Thursday, June 26, 2008

Top End Graphics

AMD's newest graphics card has definitely sent shock waves rolling across the graphics landscape.

It is clear from reviews such as Anandtech's The Radeon HD 4850 & 4870: AMD Wins at $199 and $299 that AMD is doing much better in terms of graphics. To be clear, nVidia's GT 280 is still at the top but HD 4870 has managed to surpass GT 260 to take second place. Now, with performance midway between the $400 GT 260 and $650 GT 280 the HD 4870 should be priced at $525. Instead the HD 4870 is a phenomenal bargain at just $300. It looks like nVidia's shiny new GT 200 series is now greatly overpriced. Based on the comparison GT 260 should probably be priced at about $280 and GT 280 at perhaps $350. Nor is nVidia's last minute entry of GeForce 9800 GTX+ much help. This product at $230 is not really better than the $200 HD 4850. Anandtech says:

The Radeon HD 4850 continues to be a better buy than NVIDIA's GeForce 9800 GTX, even if both are priced at $199. The overclocked, 55nm 9800 GTX+ manages to barely outperform the 4850 in a few titles, but loses by a larger margin in others, so for the most part it isn't competitive enough to justify the extra $30.

I have to say though that even when AMD does well there is still subtle bias at Anandtech. For example in the final comments they have to get in this sour note:

You may have noticed better CrossFire scaling in Bioshock and the Witcher since our Radeon HD 4850 preview just a few days ago. The reason for the improved scaling is that AMD provided us with a new driver drop yesterday (and quietly made public) that enables CrossFire profiles for both of these games. The correlation between the timing of our review and AMD addressing poor CF scaling in those two games is supicious. If AMD is truly going to go the multi-GPU route for its high end parts, it needs to enable more consistent support for CF across the board - regardless of whether or not we feature those games in our reviews.

To be perfectly honest I'm a bit floored by this comment. Last minute driver fixes for products that won't even be available for another month shouldn't be too unusual for people who claim to have experience with reviews. That is pretty much the nature of trying to get your reviews out quickly in competition with other websites. Newspapers do the same thing when they chase a breaking story and sometimes they are editing copy right up until it goes to press. This negative comment is even more odd when contrasted with the comment further back in the article:

It is worth noting that we are able to see these performance gains due to a late driver drop by AMD that enables CrossFire support in The Witcher. We do hope that AMD looks at enabling CrossFire in games other than those we test, but we do appreciate the quick turnaround in enabling support - at least once it was brought to their attention.

This seems to be Anandtech schizophrenia at its finest when appreciation for "the quick turnaround in enabling support" turns into suspicion at the "correlation between the timing of our review and AMD addressing poor CF scaling ". Guys, if you read your own text it says that you brought it to AMD's attention and then they sent you an update. We can also see that Anandtech was careful not to end with this comment:

We've said it over and over again: while CrossFire doesn't scale as consistently as SLI, when it does, it has the potential to outscale SLI, and The Witcher is the perfect example of that. While the GeForce GTX 280 sees performance go up 55% from one to two cards, the Radeon HD 4870 sees a full 100% increase in performance.

I assume nVidia must be sweating a bit over a comment like that. This suggests that some more driver work could put dual HD 4870 ahead of nVidia's dual GT 280 in a lot games. This seems especially true when we recall that unlike 3870 X2 the new 4000 series uses a special proprietary inter-processor link instead of using Crossfire. I think we can give credit for that to AMD whose engineers no doubt have lots of experience doing the same thing with CPU's. We've certainly come a long way since AMD's 2900XT last year which although worse in performance than nVidia's 8800 GT still drew a lot more watts. This should also be very good news for Intel. Intel motherboards have by far the weakest integrated graphics compared to nVidia and AMD and really need a boost with an added discrete graphic card. However, Intel motherboards can't do SLI so although Intel is loath to admit it the best graphics on Intel motherboards can only be had with AMD graphic cards using Crossfire. This means that AMD's huge leap in performance with HD 4870 has also just given a big boost to Intel motherboards for gaming.

