backtop

Hardware AMD Preps "R600" Die-shrink
Gabriel Ikram (Blog) - September 27, 2007 11:04 AM
Print E-mail del.icio.us 48 comment(s) - last by murphyslabrat.. on Oct 1 at 1:37 PM

The non-working MSI RV670 sample with heatsink.  (Source: Coolaler.com)

Although the GPU is a non-working shuttle, the memory layout still sheds some hints on ATI's plans for RV670.  (Source: Coolaler.com )
New ATI documents and images shed more light on AMD's newest processor core

AMD continues to develop its new RV670-based cards in the wake of the recent AMD ATI Radeon HD 2900 PRO release. The RV670 GPU is the current code name for a mid-range graphics core slated to fill the gap between the current Radeon HD 2600 and Radeon HD 2900 offerings. 

On the surface the RV670 core is very similar to the R600 core featured on Radeon HD 2900, though where the R600 was manufactured on TSMC's 80nm process, RV670 will feature a 55nm process node.

According to early board partners, including MSI, the GPU comes clocked at 600MHz and features 320 stream processors -- the same featured on the Radeon HD 2900 series. However, test boards make use of 256MB of Hynix GDDR3 900MHz memory -- R600-based boards feature 512MB and 1GB GDDR3 or GDDR4.

Photographs of the card show a single-slot cooler. Despite the single-slot cooler, the card is quite large and appears to require a 6-pin PCIe connector. 

AMD guidance claims the RV670 die includes integrated Shader Model 4.1 capability along with PCI Express 2.0 support.  Corporate guidance does not mention support for Crossfire, but newest images of the card clearly reveal interconnects across the top of the board found on other Crossfire cards.

As of right now the card is expected to make its way into AMD’s lineup in Q1 2008.  Vendors have not issued pricing for the card, partially due to the fact NVIDIA is also expected to launch new high-end adaptors before the end of the year.

Comments     Threshold


This article is over a month old, voting and posting comments is disabled
What the heck
By Polynikes on 9/27/2007 12:17:17 PM , Rating: 4
Why don't they make a high-end enthusiast card with the die-shrink? Maybe then they'd give Nvidia a run for their money.




RE: What the heck
By m3rdpwr on 9/27/2007 12:32:34 PM , Rating: 5
My guess is because mainstream parts sell much more.
High end cards are a niche market.

The die shrink on mainstream cards would be more profitable and perhaps they could lower the price as well...

Just my thoughts...

-Mario


RE: What the heck
By ira176 on 9/28/2007 2:07:08 AM , Rating: 2
Mario,
I agree that a die shrink will surely increase profitablity for AMD and they should still be able to make a profit with a fairly priced card which performs well. My other thought on this is that, I hope that they will include the Unified Video Decoder on the chip for dedicated HD video playback.


RE: What the heck
By herrdoktor330 on 9/29/2007 10:01:18 PM , Rating: 2
I second that about the HD Playback!

Just as long as they're not forgetting to work on their linux drivers while they do this. I know that the open source driver issue isn't exactly associated with this topic, but I hope they don't forget about that one. I'd be more apt to buy ATI if they had better linux drivers. I've been their customer before when I bought a 9800 pro card. I'd be more than happy to buy their product again and once this generation of the product is refined, I will consider buying again.


RE: What the heck
By JAB on 9/27/2007 12:38:13 PM , Rating: 2
I doubt they can compete on the top end yet but if they pull off that die shrink it will be serious competition for the mainstream market. Not exciting but Both Nvidia and ATI have learned the lesson of not bringing out new top end cards and a die shrink at the same time the hard way. That is rarely pretty.


RE: What the heck
By murphyslabrat on 9/27/2007 12:59:44 PM , Rating: 3
What would save AMD's graphical neck (as opposed to a text one?) is not a top-end card, that 2% of consumers will purchase (wild guess), but a better price-performance competitor. You have here a card that, while cheaper, performs relatively well, and you succeed. And, at least for mainstream games, 256MB is not that bad, especially if they're going to be selling at $200-ish. A 256-bit BUS? A rumor, but would still be ~okay~. I would expect more, but still...

