AMD: R7xx Speculation

[Bouncing Zabaglione Bros.]

AMD Will Show R700 on Cebit08 (2008 march 4-9)

Yeah, but when will they ship it? Three months later? Six months later? Anyone can show a limited early engineering sample, but when will they get it running at high speed/performance, working properly without bugs or issues, and yielding well? AMD have shown over the last year that they can only execute two out of three of those at any one time.

 

[Speccy]

Good news if true. Problem is it's almost impossible to believe anything good coming out of AMD these days.

RV670 / 3800 wasn't good news?

 

[Bouncing Zabaglione Bros.]

RV670 / 3800 wasn't good news?

In general, yes. It has good price-performance, but there seems to be supply issues, driver issues (especially in it's performance configuration as Crossfire), it doesn't compete at the high end, has AA performance issues, etc. It's a good product, but it's not a great product. It looks better than it is because of the dropped ball that R600 was in so many ways. As part of the Spider platform, it's the only bit that actually works well, but that's also more a reflection of the relative problems/lateness with regards to Phenom and RD790.

In short, R670 looks good because everything else around it from AMD looks so bad, especially in the face of the successes from Intel and Nvidia at the same time.

The best thing about RV670 is that it replaced and fixed a lot of the things that were wrong with R600, and a lot of that is down to how fortunate ATI was with a good first spin.

 

[Acert93]

Good news if true. Problem is it's almost impossible to believe anything good coming out of AMD these days.

Good thing it is coming out of the ATI wing then Outside of R600 and the troubles ATI had with the R520 bug, ATI has competed well in the market. Since R300 ATI has had the R420 which was a solid performer with good IQ, the biggest problems being XT PE availability and how the midrange (X1700) didn't fair as well against the competition (NV 6600GT). R520 had the afformentioned bug, but R580 was a really nice part for ATI in terms of performance and features. The midrange, again, is an area that seems to have been slightly misgauged. And in this same timeframe there was Xenos which performs well, is making ATI money (at the cost of more PC resources, ala NV2A???) which punctuates the question: Were the whispers in ATI that R580 should have been unified the correct move? NV really did steal ATI's thunder there.

Anyhow, ATI has been competitive in the GPU space for quite a while now. R600 has been a mess for sure and until recently midrange GPUs didn't get the proper TLC they deserved (in my opinion).

Outside of potential R&D cutbacks and key defections I think ATI has proven they can develop strong technologies that compete in the market. Obviously they have their work cut out for them (multi-GPU systems still seem to have issues on both sides of the fence, the Havok buyout really puts a damper on ATI, devrel still seems to be a major weak spot at ATI, will they be able to push out competitive parts in the midrange and retake the performance crown... and start generating better margins, etc) but I wouldn't count ATI out completely at this point.

 

[Arnold Beckenbauer]

Good thing it is coming out of the ATI wing then Outside of R600 and the troubles ATI had with the R520 bug, ATI has competed well in the market. Since R300 ATI has had the R420 which was a solid performer with good IQ, the biggest problems being XT PE availability and how the midrange (X1700) didn't fair as well against the competition (NV 6600GT). R520 had the afformentioned bug, but R580 was a really nice part for ATI in terms of performance and features. The midrange, again, is an area that seems to have been slightly misgauged. And in this same timeframe there was Xenos which performs well, is making ATI money (at the cost of more PC resources, ala NV2A???) which punctuates the question: Were the whispers in ATI that R580 should have been unified the correct move? NV really did steal ATI's thunder there.

Anyhow, ATI has been competitive in the GPU space for quite a while now. R600 has been a mess for sure and until recently midrange GPUs didn't get the proper TLC they deserved (in my opinion).

Outside of potential R&D cutbacks and key defections I think ATI has proven they can develop strong technologies that compete in the market. Obviously they have their work cut out for them (multi-GPU systems still seem to have issues on both sides of the fence, the Havok buyout really puts a damper on ATI, devrel still seems to be a major weak spot at ATI, will they be able to push out competitive parts in the midrange and retake the performance crown... and start generating better margins, etc) but I wouldn't count ATI out completely at this point.