I'm also a bit doubtful about nVidia's current strategy. nVidia must have been feeling pretty smug last year compared to 2900XT and even the recent shrink to 3870 can't have been too much of a concern. But HD 4870 is a big problem because it performs better than GT 260 but has a much smaller die. This means that nVidia has now just gained a huge liability for 200 and 8800 series inventory out in the field. nVidia will probably have to give rebates to recent wholesale buyers to offset what will almost certainly be substantial reductions in list prices. This also means that nVidia's current inventory has taken a hit as well. Although the 200 series is due for a shrink, the die is so much larger AMD's 4000 series that it is difficult to imagine that nVidia can turn a good profit even at 55nm's. With GT 280 designed for a price point above $600 nVidia is probably going to take a good revenue hit in the next quarter. I wouldn't be surprised to see nVidia change course and start thinking about reducing die size like AMD.

Finally, there should be no doubt at this point that the ATI purchase has turned ATI around giving them both more resources for product development as well as access to a much larger sales force. A rising tide for ATI products is probably some benefit to AMD but obviously AMD has put a great deal of money and resources into the merger. It still appears that AMD had no real choice in the matter but we'll probably have to wait to find out if this relationship has indeed become greater than the sum of its parts.

Some apparently feel that I pick on Anandtech too much so I'll link to some other reviews. Tech Report is considerably more upbeat in their conclusions:

The RV770 GPU looks to be an unequivocal success on almost every front. In its most affordable form, the Radeon HD 4850 delivers higher performance overall than the GeForce 9800 GTX and redefines GPU value at the ever-popular $199 price point. Meanwhile, the RV770's most potent form is even more impressive, in my view. Onboard the Radeon HD 4870, this GPU sets a new standard for architectural efficiency—in terms of performance per die area—due to two things: a broad-reaching rearchitecting and optimization the of R600 graphics core and the astounding amount of bandwidth GDDR5 memory can transfer over a 256-bit interface. Both of these things seem to work every bit as well as advertised. In practical terms, what all of this means is that the Radeon HD 4870, a $299 product, competes closely with the GeForce GTX 260, a $399 card based on a chip twice the size.

You may also get a bit better technical description of the architecture at Tech Reports and you get graphs of games at multiple resolutions.

AMD decided a while back, after the R600 debacle, to stop building high-end GPUs as a cost-cutting measure and instead address the high end with multi-GPU solutions.

This is mostly incorrect. The shaders were increased from 320 to 800 which is definitely a brute force approach to scaling. It appears that what AMD actually did was revamp the memory bus and then apply those power savings to additional shaders. In other words, it appears that AMD's limit is power draw (as born out in a number of reviews) rather an arbitrary stopping point. We also disagree in terms of die size as I believe that AMD has finally gotten dual GPU right by using a proprietary link.

The Legit Reviews article is not nearly as good. There benchmarking is only a fraction of what TR and Anandtech do nor do they have a GT 260 for comparison. However, their most bizarre statement concerns power consumption:

The GeForce GTX 280 has impressive power savings features as you can tell above. The HIS Radeon HD 4870 uses GDDR5 that is supposed to save energy, so they must have had to really increase the core voltage to reach 750MHz. Both the Radeon HD 4850 and HD 4870 use a little more power than we want to see.

From this description you would assume that HD 4870 draws more power than GT 280; this is not the case. In fact, GT 280 draws 20 amps more than HD 4870 under load. Those "impressive power savings" are only seen at idle. So, this characterization is certainly debatable since most computers have sleep or standby modes when they are idle.

The Hexus Review is also missing a GT 260 comparison. I find the author's style of English to be tougher to get information from. I'm not sure whether that is due to his UK heritage or just his heavy tongue in cheek and somewhat meandering writing style. It's probably a bit of both.

Then we have the HardOCP Review which thankfully does include a comparison with GT 260. Their conclusions are:

AMD’s new Radeon HD 4870 and Radeon HD 4850 offer a gigantic performance improvement over their last generation of GPUs and offer great values when compared to NVIDIA new GTX 280 and GTX 260. These performance improvements translate into real-world gaming benefits being able to play at higher resolutions with higher in-game settings utilizing AA and AF.

We were surprised to see the Radeon HD 4870 competing well against the GeForce GTX 260. It not only matched the gameplay experience, but in Assassin’s Creed it edged out with faster framerates too. Count on drivers to mature in both camps as well.

The Radeon HD 4850 is the new sub-$200 video card to beat as it provides the best gameplay experience for the money. It provides a better experience than a GeForce 8800 GT and a GeForce 9800 GTX and is on par with GTX 260 while being less expensive.