I, for one, had been planning a March '08 system-rebuild, for which this aught to be perfect. Maybe I will be able to get a decent computer for $700.


RE: What the heck
By wordsworm on 9/27/2007 1:49:48 PM , Rating: 5
You're missing a big part of the equation. While your logic is sound on the basis of a smart consumer, that smart consumer is probably between 2~5% of the market (my own wild guess). They only listen for one thing: who has the best ( fill in the blank ). They then assume that that's the company to buy from, regardless of whether they're picking a high or low end product. My boss is about to open a new English academy (for about 200-300 students), and a part of that includes a small computer room containing 8 computers for the students. He thinks he ought to buy Intel machines because they're the best. On the low end, anyone who's an enthusiast (I don't necessarily mean the high end - could be the low end) knows that AMD has that market. It's possible to buy an AMD Sempron 3000+ (39 watts) machine for about $150 with 512MB of RAM, 80GB hard drive, motherboard with video built in. The alternative is a much inferior Celeron. That Celeron (89 watts) will cost, per computer (based on 12 hour per day usage), about an additional $10-15 per year because of the difference in power usage. Furthermore, that AMD chip will outperform its Celeron competitor (if someone has stats to prove me wrong, please, by all means - I haven't actually tested them. I'm just going on 'gut instinct' The Sempron was peppy on basic MS productivity software, whereas the Celeron always seems to drag its feet).

He has a friend who's big on networks and he's an IT guy. This guy can run circles around me in probably every area except hardware. (He didn't even believe me when I was mentioning a then 3 week old 8600 Nvidia series). He just tells him that Intel and Nvidia have the best platforms. While this is true on the high end, it's not so clean cut on the low end. We all know that AMD rules the bottom. I have an Intel E2160 at work and an AMD BE2350 at home, and the integer calculations are about 10% faster on the AMD, and 40% faster on floating point, all for about the same cost (about $82 Intel - 89 watts - and $85 for AMD - 45 watts) and much lower watt consumption. That $3 dollar difference will be eaten up in about 3 months. My computer will last me probably 3 years, and so that amount adds up to $36, not to mention the air conditioning months which again adds about $3 for about $9 to the life span of my machine. That $45 might not seem like much to most folks here, but it counts in my book. With that money I can buy the blue ray/HDVD combo player when it comes out for $99 at Walmart in about 6 years. :)

Being an enthusiast isn't always about what's the best at the high end, it's about getting the best for the money that you're ready to spend. People like us, the enthusiast, are keen on this, but the majority of the people, though buying on a lower end than I do, only pay attention to the few words they hear about the high end.

My point is whoever has the best stuff wins bragging rights that win customers who are ignorant. My boss will never even consider buying a mid end machine, let alone a high end machine. Regardless, he's itching for a Celeron because his friend tells him Intel or Nvidia makes the best chips. He's not a minority, he is just one of many lemmings. It might sound like I am disrespecting him, but he's a pretty smart guy. He just isn't an enthusiast like I am. I will probably be able to save him from his mistake. He's cheaper than I am.


RE: What the heck
By murphyslabrat on 9/27/2007 2:02:39 PM , Rating: 5
quote:
He's not a minority, he is just one of many lemmings.

It is true, you are right. And, I guess, it's just part of getting things done. Unfortunately for AMD, it means they're screwed either path they go.

On the flip side, the same Lemming mentality did get AMD 60% of the oem graphics market because of "Full-featured HDMI," that happens to include the exact same sound chip as is on modern mobos.

So, lemme change my previous statement: I am glad AMD went this way, so that I can get a better video card for my money. ;j


RE: What the heck
By plewis00 on 9/27/2007 4:03:22 PM , Rating: 3
Whilst this is obviously massively off-topic from the AMD die-shrink discussion - you state the Celeron has a large power dissipation of 89W, this is wrong firstly, even for the Celeron D (Cedar Mill) which was rated at 65W. Anyway, these days the current Celeron chip is the 4xx-series based on the Conroe-L core and hence the Core architecture. It's rated at 35W and will slap the Sempron into oblivion within that price bracket. It's also cheap, priced between $40 and $60 depending on the model.

