware, being not only the primary supplier of processors, but also the #1 supplier of motherboard chipsets, video solutions, and being well up there if not #1 when it comes to audio and NICs (most of this is now integrated into motherboards). They are also the driving force behind most buses and interconnects, ie PCI and AGP (AMD's Hypertransport is a notable exception here), not to mention the fact that they defined the ATX form factor that virtually all current PCs make use of (albeit in a somewhat basterdized format for some of the big OEMs).
Microsft, on the other hand, is the dominant force in software, having the most common operating system, office software, web browser, e-mail program, etc. etc. ad nauseum.
The end result is that Microsoft is the only company in the PC world that Intel doesn't have a fair degree of control over (AMD fights, but ends up adopting Intel-compatible technology in the end, ie MMX/SSE/SSE2). Similarly Intel is the only company that Microsoft doesn't exert a signicant amount of control over in the PC world, though even there I'd say that MS mostly won that battle a while back.
Anyway, long story short, MS and Intel, not so much allies as simply two companies forced together by circumstance.
The P4 does indeed do 4 transfers per clock on it's bus. It does this by having two clocks 90 deg out of phase with one another, and sending data on the rising and falling edges of both clocks. It's definitely trickier, from both a conceptual point of view and from a design point of view then DDR, but it's not magic, and it's definitely not just a marketing term to try to hide things or anything like that.
Ok, the C3 may not be besting the Athlons and P4s of the world, but it's a whole heck of a lot faster then you're making it out to be!
The C3 at 800MHz will have no trouble at all decoding DivX, let alone mpeg-1. Many people have quite successfully used these chips in set-top boxes for playing DVDs and DivX movies.
Ok, totally off-topic, but something I've been asking for a while, where the heck are PDAs based on PalmOS 5.0?!
The operating system was released back in June, and (silly me) I figured that PalmOS had worked with PDA manufacturers (or at the very least Palm themselves) beforehand so that actual products using this new operating system would be out shortly there-after. Well it's now 3 months later, and the best I've heard are some far-off rumors that the new handhelds MIGHT be coming sometime sorta not too long from now, like next year!
Come on Palm, Handspring, Sony et al. EVERYONE knew basically what was going to be required for this new OS LONG ago, why is the hardware lagging so far behind the software?! It's not even like the hardware would require that much in the way of development, since any current PocketPC PDA should be able to run PalmOS 5.0 with virtually no hardware modifications.
AMD tried that, and their popularity dropped, profits dropped and they were barely able to sell their top-end chip for $100. Apple is still trying it, and still failing miserably (though Apple has the unfortunately downside of actually have very slow chips, regardless of whether you look at their clock speed).
Face it, the few that know enough to bother looking beyond a single number know enough to realize that clock speed is a completely useless number.
Personally I don't really care too much one way or the other. AMD replaced one totally meaningless number with another totally meaningless number. Ideally this sort of thing would just encourage people to actually look beyond the pretty number and try to figure out what it actually means, but both myself and AMD's marketing dept. are well aware that that sort of thing is WELL beyond what 99% of the buying public are going to do. So, model numbers it is, and they worked. AMD's profits and average selling price have increased a lot since the indroduction of their model numbers, even now in a bit of a PC market downturn.
Well, 10 years ago everyone was saying the EXACT same thing about how we really didn't need these new fandangled Pentiums, and that a 486 was really fast enough for everything, and that in 10 years time we would all be laughing at the people who thought that they needed a 100MHz processor!
The more times change, the more they stay the same. 10 years from now, we'll all be looking back at these AthlonXP 2700+ systems and wondering how we managed to get by with such SLOW processors, because runs REALLY slowly on anything less then a 10GHz chip.
It's been happening for the past 20 years, and I don't see anything today that makes this look different.
I don't know that the Barton core will use any lower latency, or lower power at all.