X700.
RV410's problems were: heat (no X700XT) and it's buggy memory controller.

 

[Acert93]

Ah yes, X700 not X1700. Brain fart

 

[ShaidarHaran]

RV670 / 3800 wasn't good news?

Good value proposition at MSRP, but the lack of availability and the price gouging by e-tailers means it's not really an option.

In general, yes. It has good price-performance, but there seems to be supply issues, driver issues (especially in it's performance configuration as Crossfire), it doesn't compete at the high end, has AA performance issues, etc. It's a good product, but it's not a great product. It looks better than it is because of the dropped ball that R600 was in so many ways. As part of the Spider platform, it's the only bit that actually works well, but that's also more a reflection of the relative problems/lateness with regards to Phenom and RD790.

In short, R670 looks good because everything else around it from AMD looks so bad, especially in the face of the successes from Intel and Nvidia at the same time.

The best thing about RV670 is that it replaced and fixed a lot of the things that were wrong with R600, and a lot of that is down to how fortunate ATI was with a good first spin.

Exactly.

Good thing it is coming out of the ATI wing then

Even then I still can't get excited. I really looked forward to R700, but everything in the R6xx family has been a huge disappointment, speaking purely from the perspective of an enthusiast that demands the best performance and IQ. There are some good value propositions on the ATi side, and features that are unmatched by the competition, but the performance just isn't there.

Outside of R600 and the troubles ATI had with the R520 bug, ATI has competed well in the market.

They've remained competitive on the strength of OEM design wins. ATi cards aren't selling to enthusiasts through retail channels, and that's going to come back and bite them in the ass eventually (I think it has already, actually).

 

[Shtal]

but everything in the R6xx family has been a huge disappointment.

For example: If HD3870 RV670XT came out last year - exactly "Nov 2006" instead this year; then it might not be complete disappointment based on time frame when it was delivered.

 

[ShaidarHaran]

For example: If HD3870 RV670XT came out last year - exactly "Nov 2006" instead this year; then it might not be complete disappointment based on time frame when it was delivered.

Again, from a pure enthusiast performance perspective, it was disappointing whenever it came out, unless it was released well in advance of G80. Since none of us lives in an alternate timeline where that occurred, RV670 is not an exciting product for enthusiasts.

 

[Shtal]

Again, from a pure enthusiast performance perspective, it was disappointing whenever it came out, unless it was released well in advance of G80. Since none of us lives in an alternate timeline where that occurred, RV670 is not an exciting product for enthusiasts.

Sure, based on comparing to G80 ATI did failed to deliver competitive product.

 

[nexus_alpha]

Again, from a pure enthusiast performance perspective, it was disappointing whenever it came out, unless it was released well in advance of G80. Since none of us lives in an alternate timeline where that occurred, RV670 is not an exciting product for enthusiasts.

The rv670 is not targetted to the enthusiast crowd

 

[BRiT]

The rv670 is not targetted to the enthusiast crowd

AMD/ATI has nothing targeted to the enthusiast crowd which is very disappointing.

 

[ShaidarHaran]

The rv670 is not targetted to the enthusiast crowd

I'm well aware of this.

AMD/ATI has nothing targeted to the enthusiast crowd which is very disappointing.

Exactly my point. The very fact that AMD has no presence in the enthusiast market is what my gripe is about.

 

[AlexV]

I'm well aware of this.



Exactly my point. The very fact that AMD has no presence in the enthusiast market is what my gripe is about.

Waiting for WaltC and the boys to come and educate you on the fact that the enthusiast market is irrelevant, it's only the high-volume parts that matter. That would be fun.

I'm hoping that they'll actually have some kind of stab at the enthusiast market with the 3870X2. Because the fact that the GTX and the Ultra have been fairly constant in terms of pricing ever since their release(not truly significant downwards shifts, I think the GTX cheapened somewhat, but that's all) is an indication of stagnation and what the lack of competition brings.