I don't know where you are producing this 89W figure from but it's wrong, for both the current Celeron line-up and Pentium Dual-Core E2xxx series, it's 35W and 65W TDP. So for one of the first times in this industry, Intel should have some kind of right to the lower-end market. As for your BE chip, all it is is a binned chip, you can undervolt the Core 2 chips as well and get a similar result.


RE: What the heck
By wordsworm on 9/28/2007 6:51:39 AM , Rating: 2
quote:
Whilst this is obviously massively off-topic from the AMD die-shrink discussion
I was mostly discussing customer psychology and how having the top notch product has an effect for all sales on the broad spectrum, and not just on the top tier. This was in response to the previous post suggesting that people would see the true value in the AMD product because of its merit. I counter argued that the merit in the component never really reaches the mind of the consumer.

As far as comparing Sempron to Celeron, I'd be interested in reading comparisons between the 2. I did a fair amount of searching for good information.


RE: What the heck
By murphyslabrat on 9/29/2007 4:43:33 PM , Rating: 2
quote:
you state the Celeron has a large power dissipation of 89W, this is wrong firstly, even for the Celeron D (Cedar Mill) which was rated at 65W.

I am not going to argue the merit of either method, but AMD and intel use different states of usage to find the TDP, or ?total disipated power?. AMD measures absolute stress, where Intel measures maximum normal stress. And, an AMD TDP of 89 watts is about the same as an Intel rating of 65 watts. It is not exact, but it is fairly close.

Again, I am not arguing merit, but that is where he was coming from (or at least what I assumed Wordsworm to be talking about).


RE: What the heck
By coldpower27 on 9/28/2007 9:11:16 AM , Rating: 2
I am curious where you got the information for the Celeron as it is just plain wrong in terms of power consumption. Intel doesn't have 89W TDP's for it's desktop processors that is a AMD TDP value.

It's also perfectly possible to buy a cheap system for Intel using a cheap Celeron.

Celeron 420 45USD
80GB GD 37USD
G31 Chipset 70USD
512MB DDR2 15USD.

The AMD Sempron has a TDP of 62W at least is I remember correctly. You would have to show me a direct link for proof if your talking about the SFF Sempron models. The SFF models in general are pretty rare, you can find the 45W or 62W stuff pretty easily.

The pricing of the BE-2350 and E2160 are fairly comparable I can agree in retail, as I just cross verified it on Newegg which has both processors.

The difference between the E2160 and BE-2350 is only 20W in terms of thermal design power, however real world testing shows there isn't much of a difference once you compare them on the same platform.

http://www.techreport.com/articles.x/12872

As shown there the E4300, (which is the the same core as the E2160 but has the rest of the cache activated) is only 4-7W away depending which situation your looking at.

Don't try to impress your ideals on others or how you spend money as the only right way, people who buy the best want the best and are willing to pay for it. Not to mention Nvidia and ATI know having a strong high end has the effect of filtering that perception to the mainstream.


RE: What the heck
By Melted Rabbit on 9/27/07, Rating: -1
RE: What the heck
By murphyslabrat on 9/27/07, Rating: 0
RE: What the heck
By wordsworm on 9/27/07, Rating: 0
RE: What the heck
By Lord Banshee on 9/27/2007 2:41:20 PM , Rating: 5
Thats not High End that is Pro End... I hope you know that?

FireGL series are doing very well this generation but the topic is about gamer cards which FireGLs do no difference than its Radeon counter parts in games. CAD applications bottlenecks are not in shaders.

As stated above, ATI does not have any card as fast as the 8800Ultra in the gaming world.


RE: What the heck
By murphyslabrat on 9/27/2007 3:19:08 PM , Rating: 4
quote:
Why do people pay $2,000+ for ATI products then?

The answer? For absolute error-absence. The purpose of the FireGL and Quadro FX is for absolute stability. Actual performance has little to do with the price. For instance, according to Hardspell.com, The Radeon HD 2900xt 1GB performs almost identically to the FireGL V8600 1GB. The FireGL card costing almost $2000.
http://www.hardspell.com/english/doc/showcont.asp?...
Now, there was very little improvement with the step from 512MB to 1GB for the Radeon HD 2900XT, so I imagine that the performance would change even less when going from 1GB to 2GB's. If you can contest this with a citation, please do.
http://techreport.com/articles.x/12956/5
quote:
Quote: Surely, even ATI wouldn’t believe that its FireGL V3600 is three times faster than Nvidia’s Quadro FX 1500 under UGS NX, for example!