While NVIDIA still ultimately holds the single-GPU performance crown with the GeForce GTX 280 it also holds the “Hard to Justify the Price” and “More Money Less Value” crowns with the GTX 280 as well. AMD is now offering the best value in video cards with its 4800 series GPU. And we also see AMD’s drivers maturing more for the 4800 series than we do for the new 200 series from NVIDIA so it is our guess that AMD’s performance will get even better with its new GPUs comparatively.


I pretty much agree with this. We further have yet another reference to maturing drivers. Peddie of Jon Peddie Research even suggested, "Nvidia and ATI keep improving their drivers so they'll seesaw back and forth with their scores, almost from week to week". So, again the notion of some kind of conspiracy as suggested by Anandtech seems to have no connection with reality. Peddie also seems to agree with the idea that nVidia may have to shift to smaller dies.

25 comments:

Polonium210 said...

It gets much better-the HD 4800 products support ray-tracing!
LINK

Scientia from AMDZone said...

orly

"The 4850 is on newegg."

Right, I was referring to the 4870.

"I think you missed the point that they had a quick turn around on a big game thats benchmarked everywhere and a no name game thats pretty much only benchmarked on anandtech."

I don't know what you are talking about. There were not just two games. In reality Anandtech reviewed multi-GPU peformance on: Crysis, Call of Duty 4, Enemy Territory: Quake Wars, Assassin's Creed, Oblivion, The Witcher, and Bioshock. That would be 7 games reviewed.

"Meanwhile theres games where the 3870x2 has scaling issues and are not fixed"

3870 X2 communicates via a Crossfire link. AMD realized that this was a problem so they improved it. The 4000 series now uses a faster, proprietary link. It is impossible for 3870 to get the same speed in Crossfire regardless of driver updates.

"How many other games are out there have scaling issues as well? How many other bugs are there with whatever that need addressing that could also be hotfixed?"

There may be a number. And, I'm sure the driver updates will continue as suggested by pretty much everyone except Anandtech.

"Yeah that seems rather convenient to me."

I guess I'm puzzled where your tinfoil hat has been with the driver updates for 2900 and nVidia's driver updates. There has been no real change in updates or hotfixes.

"For someone who hates anandtech so much, I find it funny that you only link to their review as well."

Well, people have complained at times that my articles are too long but since you feel I am picking on Anandtech I added another four review links.

"Oh wait, I see why you did now."

I doubt you do but keep trying.

Unknown said...

I'm very pleased that finaly AMD is making waves in..i would say consumer market and giving you a mouthfull, so you can keep all the interesting things coming our way.I am reading your blog frequently, and would like to say...thank you.

Scientia from AMDZone said...

nebojsa

"I'm very pleased that finaly AMD is making waves in..i would say consumer market"

Well, AMD hasn't really made any stir since early 2006. I guess after two years it is about time.

The puzzling thing though about some Intel fans is that they seem to really rejoice in the current performance disparity. However, even if you are a die-hard nVidia fan the HD 4870 should be welcome news since it will almost certainly push down the price of nVidia cards. The same thing should apply to Intel processors if AMD could get its performance up a bit more.

For example, you can see that AMD's dual and single lineup has been shifted downwards which has eliminated the FX segment at the top as well as the Sempron range at the bottom which is now filled with former midrange Athlon 64's. Intel in contrast still has both its Celerons and reduced cache Allendales to cover the low end.

There has only been a slight price drop for Intel quads presumably due to the recent small overlap of top clocked AMD quads and low clocked Intel quads. I assume that additional overlap would benefit Intel fans with lower prices.

gdp77 said...

scientia what are the chances that AMD's finances will be improved from the 48xx sales?

muziqaz said...

theinq has an article about nvidia's tesla :D
link


I don't know if the guy who wrote the article is same nebojsia who posted here.
I think it is quite good article about some performance numbers:)

Unknown said...

Another point is that without second opponent the market is coming to a halt.2-4 months ago anandtech, tomshardware and almost anybody, was writing about PSUs, diferences between p35 and p45, speedup from 8800gt to 9800gtx without new significant innovations. It's like watching a football with only one team on the field. No excitement, no discussions. We need two players minimum to make us at least buy some pop-corns and watch, maybe learn something on the way.