The Barton core won't actually be manufacturered by AMD, "Barton" is the code name that is now being used for the AthlonXP processors that UMC will be building for AMD. First off, UMC is still a few months away from shipping chips. Secondly, they might not be able to manage as high clock speeds as AMD manages in their own fabs. AMD's Fab 30 in Dresden (where the Athlons are currently all produced) is quite a high-end fab that is really tuned for high performance/high clock rate chips. UMC's fabs tend to really emphasize low cost/high yields, so they might not get the same sort of bin splits that AMD can get on their own.
The real benefit of the Barton core should be for multiprocessing setups, where the extra cache should really help out a lot. For single processor setups, I doubt that we'll see that big of an improvement over the current Thoroughbred core, probably only about 3-5%, or about the same as the improvement seen when increasing the bus speed from 133/266MHz DDR to 166/333MHz DDR.
Well, FWIW the Thermal Design Power of the P4 at 2.8GHz is 68.4W. This number is kinda-sorta-but-not-quite the maximum power that the chip will consume (I've ranted about this in other threads:> ). AMd's AthlonXP 2600+ has a maximum real-world power draw of about 68W as well (AMD does document the real exact figure, but I don't have the data sheets in front of me and they don't open properly in Mozilla anyway due to a bug in the Acrobat plug-in... but again, that's for another rant).
So, is AMD going to have to redesign their motherboards to accomodate this new chip? Well, maybe and maybe not. The Athlon 1.4GHz processor is the highest power consuming chip of the Athlon line to date at 70W maximum, and that's probably going to about what an AthlonXP 2700+ will require. However the 2700+ will get that 70W at a lower voltage but higher current. Long story short, it depends on the motherboard. AMD tends to leave this sort of thing more up to the motherboard manufacturer's designs, while Intel kinda forces motherboard manufacturers to do everything the Intel way. So, chances are that some AthlonXP boards will be able to accomodate this chip without change, while others will need a new voltage requlator.
Re:AMD HEAT issues AC is not cheap!
on
AMD's Athlon XP 2700+
·
· Score: 3, Informative
Processor myth #1: AMD chips generate more heat then Intel chips.
Check the data sheets sometime, the amount of heat that an AthlonXP and a P4 put out is nearly identical. Both are also only in the 60-70W range, or about the same as your typical light bulb. We all know how turning on a single light in your house cause your AC bills to skyrocket!!!
Slashdot Mirror
Ohh I love the uniformed idiots that just mindlessly repeat things without ever bothering to check for facts.
Fact: Heat is directly releated to two things and two things along. How much power a chip uses, and how good your cooling system is at getting rid of that heat.
Now, check out the power consumption of AMD and Intel processors, it's quite clearly documented in their respective tech docs (or at least it is in AMD's tech docs, Intel uses the rather ambigious "thermal design power" number to describe how much power the chip uses, with the TDP being a little bit less than the actual maximum power the chip can use).
Here's a sample for ya:
AMD AthlonXP 2000+ (180nm fab process): 70.0W Max, 62.5W typical
Intel P4 2.0GHz (180nm fab process): 75.3W TDP
AMD AthlonXP 2600+ (130nm fab process): 68.3W Max, 62.0W typical
Intel P4 2.60GHz (130nm fab process): 62.6W TDP
Long story short, these chips are all in the same basic range, all within about 10% of one another except for the old 2.0GHz (Willamette) P4s.
So, if the power used is roughly the same, than the only real differences comes down to how good your cooling system is, and this is why the P4 seems better. See, almost all P4 heatsinks are at least 70x70mm at their base, and most are around 80x80mm. By contrast, a lot of Athlon heatsinks are much smaller 60x60mm. However, if you use one of the 80x80mm heatsinks for an Athlon, guess what? The chip is nice and cool, even with a slow-spining fan that hardly makes any noise.
* Data Starved Processors - and this is all about latency
That's mostly a function of main memory. With 512K of cache, your hit rate is typically in the 95-98% range. Throwing tons of cache on the motherboard rarely helps much, since it usually only bumps your hit rate up to around 96-98%. Generally speaking, if your data set doesn't fit into 512KB of cache, it usually won't fit into any amount of cache, no matter how big. So what we really need is a type of memory that offers very low latency but is cheap enough that it can be used as main memory. Some technologies like prefetching can help hide this latency, but sooner or later, all those break down.