I wouldn't either, as you failed to produce a citation.
quote:
Let's face it, on the ultra high end, AMD has the best in the business.

You never gave any performance figures, just price. Just because I am paying nearly $3000 for FireGL 8650, doesn't mean it will beat out the $530 Radeon HD2900XT 1GB by any appreciable amount. Post some benchmarks, and I could change my tune, but until then I maintain my stance that AMD does not have the top. This is not to say that AMD cannot compete, just Intel owns the "ultra" tier.


RE: What the heck
By Lord Banshee on 9/27/2007 5:02:24 PM , Rating: 3
Did you miss the specview benchmarks from your link?

But again... people need to stop comparing Pro cards and Gamer cards. In games they perform 95-100% identical. In CAD applications the Pro cards get 200-400% better performance due to driver features turned on via the drivers seeing the correct card installed, FireGL or Quadro.


RE: What the heck
By wordsworm on 9/27/2007 8:39:59 PM , Rating: 2
quote:
you failed to produce a citation.
http://www.beyond3d.com/content/pr/62/
How could you miss my citation? The guy said he didn't want to provide any benchmarks because he couldn't believe his eyes.

I followed your link to hardspell.com and found SPECviewperf reports to be in the 2-3x better range. Unfortunately, like you, I couldn't find any publications showing how this 'pro' card from AMD could do in a gaming situation. It's too bad Anandtech doesn't test something like that just for the sake of argument.


RE: What the heck
By coldpower27 on 9/28/2007 9:28:02 AM , Rating: 2
Did you not see the benchmarks in 3D Mark 2006 in comparison the HD 2900 XT, it performs effectly identically. In gaming there is pretty much no difference.

The professional cards have much higher performance in rendering and that is what your paying for, not the gaming performance, businesses use these cards, not home users.


RE: What the heck
By murphyslabrat on 10/1/2007 1:37:07 PM , Rating: 2
quote:
http://www.beyond3d.com/content/pr/62/
How could you miss my citation? The guy said he didn't want to provide any benchmarks because he couldn't believe his eyes.

I am not trying to be a jerk, but I can't find what you are quoting. Could you give me a paragraph number, or tell me where he makes his link?


RE: What the heck
By theapparition on 9/28/07, Rating: 0
RE: What the heck
By Melted Rabbit on 9/27/2007 2:50:34 PM , Rating: 2
Hmm... the grandparent comment I made should have included some benchmarks to back it up: http://www.xbitlabs.com/articles/video/display/dia...

Also to the parent, I wasn't talking about just about single card performance. I was talking about the nearly absolute high end performance in Vista. It seems that many reviewers at one point used a dual card setup, because it favored nVidia. At least it used to be before SLI was broken on Vista. For that matter ever notice that many reviewers, HardOCP for one, don't bench dual card performance in Vista at all. The review becomes not as useful and even a bit misleading as it does not include the dual card benchmarks.

Also, don't change the metrics on me, you shift the comparison from a dual HD 2900XT 512MB versus a single 8800Ultra to a single HD 2900XT and a single 8800GTX. Not a reasonable shift.


RE: What the heck
By murphyslabrat on 9/27/2007 3:32:58 PM , Rating: 3
quote:
Hmm... the grandparent comment I made should have included some benchmarks to back it up...

That still supports my statement: The 8800GTX vs. 2xRadeon HD2900Xt (1GB or 512MB) is a fairly even match, with the occasional benchmark or game in which one crushes the other.
quote:
Also to the parent, I wasn't talking about just about single card performance. I was talking about the nearly absolute high end performance in Vista

Mind you, NVidia just released beta drivers that support SLI in Vista. So, the option is now there to compare 2x8800Ultra. In addition, I was also talking about absolute performance roof.
quote:
Also, don't change the metrics on me, you shift the comparison from a dual HD 2900XT 512MB versus a single 8800Ultra to a single HD 2900XT and a single 8800GTX

Not at all. I added that to the comparison, as well as adding 8800GTX vs 2xHD2900XT. A perfectly reasonable shift, and one that, again, illustrates the inaccuracy of your position: that AMD beats out NVidia in the ultra-high end.