 

[ShaidarHaran]

LOL. Hey, up until a few short weeks ago I *was* "one of the boys"

I hold out no hope for the 3870 x2. We know what the 3870 can do, we know what it can do in CF. CF "on a stick" doesn't bring magic pixie dust with it that increases performance super-linearly so unless ATi can pull an uber CF driver out of their collective arse (ha, not bloody likely) I see 3870 x2 as an altogether unexciting product.

 

[Bouncing Zabaglione Bros.]

I can't see Crossfire or X2 being a real alternative for the high end (even though ATI likes to paint it that way) while we have to wait months after a game's release before we get drivers with the correct profiles to make it work properly.

It can't be good to pay out for Crossfire or X2 and find that your second chip/card spends most of it's time idle - until several months after you've finished that hot new game.

 

[Valzic]

After Intel's fiasco with Netburst and the Pentium's not being able to massively upscale as they used to they released the dual core architecture.

They basically threw away the performance crown.

/sarcasm WHO would want their new multi-cpu architecture each individual component slower than their old abandoned single processor CPUs?

Things have changed.. Intel and AMD (at least trying) are pushing multi-cpu architectures.

It may be that multi-GPU architecture may mimic the same path. It may take some time but it may the way it is going whether we want it to go that way or not.

 

[3dilettante]

I'm not clear on which products you are talking about.

No successor to the Pentium 4 backslid on single-threaded performance.

The dual-core Conroe chips in the vast majority of instances are more powerful on a per-core basis than the P4, not that it was particularly hard after the thermal problems with Prescott.

 

[AlexV]

After Intel's fiasco with Netburst and the Pentium's not being able to massively upscale as they used to they released the dual core architecture.

They basically threw away the performance crown.

/sarcasm WHO would want their new multi-cpu architecture each individual component slower than their old abandoned single processor CPUs?

Things have changed.. Intel and AMD (at least trying) are pushing multi-cpu architectures.

It may be that multi-GPU architecture may mimic the same path. It may take some time but it may the way it is going whether we want it to go that way or not.

You lack understanding of the issue, the comparison is invalid. The move for multi-cored CPUs was a logical one as straight clock-rate scaling became impossible due to process limitations. CPUs aren't(weren't) inherently multi-cored. GPUs are. The move towards multi-GPUs per-PCB/package is due to process limitations(again) and possible reduction in R&D costs and increased ease of scaling(it should be cheaper to design a 300 mln trannie part than a 1.2 bln one, in theory, and it will yield better, and you'd have, again in theory, the possibility to scale the architecture to 4 different market segments(from 1-GPU cheap-arsed to 4-GPU enthusiast).

About the 3870X2, I think i've stated before that IMHO it's more of a signal along the lines of:Hello, we're still there, we still care about top-notch performance!, done within the constraints of what can be done now, with the R6xx architecture as a baseline. The R700 should be juicier, but considering AMDs and ATis recent track records, it can go either way easily enough.

 

[Valzic]

My thoughts on this is that both AMD and Nvidia have pushed the enthusiat products to what was a mainstream price with thier 3800s and 8800GT. That has to reduce the number of people that are willing to pay for the traditional upper level products.

Also that probably moved people that did not bother to be in the game to get the products because of their new price point, They are willing to spend 170-240 on product that basically plays the majority of the games well.

I figure that both nVidia and AMD will provide a bump this next generation, but IMHO I think they will concentrate on SLi/CF drivers (make then viable) and that may harbor a change in direction.

That is the creation of a product that basically fulfills a large portion of mainstream requirements without huge cost and use of that product in a multi- product scenario, either multiple GPU's on 1 PCB or traditional SLi/CF. It really depends on them getting the drivers working well.

I see IGP for the masses (looks like they will be not bad with the next ones being released), a mainstream part for people that want more and then multple GPU's for the enthusiasts. That is just my opinion.