AMD Cayman info (or rumor)

[LordEC911]

No way AMD would even release it like this, so somethings up.

Magic drivers 10.11 borks even 6870/6850 results, he needs better drivers.
Cayman really likes the higher resolutions.

As for shader performance, as zerazax has been stating, synthetic performance of the shaders is roughly equal to Cypress.

[Dezmen]

[zalbard]

And again it's a bad result when compared to GTX570 and GTX580. It does love Nvidia, but still...

[SpuTnicK]

With all that disappointing results here is something to cheer you up guys, true Radeon HD 6970 specification slide.


And here is refreshed card positioning slide

[Factotum]

chiphell is king of photoshop

[Solus Corvus]

Instead of making up more reasons to believe/disbelieve and buying another ticket for the emotional roller coaster, why not just wait a few days instead?

[Biker]

Could someone please suggest to that Ducati750SS dude that it would be a really good time to update his graphic drivers and bench again?

[LordEC911]

Could someone please suggest to that Ducati750SS dude that it would be a really good time to update his graphic drivers and bench again?

Not available to the public, yet. You have to either get really lucky at certain websites or know someone that has them, which is a pretty small list.

[tajoh111]

I don't know about you but don't those numbers seem a bit disappointing?

I mean if this card is 450 dollars. It doesn't seem to really bring anything new to the table over NV. And a 250 dollar 5870 seems like a much better deal.

What happened to ficherodel post with all the benches?

[Nintendork]

Want to know numbers? Sleep or go travel for 4 days. That's the Cayman way of life.

[zalbard]

Want to know numbers? Sleep or go travel for 4 days.

I pick travelling. Hm, Emirates, perhaps? Where do I pick up my tickets, sir?

[Nintendork]

Dubai seems the place xD

[zerazax]

I'm going to say this again: stop looking at synthetics!

First of all:

6950 - 1408 SPs * 800 * 2 = 2.252 TFlops
6970 - 1536 SPs * 880 * 2 = 2.703 TFlops

Now compare this to the other cards:
5850 - 1440 SPs * 725 * 2 = 2.088 TFlops
6870 - 1120 SPs * 900 * 2 = 2.016 TFlops
5870 - 1600 SPs * 850 * 2 = 2.720 TFlops

So in terms of theoretical shader power, Cayman and Cypress are a wash.

Why is this important? Well, look at how 5VLIW is implemented in Cypress and Barts:


For those who don't get the Cayman change to 4VLIW, it goes like this: R600,RV670,RV770, Cypress, and Barts all use the above image's shader configuration. Each SIMD has 4 simple shaders and 1 "fat" complex shader - the transcendental one - that can do complex math (such as calculating pi, sine/cosine, etc.). The other four shaders only do simple MADD (multiply-add). Since most shader operations are just MADD, the 5th shader often sits idle and thus 5-VLIW wasn't as efficient.

The move to 4-VLIW is this: instead of 4 simple + 1 complex, there's 4 moderately complex shaders. Each can still do MADD. However, to do a transcendental, I'm not sure how AMD is going to handle it - it might have one of the 4 shaders iteratively calculate a value, or it might use all 4 to do it at the same time so it requires fewer clock cycles.

Now why is that important? Because synthetics can utilize the 5th shader because they're designed to run em! So, synthetics often got to utilize the 5800's / 6800's up to their full shader potential, which is actually more or less a wash with the 6900's.

However, supposedly real gaming wasn't the case - most vector operations are shorter than 4 values, and so the 5th shader in Cypress/Barts is under-utilized.

Hence, in synthetics, the fact that Cypress is close to Cayman is a reflection of the fact that shading power in Cayman isn't vastly improved for benchmarks. The real key will be how gaming performance goes, since supposedly in games, going to 4-VLIW SIMD brings it to 98% the same performance as a 5-VLIW SIMD. The leaked gaming benchmarks seem to suggest that gaming performance is much higher than 3dMark is hinting at.

This can all be proven wrong of course, but it's the best reason for the gap between 3dMark/synthetic scores (to say nothing of Nvidia's optimizations in the tesselation benches) and gaming scores that were leaked

Also, people forget that the 5800's got HUGE boosts in 3dMark Vantage through drivers. In fact, a few boosted it scores by a considerable few hundreds of points. Who's to say that AMD doesn't do the same and pushes it upwards? The Extreme score on the 6970 already puts it above the 570, so we'll see where it goes. And if this slide is accurate, we still have drivers to use:




But that's why its pissing me off the guy won't use his GTX 480 to bench against the 6970 in games, and will only bench useless synthetics

[tajoh111]

I don't know. A guy that has a with a picture like this



seems to be more reliable than the random screenshots we have been seeing.