RE: What the heck
By 3kliksphilip on 9/27/2007 3:27:21 PM , Rating: 2
I think it's a great idea! They mucked up the X2900 XT (At least, in my eyes). If they shrink the PRO card, it will be faster, cooler and more efficient, leading to more over clocks and BINGO! A card that's superior to the XT in every way. For a mere £170 in England (About $340 Dollars) people can have a card which can compete with the Geforce 8800's, with the added advantage of an extra .1 on the DX version.


RE: What the heck
By murphyslabrat on 9/27/2007 3:46:46 PM , Rating: 2
The added bonus is, of course, that with a smaller node comes cheaper production costs, which in turn means that we could see price drops in the months after release.

Plus, that extra DX revision has got to count for a 50FPS boost!</joke>


RE: What the heck
By aznstriker92 on 9/27/2007 7:09:26 PM , Rating: 2
Dont worry, havent you heard of the R670dual card? basicly its a card with two R670s on it.
Thats going to be the highend for ATI


RE: What the heck
By erikejw on 9/29/2007 6:32:28 PM , Rating: 2
For the same reason AMD and Intel don't use new cores together
with a dieshrink. It is hard enough to get the parts to work in the first place and to add a die shrink adds to the complexity tenfold.


By Mitch101 on 9/27/2007 1:05:53 PM , Rating: 1
NVIDIA will be dead or hurting massively in 2-3 years unless they merge with AMD

http://freshscoop.com/modules.php?name=News&file=a...

If AMD had enough fab to produce the chips inhouse than NVIDIA would be dead quick but even still being able to leverage 2 products CPU and GPU using a team that already has mastered the die shrink of the CPU on the GPU. NVIDIA is in deep trouble sooner than later.




By rogard on 9/27/2007 1:47:38 PM , Rating: 2
Hm. The article behind the link does not impress me too much. We have known intel to be the "big bad wolf" for years now, and if they really take on something, they have all the resources in the world. A deadly foe. On the other hand, it seems to that it is not an easy feat to just produce a competitive graphics chip/card, just like that. If it were easy, SiS, Matrox, Tseng and last bút not least AMD would have the latest and greatest cards on the market. Funny they don't.

Your argument is not valid, because if nvidia had huge inhouse capabilities, they would be even better off right now, flooding the market with cheap parts if necessary or enjoying the benefits of having the fastest product that sells for the best price (as they do now) even more so. What is your point again?

Right now, neither AMD nor nvidia have their own fabs, but at least nvidia has motherboard chipsets that sell, and they have had a huge advantage on the gfx market over AMD. We all know that this could change within a year, as it did in the past (remember Ati 9700? That was something!)

I really cannot see the benefit of a merge of AMD and nvidia, not for the customers and not for the companies. What will happen in 3 years, I do not know. Maybe intel really decides to stop making shitty graphics cards one day, who knows?


By Mitch101 on 9/27/2007 2:25:12 PM , Rating: 2
ATI produced 55nm production ready chips on the first try. This has never happened before with ATI. So you dont feel that the AMD engineers had any influence on getting it right the first time around after dealing with the die shrink of the Phenom or even the Opterons at 65nm?

Might as well say if BitBoys had huge inhouse capabilities.

AMD has fabs they just dont produce enough chips to support the graphics lineup and are still going to outsource it.

You also feel Intel engineers must be less capable then SIS, Matrox, Tseng? Anyone worth thier money in the buisness of graphics is working for AMD/ATI, NVIDIA, and Intel now there are already a few confirmations of former ATI people who didnt stay with the merger who went to Intel. Somehow I think Intel could have had some internal people plenty smart enough to be able to come up with something amazing now that Intel is seriously getting into graphics instead of seeing it as a niche market. I should add that Intel doesnt have to back support a graphics line it can go from the ground up and develop drivers that work on one serious card.