Ohh, and to make L3 cache really effective, it would probably have to be hanging off a backside bus of the processor anyway, not off the chipset. I know that Intel talked about doing this with their current line of Xeons, but I don't know what ever came of those plans.
# Improved hyperthreading support
Hyperthreading performance is about 49.5% CPU, 49.5% software, and about 1% chipset. There is virtually nothing that can be done on the chipset to specifically improve hyperthreading performance. All the chipset manufacturers could do here are fairly generalized improvements that would end up helping out chips both with and without hyperthreading.
where are nice stuff like block data copy between video and RAM (like the SGI chipsets for the Indy/O2 had)
Uggg.. copying data back and forth between main memory and the video controller? That's a sure way to hurt your performance! The SGI solution only made sense because it was cheaper/easier for them to have a single high-bandwidth bus with and a single chunk of a GB of memory or so. However if they could have had a GB of video memory and a GB of main memory, with each having tons of bandwidth, they would have been better off. PCs are in that situation. These days, having a video card with 10-20GB/s of memory bandwidth and 128MB+ of memory is cheap, which essentially eliminates the need to read/write between (slow) main memory.
There's really nothing wrong with AGP other than the fact that it's original design idea has become obsolete by the fact that memory is dirt-cheap now. Otherwise it offers over a GB/s of bandwidth for what amounts to essentially 1-time read/writes. After that, all the magic happens on the video card itself.
# Standard interfaces for custom silicon
Umm.. ok.. whatever. The market for this is approximately 2 people. Still, believe it or not, you're actually going to see just such a thing in about 6-months time with AMD's Hammer processors. These chips/chipsets will have Hypertransport links, which will offer a high-bandwidth connection directly to the chipset if you so desire. Of course, if you want to make use of it you're going to have to design your own motherboard from the ground up, because the market for what you're looking for is TINY, and no motherboard manufacturer is going to waste their money on such a thing.
You're basing your entire opinion on integrated graphics on a PC-Chips motherboard?! PC-CHIPS?!?!
That's the same company that put fake chips on a board that pretended to be cache but didn't actually do anything at all! This is the same company that refuses to put their own name on any board that they sell for fear that someone might be able to track the things back to them!
And you were surprised that the board was crap?! It doesn't matter how good the chipset is, a sufficiently crappy job on motherboard design will fuck it up. PC-Chips made some TERRIBLE boards based on the Inel 440BX chipset (arguably the best ever PC motherboard chipset). They are quite capable of screwing up anything they touch.
That's it!? Geez, how can you survive, your minimum frame rate MUST drop bellow 100 f/s on that setup! Hell, it might even get bellow the 85Hz refresh rate of the monitor, or if you're really unlucky, bellow the 60 f/s which is the maximum the human eye can perceive!
It's no good unless you're frame rate NEVER drops bellow 200 f/s!
</sarcasm>
Well, since the article is/was slashdotted, I guess I can't fault you TOO much for not having read it :>
ALL of the boards tested in this article have an external AGP slot. In fact, it's quite rare these days for a board with integrated graphics to NOT have an external AGP slot. Yes, they do exist, and you could probably find one if you bought a HPaq or a Dell, but in the retail or white box market such a beast is VERY rare.
Hmm.. that's interesting. I use an nForce board at home with it's integrated graphics and find it quite crisp and clear at the 1024x768 I usually use it at, and it also seemed fine at 1152x9??. It may not be a top-end, super-high-quality 2D card, but in my experience it's been better than the vast majority of systems I've used at various jobs (which mostly had el-cheapo video cards installed).
What resolution do you run the thing at?
A GeForce2 is also more expensive than a GeForce4 MX, and the GF3 Ti is MUCH more expensive than the GF4 MX.
Is the GeForce4 MX poorly named? Yes. Is it a bad card? Not really when you consider it's price.
And just how hard is it to stick a video card in the AGP slot included with these boards?