Not simply benchmarks, but that crysis score is a bit disappointing. Its scores 10% over a 5870 and matches more or less a gtx 580. Crysis recently has been a strong suit for AMD so I expected much better.



Heaven should be better off the bat, but this still seems well below a gtx 580.

[zalbard]

Perhaps he's sponsored by Nvidia.

[zerazax]

Well, it confuses me why he refuses to bench against the GTX 480 he has - it would be a simple comparison then

edit: some Bulgarian reviewers with a 6950 leaked (then took down the quote) saying it reached 30-40% best scenario faster than the 6870 @ 1680x1050. Well, the translation might be bad, but if that's what they meant, they would place it ~570 performance

And they had a pic of it running too so no reason to doubt they're testing it

[tajoh111]

That ducati guy should have good drivers in the box shouldn't he?

I mean it does appear to have a retail sample which should have good drivers in the box.

[zerazax]

Not simply benchmarks, but that crysis score is a bit disappointing. Its scores \10% over a 5870 and matches more or less a gtx 580. Crysis recently has been a strong suit for AMD so I expected much better.


Heaven should be better off the bat, but this still seems well below a gtx 580.

Regarding Heaven: *everything* AMD scores lower. The 5970 with a fast i7:



But where does the 5970 game compared to the 580? Exactly why synthetics don't represent real world gaming and why Heaven is pointless to compare overall cards with

FWIW this 5970 score is ~5% more than the one provided, which ought to be proof that the 6970 does have vastly improved tesselation compared to Cypress, though probably not up to nvidia levels

Regarding Crysis, I'd want to see the scores based on the same computer and same run and min FPS.That said, there's two games now (F1 and Crysis) where the Cayman is at least around 580 performance, but we need more games to see where the trend is

As I've stated, Fermi saw some games where it took a ridiculous lead over Cypress, and in some games Cypress was awfully close. There was a huge performance delta / wide variance. What'll be important to see is if Cayman has that same performance delta, or if Cayman performs much closer on average (smaller standard deviation) and thus in some games performs close to Cypress, and in others performs higher.

That ducati guy should have good drivers in the box shouldn't he?

I mean it does appear to have a retail sample which should have good drivers in the box.

The 10.12 drivers are supposedly 1GB in size and have a new Catalyst Control Panel interface, which doesn't appear to be what he has. And PowerTune and AMD Overdrive are supposedly coming with that. Wouldn't be the first time drivers shipped with the CD were out of date

[LordEC911]

That ducati guy should have good drivers in the box shouldn't he?

I mean it does appear to have a retail sample which should have good drivers in the box.

Nope, he has the old drivers, 8.79.6

[SKYMTL]

I needed to chime in here.

Nope, he has the old drivers, 8.79.6

And you expect massive performance improvements with a driver that was sent for WHQL certification a few days later?

[Aten-Ra]

The move to 4-VLIW is this: instead of 4 simple + 1 complex, there's 4 moderately complex shaders. Each can still do MADD. However, to do a transcendental, I'm not sure how AMD is going to handle it - it might have one of the 4 shaders iteratively calculate a value, or it might use all 4 to do it at the same time so it requires fewer clock cycles.

Transcendentals will be calculated with 3 out of 4 Shaders inside the SP of the VLIW-4 architecture in Cayman. The problem is that calculating Trancendentals in this way could be slower by 10% vs the old VLIW-5 in Cypress/Barts.

[LordEC911]

I needed to chime in here.

And you expect massive performance improvements with a driver that was sent for WHQL certification a few days later?

Did I say that???

[hurleybird]

I needed to chime in here.



And you expect massive performance improvements with a driver that was sent for WHQL certification a few days later?

Everyone is expecting greater than slightly faster than HD5870 performance. AMD has had over a year since Cypress! Of course seeing these kinds of numbers people are going look for any possible explanation, because frankly the numbers we're seeing right now don't make sense.

I'm still holding onto some hope that Cayman will be much faster, and I'm ready to lay down 500-600 dollars if that is the case. But if this is all that Cayman has to offer, I'm holding off until 28nm.

[Motiv]

I needed to chime in here.



And you expect massive performance improvements with a driver that was sent for WHQL certification a few days later?

I wouldn't, unless AMD are sandbagging. I would expect a far greater improvement over the coming months, due to the different architecture though. Possibly something similar to the "big bang" drivers, that you reviewed for Nvidia.

Wouldn't you say, in your non biased view?

[LordEC911]

Wouldn't you say, in your non biased view?

That's not available at this time...