Intel is not going to go out of buisness at the hands of NVIDIA. Intel will allways be there. They have never focused on graphics as much as they have recently. Its going to happen. Intel has 45nm silicon and has shown 32nm silicon. Thats 2 steps ahead of NVIDIA in manufacturing.

You didnt read through the link. Intel is making some of the best motherboards they have ever made. The best overclocking motherboards are not NVIDIA but Intel.

The fact of the matter is NVIDIA needs to start thinking of a backup plan now instead of 2-3 years from now because mergers take time for the products and benefits to merge together. Merging when they are at a peak might be better than when Intel has a competing product sucking revenue from NVIDIA which is still capable of making money.


By PlasmaBomb on 9/27/2007 4:02:01 PM , Rating: 2
quote:
ATI produced 55nm production ready chips on the first try. This has never happened before with ATI.


quote:
RV670, ATI's first 55 nanometre chip turns out to be better than anyone inside ATI expected. ATI planned to have the chip ready in very early Q1 next year but for the first time in ATI's history the A11 silicon doesn't need a re-spin.

Traditionally A12 was ATI's production revision but it looks like that A11 is bug free and that it is ready for production.


A11 is hardly the first try


By MetaDFF on 9/27/2007 5:04:15 PM , Rating: 2
quote:
A11 is hardly the first try


Actually, A11 is the first try.
A12 refers to the first metal spin.


By PlasmaBomb on 9/27/2007 4:11:46 PM , Rating: 3
quote:
Intel has 45nm silicon and has shown 32nm silicon. Thats 2 steps ahead of NVIDIA in manufacturing.


They showed 32nm SRAM, which is essentially a test, they haven't got their 32nm process ready yet. Intel is also ahead of AMD/ATI


By Master Kenobi (blog) on 9/27/2007 4:25:30 PM , Rating: 2
Intel has always lead the graphics market by a huge margin. It's just that graphics doesn't just mean discrete. By stepping into the discrete graphics market in a serious fashion they can further increase their graphics marketshare lead and besides the margins are a lot better in discrete graphics and Intel wants a piece of the action finally.


By rogard on 9/27/2007 4:38:07 PM , Rating: 2
Well, Sir, did I say anything about nvidia putting intel out of market or even business? No. That's ridiculous. Read my text again.

Did I say, intel was less capable than the other companies? No. Read my text again.

What I said is, that so far nvidia is doing just fine, and if the had their own fabs (or more capacity in general) they would even be doing better. Read my text again.

Sure, every companie in every business in the whole world needs to think what they'll be facing in 2 or three years time. What a cliché. So does, intel, so does nvidia, so does amd and so on.

It certainly escapes me what making overclocking boards has to do with nvidias "certain death" in 3 three years.

I did read through the article, and I agreed already that intel is a power one should never underestimate. But nonetheless, they cannot conjure up something out of thin air (If they could, they would. Would be much cheaper...)

All I said is that the article didn't impress me much, and that I didn't see it as a fact that nvidia is "doomed".

I don't know about your connection to intel or amd or nvidia, but I am sure they don't need you to defend or advise them, they will just be doing fine whatever I or you think. If they should merge, I'd be surprised, I admit it, but what the hell? Seems to me you don't like nvidia for whatever reason, but I don't care, really. Have a nice nice day.


By MultiSync on 9/28/2007 2:03:48 PM , Rating: 2
Nvidia really is killing themselves by only enabling SLI on its own chipsets. Now ATI is own by AMD and Nvidia not letting them use SLI(possibly other things in the future as well like physics), Intel has no choice but to enter the discrete GPU market on their own. With their own fabs intel will slaughter ATI Nvidia on the low mid range markets. What Nvidia should do is to try to get into a join venture of some sort with intel in the future. If not they will be just be push into the niche market very soon.