Scientia from AMDZone said...

george

"scientia what are the chances that AMD's finances will be improved from the 48xx sales?"

I imagine they will improve a bit. For example, the graphics section will probably turn a profit instead of losing money. This will help but I don't think it would be enough on its own to reverse AMD's $400 Million a quarter losses even with the layoffs.

Scientia from AMDZone said...

orly

I've been trying to make some sense of your comments but I'm not making much headway. You seem to have some crazy notion that AMD only issues fixes when it improves review benchmarks. There have been a number of updates to the Catalyst drivers as well as hotfixes. Presumably if these always coincided with reviews at Anandtech they would have noticed it by now.

Some of your other comments are simply delusional.

"they got hotfixes for 2 games that are regularly benchmarked while other games continue to have issues that are not used in benchmarks."

In reality, Anandtech says:

"due to a late driver drop by AMD that enables CrossFire support in The Witcher."

The Witcher does not appear in the reviews at Hexus, HardOCP, LegitReviews, or TechReport. So, much for your idea that this is regularly benchmarked. Another odd comment by you is:

"the GeForce 9800 GTX+. If Nvidia can deliver those as expected by mid-July and cut another 30 bucks off of the projects list price, they'll have a very effective counter to the Radeon HD 4850"

I agree with this quote from Tech Report. At a lower price 9800GTX+ is a good deal and being made on 55nm should be more cost effective for nVidia. As such it would be great competition for 4850. This is just common sense. Why would you erroneously think that you had somehow blinded sided me with this?

"How is adding more hardware to use the bandwidth you have brute force?"

If adding 150% more shaders isn't brute force then I can't imagine what would be. This contrasts completely with nVidia's approach where the shaders have actually decreased from 1836 in 9800GTX+ to only 1296 in GT 280.

Frankly I'm not sure why you waste your time trying to troll here. You can make up anything you like (as long as it is anti-AMD) and post on roborat's blog without restraint.

The main point which seems to have completely eluded you is that AMD is today considerably more competitive in terms of discrete graphics. Note that this is in complete contrast to the Intel fans who predicted that AMD would make no gains with their next GPU because supposedly nVidia's next GPU would maintain the same lead.

Scientia from AMDZone said...

muziqaz

"I don't know if the guy who wrote the article is same nebojsia who posted here."

I would be surprised if anyone from the INQ had ever heard of my blog. However, there are some errors and omissions in the article.

It is never mentioned that AMD GPU's have some global memory whereas nVidia GPU's do not. This tends to make AMD's GPU's more suitable for general purpose computation.

He doesn't mention that it is possible on AMD GPU's to perform operations during a memory transfer to hide some of the latency. He doesn't mention that AMD's programming model can span across both processors and GPU's.

He doesn't seem to mention that the FireStream 9250 can do 200 double precision GFlops at $1000. This would give you 400 GFlops for just $2000 so the box would cost half of what nVidia's or a Skulltrail would.

He is correct though about extracting performance. Both Intel's and Portland's compilers do have some automatic parallelization. That is the kind of thing we really need for GPGPU. This could improve however with Khronos or OpenCL.

He also doesn't seem to mention that local memory is a big problem for IBM's Cell which is current top architecture for HPC.

muziqaz said...

reacently as you know(maybe). Stanford released folding client for nVidia GPUs. nVidia client was based on hard work of ATI programmers and Stanford programmers(Mike Houston, who now works for AMD, though I do not take any credits away from others involved in that project), thus the client being so mature(in performance). The client is written using CUDA. And nvidia cards took folding scene by storm, because 8600GT generates same amount of points as 3850/3870.
As we all know 3850 eats any 8600 based card in games and in workstation market. And in workstation market 3850 eats any g80/g92 card for breakfast.
So I was interested to know if nvidia finally found some magic in their driver team, that those cards perform so much better in Stanford project. I asked around about that, but I got only one proper answer, that folding WUs like nvidia architecture and programming architecture, thus generating so many points. I asked again about 8600GT being equal to 3850, but did not get any answer at all. Well as I am involved passionately in this project I do not want to say that the performance numbers and points on G80/92 are greatly exaggerated, but no answer does raise a lot of questions for me. Is nvidia new driver that good? Or is it just nvidia doing some proper PR.
I do not question gtx 260/280 cards, they are much faster in folding than 3850/3870 cards(almost twice). But ATI cards in that project are bottlenecked by CPU, while nvidia's are not. So there is another question: is AMD programmers cannot figure out the way to harness full potential of their cards? Let's take 4850, it almost idles while folding :D

Ho Ho said...

polonium210
"It gets much better-the HD 4800 products support ray-tracing!"

