• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

ATI cuts 6950 allocation

First post on this board, and its a long one. I wrote this elsewhere, but am posting it abridged here:

My analysis of all the reviews

Going off these 3dMark feature scores:

http://h-5.abload.de/img/69704mxh.jpg

The 6970 trounces the 5870 in everything, except Perlin Noise. Taking a look at Perlin Noise, which is a score reflecting shader power, we get:

5870: 175.42
6970: 146.02

Thus, 6970 is just 83.24% of the 5870's power in shaders. Now why is this significant?

As I wrote before, 5870 is VLIW-5 which is (w,x,y,z,t) where t is the transcendental unit. 6970 is now VLIW-4 which is (w,x,y,z) with 3 of the 4 shaders being used to calculate a transcendental. Now, why the big gap in Perlin Noise if VLIW-4 Cayman has 384 SIMDs and Cypress has 320 SIMDs?

Well first, 1536 is 96% of the 1600 shaders that Cypress has assuming all shaders are firing. If you factor in clocks, it's near exactly 100% of the 1600 shaders. However, that doesn't account for the 83% gap. Two possibilities:

a) The VLIW-4 compiler isn't doing as great a job yet, in which case drivers and optimizations may improve performance

or

b) 3dMark uses a lot of transcendentals, and hence Cayman isn't able to take advantage of the complex t-unit and is getting a performance decrease

So either way you look at it, the 6970 still has room for improvement with regards to 3dMark by improving the compiler and/or optimizing the 3dMark code for VLIW-4. Thus, at this time, 3dMark is not very indicative of actual in game 6970 performance.

(Besides, look at 5870 scores in Vantage at release and today... it's a whopping increase over a year of driver optimizations for a synthetic. Cayman should get even more seeing as how it is a different architecture)

So what's all this mean?

Well, I've been saying it for some time now, but I can see in-game performance putting the 6970 ~GTX 580 levels and the 6950 ~GTX570 levels. The key is that the 69xx improvement over Cypress will range greatly - and hence your perspective of how good the card is may differ.

One of the key things from the release of the GTX 480 and now 580 is that in some games, Nvidia has a whopping lead, and in others, the 5870 barely trails or even takes the lead. That's because Fermi's architecture enabled it to take advantage of certain games far better than the 5870 (esp. in some DX11 tessellation), whereas in others, the 5870's pure shader and texturing advantages bring it close.

However, this creates a wide variation in performance figures - some say Cypress trails only 15%, others say it trails 30%, etc. from the 480. What will be interesting to see is how "stable" Cayman performs relative to the GTX 580/570 - in other words, in games where Cypress trails heavily, does Cayman get a considerable boost over Cypress showing that Cayman is truly different from Cypress? We've seen from the Stalker benchmark, Cayman does get a 33% boost over Cypress so it's quite possible.

IMO this is what we'll see:

6970 will be anywhere from 10 to 50% faster than the 5870 based upon the game - in heavily tessellated games the lead should increase, in games with heavy shaders and little tessellation the lead is probably lower. 6970 should be close though probably trail to the GTX 580 in quite a few games overall though, and show less variation in performance relative to the 580 than the 5870 did. Likewise, in situations were CF doesn't scale well and/or tessellation is heavier (where the 5870 is weak), the 6970 might be very close to the 5970 but in other games, it will trail heavily.

In other words, my theory is that games where Cayman barely improves on 5870 are games with low-to-none tessellation and are heavily shader bound. These will also be games where the 5870 is really really close to the 480/580. Games where Cayman pulls far ahead of 5870 are games with higher levels of tessellation and/or less shader bound. DX11 titles and titles using heavy DXCompute features as well.

That's my assessment of the situation, and why so many people were giving doom and gloom when certain benchmarks were showing Cayman barely edging Cypress (3dMark and some other games), whereas in other benchmarks (such as Unigine Heaven, ComputeMark, Stalker etc.), the 6970 seems to beat the 5870 handily (and often gets close to the 5970).

Stalker COP:
comparep.jpg

Metro2033:
http://www.hardwareluxx.de/community/15873506-post1363.html

What'll be interesting to see is what games the reviewers use to compare. If they use games that are heavily 5870 favored, Cayman might not look great - however, if they show games where the 5870 struggled and 480/580 excelled, its possible Cayman looks amazing. Of course, this will show who's biased to who...

TL;dr - 3dMark and other benches are optimized for VLIW-5, and VLIW-4 isn't optimized yet, and so 3dMark isnt representative of Cayman performance yet. Where Cayman will shine and probably pull ahead is in DX11 games where DirectCompute and Tessellation is necessary. Ultimately, the suite/games tested will determine whether the 6970 looks like a big improvement. When compared to the 5870, it will probably show small improvement in games where shaders are heavy. In games where more TMUs and/or tessellation are required, 6970 pulls ahead. What's most important though is comparing it to the 480/580/570 and seeing if it is a more consistent performer relative to those cards than the 5870.