By wordsworm on 9/27/2007 3:03:36 PM , Rating: 2
quote:
Funny they don't.
Just to reiterate a previous comment I made: AMD owns the high end market. They sell a $2700 video card with 2GB of RAM. Apparently it's one of the first video cards to auto detect whatever software the user is using so that it can optimize itself, even when there are multiple applications running. Scroll up and you'll catch the link to the article that talks about this.

quote:
Right now, neither AMD nor nvidia have their own fabs
This is the comment that really got me. AMD does indeed have its own fabs. I really don't know why you got this in your head. They have 2 and will have significantly upgraded one of them to 65nm/300mm in time for Phenom to start penetrating the market: http://www.amd.com/us-en/Corporate/AboutAMD/0%2C%2...
Even if Phenom is the best processor ever, they simply aren't ready to facilitate its production. It only makes sense for them to wait until they have Fab 38 ready.

quote:
I really cannot see the benefit of a merge of AMD and nvidia
I agree with you 100% here. I can't see how this could even be legal. It would probably be deemed anti-competitive. Sometimes companies are blocked from becoming what would be a virtual monopoly. There's no way that AMD can acquire Nvidia. I think Intel was flirting with the idea but discarded it when they realized how expensive it would be. If it's too rich for Intel, then certainly it's too rich for AMD.

AMD is going to own the market shortly, I feel pretty confident in that. AMD shares have clawed their way back up to $13.75/share (up $2 from 1 month ago). I call that 'cautiously optimistic' on the commodity floors. With their new production facility (Fab 38) on the verge of opening I can't help but feel confident that they're going to bounce back in a big way. I don't know if they'll ever dominate Intel (in market share I mean), but there's every chance that they might regain the performance crown in the next 6~12 months. Intel products seem to be improving incrementally. Furthermore, they seem to be taking short cuts (dual/quad+) that I can't help but foresee as being something that'll bite their butts in the long run. Already AMD is snubbing their noses at Intel with their 3 core on the way. ie, 9 cores make a lot more sense to me than 8 cores. 3x3 rather than 2x4. With AMDs foray into controlling power so that each core has its own control is simply brilliant.
quote:
With Barcelona/K10, each core can change speed as needed potentially saving a lot of power. More interesting is the memory controller power savings. New to K10 is the ability to manage DDR channel power. If K10 sees that memory is using only writes, it can shut down read channels. If it is only reading, it can shut down write channels. This again saves a chunk of wattage.
http://www.theinquirer.net/en/inquirer/news/2007/0... All the people who see Intel pulling ahead of AMD simply don't seem to recognize how innovative AMD is actually being with their ingenious designs, and how conservative Intel seems to be with their increments. I haven't seen anything out of Intel, aside from die shrinkage, that really says 'amazing' or 'revolutionary'. Instead, I just get 'more of the same' or 'faking it'. For now this is working, but as I said earlier, it's going to hurt when everything that AMD is doing gets into motion. I'm not a fan of AMD or any particular party. I am a fan of innovation and invention. Right now, those words seem to fit more appropriately to the AMD camp. They are reading the future, and Intel just seems to be doing more of the same. Ultimately, it's short term gain for long term pain.


By erple2 on 9/27/2007 4:37:20 PM , Rating: 2
quote:
Just to reiterate a previous comment I made: AMD owns the high end market. They sell a $2700 video card with 2GB of RAM


Understand, this is in the Professional Level OpenGL market, not the consumer market. I could find a LOT of other non-nVidia, non-AMD video solutions in the professional world that will outperform any card from nVidia and AMD that will cost many tens, if not hundreds of thousands of dollars. Granted, they're usually custom one-off devices...

AMD does NOT own the high end market - I have yet to see anything that suggests that AMD's V8650 (the card they're touting) can favorably compete with nVidia's current high end Professional card, the Quadro FX 5600. In fact, looking at the press release specs, it sounds an awful lot like a 2900 with 2 GiB memory. I'm sure there are some other minor tweaks in the silicone, but nothing substantial.

That STILL does not suggest that AMD "owns" the Top End, even in the Pro Graphics Market. Hint - The 8800 Ultra, or the 2900XT 1GB model are still only in the Consumer level Market.

The fact of the matter remains that at the high end in the consumer market, there's a funny tradeoff - in the single graphics card market, nVidia is clearly king. In the dual graphics card market (winXP only), nVidia is still clearly king. It's only in Vista that things get a tad muddy. EVENTUALLY, nVidia will figure out how to get SLI working in Vista. Until then, it's kind of a tossup as to who has the better solution, crossfire 2900XT's, or a single 8800 Ultra (though leans closer to ATI winning).