From the little information we have available now it seems like this isn't traditional ray tracing but some kind of clever hack that makes the image pretty but has tons of constraints.


scientia
"It is never mentioned that AMD GPU's have some global memory whereas nVidia GPU's do not. "

But NV GPUs do have global shared memory caches. Or did you think of something in the multi-GPU boards?


As for the numbers, in theory Radeons have had tons of more FP throughput compared to NV gpus over the last few generations but in real-world that performance is very difficult to use efficiently compared to NV GPUs.

Scientia from AMDZone said...

ho ho

"But NV GPUs do have global shared memory caches. Or did you think of something in the multi-GPU boards?"

This is what Anandtech said:

"AMD offers an advantage in the SPMD paradigm in that it maintains a global store (present since RV670) where all threads can share result data globally if they need to (this is something that NVIDIA does not support). This feature allows more flexibility in algorithm implementation and can offer performance benefits in some applications."

"As for the numbers, in theory Radeons have had tons of more FP throughput compared to NV gpus over the last few generations but in real-world that performance is very difficult to use efficiently compared to NV GPUs."

That kind of statement is nonsense. In reality, neither the CUDA nor Brook+ high level environments have been around that long and neither include transparent parallelization as we are beginning to see in CPU compilers. Maybe we'll get somewhere with OpenCL.

Secondly, both architectures are designed for graphics, not general purpose processing. What we should see over the next couple of years is not only maturing of the programming environments but also some tweaking of the hardware to make them more suitable for general purpose programming.

Scientia from AMDZone said...

muziqaz

I really don't know much about the F@H status. The last I had heard was that there was some contraversy over how points were calculated with the ATI client. I also have not seen any specific numbers for nVidia but my assumption was that the claim of the fastest hardware was for GT 280.

I don't know if there is even a client for AMD's 4000 series. Specifically, the old client was written in CAL whereas the nVidia client was written in CUDA. To have similar maintenance I assume the AMD client would need to be written in Brook+ but I don't know they can actually get enough performance that way. Again, these applications are very new.

Scientia from AMDZone said...

ho ho

TechReport says the same thing about Radeon's global memory:

"Another such accommodation is the addition of 16KB of local shared memory in each SIMD core, useful for sharing data between threads in GPU-compute applications. This is obviously rather similar to the 16KB of shared memory Nvidia has built into each of the SM structures in its recent GPUs . . . Uniquely, the RV770 also has a small, 16K global data share for the passing of data between SIMDs."

muziqaz said...

Scientia


I really don't know much about the F@H status. The last I had heard was that there was some controversy over how points were calculated with the ATI client.


Not as much controversy as with nvidia client :) x1950xtx used to fold in certain scenarios 30-40x as fast as fastest dual core processor, but it lacked the versatility of the cpu, which can(slowly) do all kinds of simulation scenaious, then comes PS3 with performance better than CPU, and much better variation in simulation scenarios than GPU, and after that is GPU, great performance, only useful in few folding simulation scenarios. So normalizing points was done I think in the right way. As far as I remember x1950xtx was doing 800-900ppd(my ps3 does ~900ppd). Then came along 38xx series ~2k ppd, which is again justifiable. BUT, I will never understand where does 8600GT gets those 2k ppd from. So that is controversy.

I also have not seen any specific numbers for nVidia but my assumption was that the claim of the fastest hardware was for GT 280.


Yes, that is correct, but nVidia kindly omitted this in their PR conferences. I mean there is nothing wrong with that, but most of folding enthusiasts who were nVidia fans, jumped the guns saying that every nvidia card folds much faster than 38xx series. This still needs to be confirmed by Stanford.

I don't know if there is even a client for AMD's 4000 series.

When AMD recruited Mike(Houston), I knew from the start, that AMD will try to milk the folding cow even more. Thus they released GPUv2 client which supports 2xxx series, 3xxx series and 4xxx series GPUs. I don't know what is happening with R800, so who knows :)

Specifically, the old client was written in CAL whereas the nVidia client was written in CUDA. To have similar maintenance I assume the AMD client would need to be written in Brook+ but I don't know they can actually get enough performance that way.