Come on people! Take a look at the boards before sprouting off mindless opinions! Every single board in the review had a fully functional AGP slot alongside their integrated video!
I bought a nForce board about a 8 months ago. I play games from time to time, but I'm not worried about getting 200+ f/s on all my games, so the integrated graphics is fine for me for now. In 6 months to a years time, the integrated graphics will probably be getting a touch on the slow side, so I'll probably pick up a GeForce4 Ti 4200 than (which should be nice an cheap after the GF5 comes out) and it should serve me well for at least another year, possibly two.
Well the OEMs dont' seem to agree with you. The nForce chipset is currently being used by ALL of the major OEMs that sell Athlon based PCs. Most also include a low-end SiS or VIA based solution as well, but the nForce seems to take up a lot of the mid-range with integrated graphics and the high-end without integrated graphics.
I know that people tend to really hate it when I post facts instead of blathering out mindless opinions.. however I'm going to go with the facts anyway.
Here's a link to some real and true benchmarks of the nForce 220 and 420 with their integrated graphics as compared to a GeForce2MX in Unreal Tournament and Max Payne.
For UT, the nForce 420 is 9.5% slower than the GeForce2 MX (at 1024x768), and for Max Payne the nForce is just under 15% slower. That's a far cry from "1/3rd the frame rate".
Ohh, and the original nForce was available without integrated graphics as well (the nForce 415 chipset).
Long story short, the nForce can and does play games just fine. It's not getting 200+ f/s at 1600x1200, but as long as you can get by with only 30-60f/s at 800x600, the nForce is up to the task. The nForce 2 should be about twice as fast since it has somewhat higher memory bandwidth and some new stuff which reduces it's dependance on memory bandwidth.
Yes, and for good reason.
Transmeta had something interesting in their code morphing software, but they didn't make use of it. Instead all they ever managed to produced is a butt-slow x86 processor that used an ass-backwards way of getting x86 compatability.
Their only saving grace was that, for a while, they had lower powered chips then the other companies out there, however even that has been eliminated with the ULV PIIIs and the VIA C3 chips. Once Intel's Banias chip is available, that'll be the final stake in Transmeta's coffin if someone hasn't bought them up by that time.
Ohh, and yes, I am aware that I have been speaking of Transmeta in the past-tense. They're a dead company. They'll probably be bought out by someone who has some potential uses for their code morphing software. My guess is IBM, but there are other possibilities.
2. Allocate your engineering resource to make the processor/system run cooler instead, to the point that it no longer need the active cooling measure(fan) on processor and (hopefully) whole board/system.
You mean like the Banias chip that Intel is developing?
3. Make a new small, low power, quiet PC form factor standard
You mean like Tidewater?
4. Make this combo your main production, push other heatmaker to the niche.
Now that probably isn't something that Intel is after. There isn't the money to be made in that market as their is in the high-end CPU market... at least not at this point in time. Things may change.
Anyway, long story short, there is some thought to doing what you suggest, but initially it will be all targeted towards laptop users. Maybe it'll start to bubble down to the desktop as well. I certainly wouldn't mind a TiVo/DVD/DivX type system sitting in my living room, and a full sized PC with a loud fan is not exactly what I had in mind.
The original timeframe was samples in Q2 of '02 and product shipping in the ever elusive "2nd half of '02". It was already pushed back to volume shipments in Q4 '02 or Q1 of '03. Now it's being pushed back again until (probably) Q2 of '03.
On the upside, samples have been shipping for a little while now. This suggests to me that the problem is related more to the manufacturing of the chip then the chip design itself, though that's obviously just a guess at this point in time (and it doesn't exactly matter what the cause of the delay is if the end result is the same).
Where do you think AMD has originations from?
AMD traces it's origins back to Fairchild, same as Intel. Actually AMD and Intel were formed only 6 months apart under relatively similar circumstances. They even setup shop just down the road from one another.