You MUST understand, that the FireGL and Quadro brands are NOT CONSUMER LEVEL CARDS. Those are NOT "High End Consumer Cards". In fact, in games, they won't outperform the current top end Consumer Cards. Why, then, do they cost more, you ask? They are generally only sold to businesses, where margins are a LOT higher.


By theapparition on 9/28/2007 12:48:55 PM , Rating: 2
Stop with the facts, Wordsworm's argument is clearly based on information he pull out of his ass, so why be any different?

v8650 is faster than anything.
Not a single benchmark to validate that.

AMD has its own fabs.
True, but no graphics products will be made on them, so for the record, the ATI product line still uses outside fabs.

See, why let facts get in the way of a good arguement?


By rogard on 9/27/2007 4:47:46 PM , Rating: 2
I thought we were talking about graphics cards and not cpus. As far as I know the gfx chips are not produced in amd fabs. My bad if I was wrong there. I know that amd owns fabs for cpus, one is right around the corner in Dresden. I have been there, driving by and having a look from the outside.

I did say nothing about intel pulling ahead, or amd being dead, or nvidia crushing intel or whatnot. Please, read me text. Thank you.

To make that clear, I have nothing against amd, and I always liked the David vs Goliath thing. I wish them the best and I hope they all put up a good fight in the future.
The last thing we as consumers or PC affiliates need is a monopoly and competition that is snuffed.


By Chimpee on 9/27/2007 3:42:27 PM , Rating: 2
Would be more correct to say ATI doesn't currently use AMD fab but still uses TSMC.


That's a tiny die!
By Shadowmage on 9/27/2007 12:12:16 PM , Rating: 2
According to rumors, this is a 320/16 part just like R600, but features a smaller 256-bit memory bus. It's amazing how small that die is: it's roughly as big as the X1600 die!

This should be a power/performance monster :)




RE: That's a tiny die!
By Dactyl on 9/27/2007 1:17:15 PM , Rating: 2
Sounds like a good choice for Crossfire on a single card.

What I'm wondering is, how many stream processors will their high-end 55nm card have? It sounds like 640 is within reason.

It looks like if AMD had just a little bit more imagination, they could have taken the graphics crown away from NVidia.


RE: That's a tiny die!
By Slaimus on 9/27/2007 3:32:44 PM , Rating: 1
Have you seen the die shot on the R520? The HUGE area was taken by the memory ring bus. The 512-bit ringbus must have been massive.

Of course, from 80nm to 55nm is also a double shrink (80->65->55), so that has a lot to do with it as well.


256megs????
By Sunrise089 on 9/27/2007 2:45:23 PM , Rating: 2
Great - there goes the reason to pass on the 2900Pro. If the card is really only 256megs, the "overclock and have a X2900XT" arguement is shot. In fact, I'd probably not even consider it over a 320megGTS - the market is definitely moving towards 512megs being more or less required.




RE: 256megs????
By GlassHouse69 on 9/28/2007 12:16:05 AM , Rating: 1
the reason it has less megs of ram is so that you cannot get good results with it.

if it had 512 or 1 gig you would overclock it for a much better wattage/performance solution. memory bandwidth is meaningless. what is meaning full clockspeed and how many stream processors are running. Then you need 512+ of ram to keep it all happy. hobble that in any way, you just made a typical "miduser" piece of shit.

piece o shit.

I hope its good though as ican only afford a piece o shit this time around (until a 9800gts 640 vanilla with die shrink becomes below 250 range)


Q1 2008? Boooo
By Assimilator87 on 9/27/2007 12:11:51 PM , Rating: 2
I thought this card would be out way before the end of the year. I'm kinda looking for something under the $250 price bracket, which the 2900 Pro would fit into.




We don't know how to make a $500 computer that's not a piece of junk." -- Apple CEO Steve Jobs

DailyTech Poll
Which web browser do you use on your primary personal machine? 






44 Comments









botimage
Copyright 2009 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki