AMD: R9xx Speculation

ATI Catalyst™ PC VENDOR ID (1002) LIST

http://developer.amd.com/drivers/pc_vendor_id/Pages/default.aspx

 

Wow! So Nvidia might actually release another card before AMD release their flagship and they will have stock in stores. (unlike AMD most probably?)

Damn, Nvidia are on the ball again(or AMD just dropped it again).

 

So we are really going to see HD5750 and HD5770 rebranded?

 

hmh, it's strange, it says "ATI Radeon" for those yet "AMD Radeon" for the NIs

 

Evergreen

 

I cannot wait for the next-gen consoles. The current high-end PC GPUs from Nvidia and AMD are already, at the very least, an order of magnitude (10x) more powerful than Xenos in Xbox360 and RSX in PS3. I'm not even talking multi-GPU cards from AMD or AMD's CrossFire (or whatever it's called now) or Nvidia's SLI.

Also, current AMD and Nvidia GPUs are like 100-200 times more powerful than Wii's Hollywood GPU, so I also cannot wait for Wii2 with modern AMD tech.

 

That shouldn't have effect on the actual retail branding

 

The majority of them are not for retail, some don't even exist if you compare Subsys ID's from previous drivers. These are OEM chips to show that AMD has a full suite of products for that market.

Edit: http://www.tomshardware.com/reviews/amd-fusion-brazos-zacate,2786-4.html
Some are also ID's for Zacate and Ontario APUs

 

Why double entries? Anyway, most if not all of this is Fusion so of course it's Evergreen.

 

Now this looks more legit......

 

This was in July... (link).

but...

 

I agree. Very tight supply of Barts pretty much explains why Cayman was delayed too.

 

And yet some how one would expect that they would've caught this problem a very long time ago when they were originally testing the cards which would've taken months and months.

I can understand the shortage of parts but that?

 

I don't think you're quite understanding it

They're saying the shortage is the same as the one for Barts, hence both are being hit by availability issues - not to mention that having Barts in production means that any shortage will hurt Cayman more

 

Ah, their 'Engrish' must be bad then.

 

http://www.nordichardware.com/news/...-samples-of-radeon-hd-6900-getting-close.html

Umm, who's generation? Nvidia's? The 580 maybe? (but I doubt that's what they meant)

But wouldn't AMD's current generation graphics CARD be the 5970?

Can you imagine 30% to 50% FASTER than the 5970? Yeah, I know they probably mean the 5870 but why should we assume that either?

Good times ahead.

 

Current gen is most likely 6870. Which would put it very close to GTX580.

 

I think they ment the 6870, so its some what old news. Oh, and if its only in tessellation we will see up to 50% improvments thats not much.

 

Yeah, I know, but I was hoping they meant cards that are "currently" on the market.

So a 6870 +50% would put it just below the 580 I think.

Try this site and see the percent difference of many cards. It may not be perfect but it's good for reference.

It's not in English, but you don't need it for comparing the numbers.

Here is the comparison chart. http://ht4u.net/reviews/2010/nvidia_geforce_gtx_580_test/index30.php
Here is the beginning page of the whole review.http://ht4u.net/reviews/2010/nvidia_geforce_gtx_580_test/

I just used 4x and 8x, 16AF, and 1920x1200 to see the difference of a 6870 and the 580.

It averages to about where the..

580 is 39.8% faster than the 6870 at 1920x1200, 4x, 16AF
580 is 72.2% faster than the 6870 at 1920x1200, 8x, 16AF

6870 CF is 36.8% faster than the 580 at 1920x1200, 4x, 16AF
6870 CF is 17.0% faster than the 580 at 1920x1200, 8x, 16AF

So if these are even close then the 6970's 30% to 50% gain over the 6870 will still lose to the 580.

It doesn't give 2560x1600, oh well.

 

I still believe that the 6970 will come very very close to the 580.

Two things.

No1 Speculation.

The Barts gpu is 250mm2. If you add 50% die size to that you end up at 375mm2. Still quite large for AMD's taste, but still that 125mm2 die increase would also include 50% more ROPS and 50% wider bus, ie 48 ROPS and 384bit bus, something close to it, since the don't scale linearly afaik. From what we know so far, the Cayman will only have 32 ROPS and 256 bits bus with very fast GDDR5.

So, it could be very well be the case, that the Cayman will indeed be 50% faster in Texturing and Shading power, but its fillrate will not be much faster, neither the memory throuput. The existing 32ROPS and memory bandwidth, could be very well enough, even for a Cayman level chip though.

Now according to Computerbase's GTX 580 review, the GTX 580 came 45% faster on a widely used 1920X1200+4XAA resolution.

I do believe that expecting a 40% increase on average, is really nothing for the Cayman chip. That 40% would bring the Cayman breathing down GTX 580's neck.

No2. AMD's own slide



It is very clear, that the Cayman is positioned quite higher in relation to the GTX 480, compared to what the barts and Juniper chips are compared to their respective counterparts.

Now, again from the above link of computerbase's GTX 580 review (1920X1200+4XAA), we know that the 5770 is 17% faster than the GTS 450 and the 6870 is 22% faster than the 460 1GB. If the positioning of the Cayman on that slide, is anywhere near the truth, we could be looking at a performance lead of more than 20% over the 480 and we already know that the 580 is less than that.

Hopefully for us end users, this could come in at 100$ less than the 580 price, and both we and Nvidia, would be in for a treat!

Whatever the case, the 6970 will be very close to the 580. A bit slower or a bit faster, does not really matter for me. What matters is the overall package, including the performance/price and consumption/performance ratios. I do believe that AMD will win hands down. I may still go with Nvidia if the 570 is any more reasonably priced than the 580, anyway.