I suppose you could say that AMD was just copying Intel then though, after all, Intel set up in early '69, while AMD didn't setup until mid/late '69! :>
intel puts out something that is more intuitive
More intuitive?! It's a processor! How intuitive or unintuitive can it get? You take the chip, you stick it in a socket!
Actually, no there aren't.
Take a look at SPEC CPU2000 sometime. Ok, maybe it's not the end-all-be-all of processor benchmarks, but it's the best that we've got right now. In terms of raw processing power, we get the following rankings:
CINT2000:
1. Intel P4 2.8GHz - 984
2. HPaq 1.25GHz Alpha EV68 - 928
3. AMD AthlonXP 2600+ - 839
3. (tied) IBM Power4 1.3GHz - 839
5. HP/Intel Itanium 2 - 810
As you can see, AMD and Intel are doing quite well competitng with the rest of the world here, and even beating them. Add an extra 20% to the performance here, and AMD would have a VERY fast little chip on their hands.
Now, in CFP, things look somewhat worse for the P4 and esepcially for the Athlon. This is largely due to two things. First, CFP is a very bandwidth-intensive benchmark, and bandwidth is one of those things that can easily be bought by throwing money at it. IBM threw a LOT of money at bandwidth, and it performs very well.
The second thing with CFP is Sun's new compiler, which magically transformed the rather lackluster performance of the UltraSparc III into a real performer in CFP due almost entirely to a HUGE (8000%) gain in performance on a single sub-test. In other words, their results are not really accurately reflecting the performance of the chip relative to other chips at this time, because they will either perform quite a bit worse then their CFP numbers would suggest, or WAY better if your application happens to exactly match that one sub-test (99+ times out of 100 it would be the former).
Still even here AMD and Intel are reasonably competative. The P4, with a score of 929 is behind only the Power4, the Itanium2 and the Alpha EV68 in CFP. Even the AthlonXP's score of 710 isn't too far off the pace, albeit it is trailing the leading score of 1356 (Itanium2) by a reasonable margin.
As for the Hammer? It's integrated memory controller should REALLY help for CFP, particularly with regards to the Sledgehammer, with a 128-bit wide memory controller. CINT will benefit somewhat less, probably only the 20-30%, clock for clock, over what the AthlonXP (this is what AMD is saying would be typical).
Anyway, the point I'm trying to get at here is that the AthlonXP and the P4 are actually VERY powerful chips, and really there isn't anything that is "leaving them in the dust" at this time.
They'll get beaten until Sledgehammer is released.. not Claw hammer which will have no x86-64 desktop software support right off the bat, and will have to rely solely on it's pure x86 performance.
64-bit performance is unlikely to be significantly different from 32-bit performance. Actually, all else being equal, 64-bit software is SLOWER then 32-bit software (twice as much data to read/write from memory), however the ability to properly access more then 4GB of memory can and does offset that penalty for many applications. In the specific case of the AMD Hammer and x86-64, making use of the 64-bit mode is expected to boost performance by about 5%, due primarily to the fact that AMD doubled the number of registers for x86-64 over IA-32.
Microsoft shafted them on the X-box because Intel paid Microsoft 200 million to use the Pentium III
I don't know any details of the deal between Intel and Microsoft, and in fact, I highly doubt that ANYONE reading and posting here does, just rumors heard from rumor mills. It is clear that Intel gave MS a VERY good deal, though Intel is probably not making any money at all off this deal, especailly given that X-Box sales have been well bellow expectations and profit from this sort of chip would be driven purely by volume.
Nvidia was stuck with an unused AMD integrated chipset for X-box and Nforce was born.
The chipset market is a market that nVidia HAD to get into at one time or another. Integrated video chipsets have grown from being ~5% of the market to being ~60% of the market in just over 3 years time. With an integrated video chipset, nVidia's potential market was more then cut in half. That being said, the nForce seems like a good thing for both AMD and nVidia. For AMD it has given them a stable platform that OEMs trust (as indicated by HPaq using it in their business PCs, something that is VERY rare for non-Intel chipsets). For nVidia, they have a successful product that is likely close to if not the top selling chipset for AMD processors (VIA mostly owns the retail channel, but nVidia has taken almost all the OEM sales).
Intel will pay Microsoft to shaft them again. No x86-64 Windows XP for AMD despite AMD testifying on Microsoft's behalf in exchange for anti-trust testimony
You seem to think that somehow Intel and Microsoft LIKE one another, or that by beating down AMD it would somehow HELP Microsoft? Perhaps you even believe in the whole "Wintel" alliance?! Well I hate to break it to you, but Microsoft wants/needs a company like AMD to give Intel competition. This lowers the prices of Windows PCs as comapred to the Unix boxes, it helps increase their performance and push things forward. Helping AMD makes GOOD business sense for MS, particularly if it doesn't cost them much money. And the x86-64 port of WinXP apparently didn't cost too much money, and they have already demonstrated it on Clawhammer systems.
Probably about the same time as they support Hyper threading and SSE3 for Intel.
WinXP already supports Hyperthreading. Even Win2K will work with Hyperthreading, though it appears like it might have been a touch of a hack (ie the chips show up as two separate processors, not a single processor which can run two threads at once).
As for SSE3, I'm really not quite sure where Intel is going with that, I'll probably have to read up on it a bit more. It doesn't seem to make much sense to me though, they barely got people started on writing SSE code when they released SSE2 and told people to start again. Now they're planning on doing the same thing with SSE3? I'm not even sure that SSE3 will add all that much with is actually useful.. I guess we'll just wait and see.
The whole "Wintel" relationship is mostly just a huge myth. At best, Intel and Microsoft are forced-friends.
Both Intel and Microsoft have realized for quite some time that they need each other to survive, at least for the time being, but neither of them are happy about that fact, and both are striving to change it. Microsoft has been a fairly strong supporter of AMD for some time now, while Intel was one of the first major hardware companies to jump on the Linux bandwagon.
Think about it, Intel is pretty much the dominant force in PC hardware, being not only the primary supplier of processors, but also the #1 supplier of motherboard chipsets, video solutions, and being well up there if not #1 when it comes to audio and NICs (most of this is now integrated into motherboards). They are also the driving force behind most buses and interconnects, ie PCI and AGP (AMD's Hypertransport is a notable exception here), not to mention the fact that they defined the ATX form factor that virtually all current PCs make use of (albeit in a somewhat basterdized format for some of the big OEMs).
Microsft, on the other hand, is the dominant force in software, having the most common operating system, office software, web browser, e-mail program, etc. etc. ad nauseum.
The end result is that Microsoft is the only company in the PC world that Intel doesn't have a fair degree of control over (AMD fights, but ends up adopting Intel-compatible technology in the end, ie MMX/SSE/SSE2). Similarly Intel is the only company that Microsoft doesn't exert a signicant amount of control over in the PC world, though even there I'd say that MS mostly won that battle a while back.
Anyway, long story short, MS and Intel, not so much allies as simply two companies forced together by circumstance.
The P4 does indeed do 4 transfers per clock on it's bus. It does this by having two clocks 90 deg out of phase with one another, and sending data on the rising and falling edges of both clocks. It's definitely trickier, from both a conceptual point of view and from a design point of view then DDR, but it's not magic, and it's definitely not just a marketing term to try to hide things or anything like that.
FWIW AGP 4x mode uses the same basic technology.
Ok, the C3 may not be besting the Athlons and P4s of the world, but it's a whole heck of a lot faster then you're making it out to be!
The C3 at 800MHz will have no trouble at all decoding DivX, let alone mpeg-1. Many people have quite successfully used these chips in set-top boxes for playing DVDs and DivX movies.
Ok, totally off-topic, but something I've been asking for a while, where the heck are PDAs based on PalmOS 5.0?!
The operating system was released back in June, and (silly me) I figured that PalmOS had worked with PDA manufacturers (or at the very least Palm themselves) beforehand so that actual products using this new operating system would be out shortly there-after. Well it's now 3 months later, and the best I've heard are some far-off rumors that the new handhelds MIGHT be coming sometime sorta not too long from now, like next year!