Actually GPU1(older one) was written with brook+ and was using directx for folding. Unfortunately, directx is not the most precise API for scientific calculations, thus GPU1 is retired now. Now the new GPUv2 client is written with CAL and does not use directx.
And as Mike is working for AMD now, he gets an access to internal info about every tidbit of ATI GPUs. That is why I was so in favour of Mike's move to AMD.

Again, these applications are very new.


But I dare to say, that Fah GPUv2 client(ATI one) is one of the apps in industry which is capable to use almost all GPU resources. Sure they still have to work a lot to get to the efficiency of GPU1 client, but they will get there. Sure 1st they need to figure out, how to remove CPU bottleneck :)

And by the way, Vijay Pande himself said, that nVidia drivers are written for performance, while AMD drivers are written having reliability in mind.

Scientia from AMDZone said...

I need to make a correction. Earlier I said:

If adding 150% more shaders isn't brute force then I can't imagine what would be. This contrasts completely with nVidia's approach where the shaders have actually decreased from 1836 in 9800GTX+ to only 1296 in GT 280.

I wrote down 1836 and 1296 for nVidia which are actually the shader clock frequencies instead of 128 and 240 which are the number of shaders. What I should have said was that AMD increased the number of shaders by 150% while nVidia's increase was only about half of that at 87.5%. Better.

Scientia from AMDZone said...

muziqaz

Thanks, you are obviously a lot better informed about F@H than I am.

"And by the way, Vijay Pande himself said, that nVidia drivers are written for performance, while AMD drivers are written having reliability in mind."

I hadn't heard that. That is interesting.

muziqaz said...

well, fah not only gives us some hope in fighting all kinds of deseases, but also their software shows what certain hardware can do. When hardware developer says this and that on paper about architecture, its potential, its theoretical power, FAH takes a good use of 75% of that theoretical power. If it wasn't for other bottlenecks in systems, I would say FAH would use 95% of theoretical power.
So it is good measure to see what is what in architecture.
And that excites me a lot :)

muziqaz said...

Scientia, and you say theinq does not read your blog :D
more architectural talk

Ho Ho said...

muziqaz
"And in workstation market 3850 eats any g80/g92 card for breakfast."

I'd like to see a link to that.


"But ATI cards in that project are bottlenecked by CPU, while nvidia's are not. So there is another question: is AMD programmers cannot figure out the way to harness full potential of their cards?"

My guess is NV can offload more to GPU than ATI.


scientia
"That kind of statement is nonsense"

If you'd read B3D G80 overviews you'd seen that they said it was super easy to make shaders use most of its potential, only problem was the "missing MUL".


As for the "ray tracing" demo, here is what was found in TG comments:
http://ompf.org/forum/viewtopic.php?p=8152#p8152

"yes, the camera was completely movable, just like elements in the scene.

According to Jules, scenes that you can see in the gallery have approximately three million different viewports, thus you have around three million combinations where you can point the camera from.

Camera movement is of course, completely seamless and you don't see the transition between the viewports... perfectly normal, since you have around 3M viewports in a scene that is only 0.92 MPixels."



Emphasis mine. So it is basically a tiny brebaked scene that will not work in real-world game scales.

Scientia from AMDZone said...

orly

"Wow, they were right. You do edit the **** out of peoples post"

Just yours actually.

"They're hotfixing games used in benchmarks on review sites over other games. Anandtech said it and so did other sites."

I haven't seen this mentioned at other sites. But, again, it is not unusual to get updates during a review. This by itself would not be out of the ordinary.

"ATI typically don't do this and like I've said (and you've edited out) they want to avoid the R600 problem."

ATI has released hotfixes before that improved games so that is not a difference. You may be right though that AMD realized that when they release driver updates the cards don't get reviewed again so they probably are trying to make the initial review as good as possible.

" This is something more typical of Nvidia with their beta drivers (which of course never make it to WHQL status) they always seem to release whenever ATI are releasing something."

That I'm not sure about. As far as I know, the hotfixes are normally included in the next general update of the Catalyst drivers.

"I said the witcher isnt reviewed anywhere but anandtech. Its a no name game, its sold less then 1 million copies worldwide but it gets hotfixed because it get in anandtech benchmarks (god knows why) There are other games still needing fixes, wheres the hotfixes for those?"

I'm not making much sense of what you are saying. If other games in the review needed to be fixed then AMD would have been foolish to concentrate on a less popular game. But, if the fix improved more than just The Witcher then that would remove your main point. Or you might be trying to say that other games on older cards need to be fixed. But as I've already mentioned, the X2 performance of 3800 series cards is going to be much worse than it is for 4800 series because they are connected with a completely different interface. Perhaps you could make your point more clearly by giving a link/example of a game that needs to be fixed.

"Wow, totally take out of context, how is that more upbeat?"

I didn't say it here but have mentioned on AMDZone that the TechReport author likes big GPU's and doesn't like AMD's strategy of smaller GPU's. So, no he was not completely positive. I believe I say more upbeat.

"That you're new card can be countered with an old card that had a die shrink."

Specifically, you have a point that AMD's shrunk 3870 is no match for nVidia's shrunk 9800GTX+. True indeed. However, I would say that nVidia was planning to keep its G92 series around awhile longer while AMD is probably going to shift to RV770 much more quickly.

"It'll still need the price cut because the 4850 is going for under $200, have better features and can used in crossfire to beat out the GTX200 series for cheaper."

Right, but if nVidia cuts the price to under $200 they'll take a big hit on margin. I've been talking about this at AMDZone but didn't write another article about this on my blog because the ideas mostly come from other people like Demerjian. For example, I really have no idea what nVidia's yields on G200 are.

"And I have no doubt that the 9800gtx+ will be a paper launch."

I actually doubt this because nVidia had planned to update the entire G92 series to 55nm. And, since G200 is still a very large die I'm not sure what other choice they have.

"And what would increasing clock speeds be? The eloquent approach?"

Well, AMD's doubling the performance of their texel units was pretty eloquent and so was their bus overhaul by adding a hub. This makes me wonder if Intel will follow suit and drop the racetrack bus on Larrabee.

"I did point out that needed to increase it because they were pad limited as well."

You'll have to explain that one more slowly.

It looks to me like the power savings from 2900XT to 3870 come both from the shrink to 55nm and from getting rid of the 512 bit external memory bus. There seem to be some additional power savings on 4870 by getting rid of the internal 512 bit racetrack bus. AMD also added some of its CPU power features but these only help when the GPU is not fully utilized. Overall, this makes 4870 much better in terms of power than 2900XT. I was saying that these power savings were then used to add more shaders to increase performance. Imagine what would have happened if AMD had increased the shaders on 2900XT by 150%.

"In case you weren't aware, ATI have always had more shader hardware and they've been using the 4:1 ALU:Tex ratio for quite a while now. It adds more value to cards in the long term, ATI have always been about that. Why would they suddenly turn around and change that?"

I corrected my shader numbers above.

"I'm sorry to break it to you but that point hasn't eluded me, they own the market at $200 and $300 and it shows with the 4850 out in good numbers and selling like crazy. It'll be a while before Nvidia can really do anything besides cut prices and eat lower margins. "

Yes, I think their margins will be an issue with that large G200 die, 400mm2 even after a shrink.

"The competition is much needed and I know for sure I'm picking up a 4870 when its available here."

muziqaz said...

Ho Ho, my statement was based on some old comparison of fire GL (2900 xt class) with quadroFX (8800gtx class).
link.

I do believe that thing have changed since then, but AMD still comes at the top with it's pricing.

And about the fah, yes, nvidia's cuda managed to compile the code which behaves very well in protein simulations :)
but Vijay said, that it might change, when they release new/bigger WUs.

Ho Ho said...

muziqaz
"but Vijay said, that it might change, when they release new/bigger WUs."

I'm sure things will change, just not sure in what direction things change on either architectures. My gut feeling says that as NV has more flexible HW it should scale relatively better with more complex code.

Joshua said...

I think right now AMD just made a win for Intel big time. Interesting fact: Recently, AnandTech reviewed the new SB750 and they showed that it was now possible to reach 3.3 GHz with the Phenom. I think AMD needs to straighten everything out with Phenom. Although it may not be the best CPU, it definitely is a sweet deal. AMD really needs to pull through and start re-generating revenue, if AMD goes out the window Intel is going to just jack up the prices on all the existing CPUs and stop developing anything.