Come on Palm, Handspring, Sony et al. EVERYONE knew basically what was going to be required for this new OS LONG ago, why is the hardware lagging so far behind the software?! It's not even like the hardware would require that much in the way of development, since any current PocketPC PDA should be able to run PalmOS 5.0 with virtually no hardware modifications.
Anyway, that's just my little rant for the day.
AMD tried that, and their popularity dropped, profits dropped and they were barely able to sell their top-end chip for $100. Apple is still trying it, and still failing miserably (though Apple has the unfortunately downside of actually have very slow chips, regardless of whether you look at their clock speed).
Face it, the few that know enough to bother looking beyond a single number know enough to realize that clock speed is a completely useless number.
Personally I don't really care too much one way or the other. AMD replaced one totally meaningless number with another totally meaningless number. Ideally this sort of thing would just encourage people to actually look beyond the pretty number and try to figure out what it actually means, but both myself and AMD's marketing dept. are well aware that that sort of thing is WELL beyond what 99% of the buying public are going to do. So, model numbers it is, and they worked. AMD's profits and average selling price have increased a lot since the indroduction of their model numbers, even now in a bit of a PC market downturn.
Well, 10 years ago everyone was saying the EXACT same thing about how we really didn't need these new fandangled Pentiums, and that a 486 was really fast enough for everything, and that in 10 years time we would all be laughing at the people who thought that they needed a 100MHz processor!
The more times change, the more they stay the same. 10 years from now, we'll all be looking back at these AthlonXP 2700+ systems and wondering how we managed to get by with such SLOW processors, because runs REALLY slowly on anything less then a 10GHz chip.
It's been happening for the past 20 years, and I don't see anything today that makes this look different.
I don't know that the Barton core will use any lower latency, or lower power at all.
The Barton core won't actually be manufacturered by AMD, "Barton" is the code name that is now being used for the AthlonXP processors that UMC will be building for AMD. First off, UMC is still a few months away from shipping chips. Secondly, they might not be able to manage as high clock speeds as AMD manages in their own fabs. AMD's Fab 30 in Dresden (where the Athlons are currently all produced) is quite a high-end fab that is really tuned for high performance/high clock rate chips. UMC's fabs tend to really emphasize low cost/high yields, so they might not get the same sort of bin splits that AMD can get on their own.
The real benefit of the Barton core should be for multiprocessing setups, where the extra cache should really help out a lot. For single processor setups, I doubt that we'll see that big of an improvement over the current Thoroughbred core, probably only about 3-5%, or about the same as the improvement seen when increasing the bus speed from 133/266MHz DDR to 166/333MHz DDR.
Well, FWIW the Thermal Design Power of the P4 at 2.8GHz is 68.4W. This number is kinda-sorta-but-not-quite the maximum power that the chip will consume (I've ranted about this in other threads :> ). AMd's AthlonXP 2600+ has a maximum real-world power draw of about 68W as well (AMD does document the real exact figure, but I don't have the data sheets in front of me and they don't open properly in Mozilla anyway due to a bug in the Acrobat plug-in... but again, that's for another rant).
So, is AMD going to have to redesign their motherboards to accomodate this new chip? Well, maybe and maybe not. The Athlon 1.4GHz processor is the highest power consuming chip of the Athlon line to date at 70W maximum, and that's probably going to about what an AthlonXP 2700+ will require. However the 2700+ will get that 70W at a lower voltage but higher current. Long story short, it depends on the motherboard. AMD tends to leave this sort of thing more up to the motherboard manufacturer's designs, while Intel kinda forces motherboard manufacturers to do everything the Intel way. So, chances are that some AthlonXP boards will be able to accomodate this chip without change, while others will need a new voltage requlator.
Processor myth #1: AMD chips generate more heat then Intel chips.
Check the data sheets sometime, the amount of heat that an AthlonXP and a P4 put out is nearly identical. Both are also only in the 60-70W range, or about the same as your typical light bulb. We all know how turning on a single light in your house cause your AC bills to skyrocket!!!