Oh and Antilles will be a beast
Click to expand...

Moving on...

Scougar said:
Just seem's real bizarre that they would just make the same improvements they made with 68xx (efficiency changes for shaders), improve the tessellator (the buffer increased?) and not actually actively try to improve performance levels above the 5870 significantly.
Click to expand...

First, the tessellator in Cayman is significantly different - look at the chart comparing the 6970 to 5870 and 6870 to 5870. They're different curves.

Second, efficiency in shaders doesn't scale linearly. Doubling shaders doesn't double performance. In fact, look at Cypress.

The Cypress was just 2X Juniper. However, Cypress did not perform anywhere near 2X Juniper despite being on a single die. In fact, recall that in some games, Juniper CF beat single Cypress.

Barts might very well be the "sweet spot" for the Evergreen / R600-derived architecture - that is, a balance of bus size (256-bit) with GDDR5 and a more even mix of shaders with TMUs. In fact, Barts is labeled as a 'rebalanced' Cypress.

Cayman is in much more uncharted waters, and so its efficiency is still up in the air, especially since its shader power is effectively equal to Evergreen (same TFLops).

I don't think AMD relying on a dual card solution to solve this issue is a smart move, in fact I think it's bloody foolish IF they have gone down this path.

I'm not 'pro' AMD, but it is disappointing, I really hope for them that in two days time they actually have a decent single card on the table. I'm struggling to believe they haven't frankly.
Click to expand...

I don't get it. So AMD trailed the single GPU crown from Nvidia by 15-20% each generation and Nvidia was disappointing and doomed, but when AMD trails the single GPU crown by supposedly 5-10%, its AMD that's now doomed and disappointing?

AMD's been doing the dual GPU thing for years now, so I don't see why they'd abandon it all of a sudden, and the roadmap released when Barts came out indicated the dual GPU thing was still going to happen.

AMD doesn't have the luxury to screw up like Nvidia has done because AMD has a much worse financial situation. Nvidia can also cover losses by selling cards to the HPC crowd. Until AMD sorts itself out on the CPU side, the GPU side has to play it safe in all likelihood. However, as AMD's situation has improved, they've taken more liberties too it seems. I wrote this elsewhere:

That being said, AMD has a pretty clear progression where they're going - the only question is whether this is AMD's own doing, or Nvidia's, or both:

3870 - 55nm - 192mm^2 - 70% of 9800GTX
4870 - 55nm - 256mm^2 - 80% of GTX280
5870 - 40nm - 334mm^2 - 85% of GTX480
6970 - 40nm - 389mm^2 - ??% of GTX580

If the ?? is 90-95, that's the closest AMD has been in a long while, and is quite an engineering feat for a < 400mm^2, not to mention that it's only a 15% increase to decrease the performance gap, rather than the 30% jumps required by the previous 3 cards
Click to expand...

Now consider that when the 3870 was released, the 3870X2 was required just to fight the 9800GTX. Then when the 4870 was released, the 4870X2 topped the GTX280/285 and Nvidia had to pull out its own GTX295. Then with the 5870, the 5970 trounced Nvidia to this day. If the 6970 is even closer to the 580 than the 5870 was to the 480, Antilles/6990 is probably absurd overkill.

Personally, I think one needs to look at Cayman like one looked at the 4890. The 4890 was released 10 months after 4870 and was 282mm^2, or 10% bigger than RV770's 256mm^2. The 4890 was released to meet the GTX280 and close the gap from the GTX285 to the 4870. The 4890's changes were mainly in the realm of design for higher clocking cores, which was later carried on in the 5xxx / Evergreen series.

Now Cayman is 15% bigger than Evergreen, and features changes which will later be carried on in future AMD cards, and is designed to meet the GTX 480 and close the gap from the GTX485.. er 580 to the 5870. The card is released 15 months after 5870 (though its possible they shot for October like Barts) but has significantly more changes and potentially performance compared to what the 4890 did.

Thinking like this, the situation is eerily similar.

We'll find out soon enough on the 15th, but I think it's easy to get caught up in some leaked retail reviews without looking at the big picture of why these numbers might be.
 
Last edited:
RavenXXX2 said:
IF this performance is correct where does this leave the 6950, that could be the card to get and overclock if it comes in well cheaper than the 6970. Still does not make sense the small jump in performance we are seeing here, I mean the 5870 was on the 480's heels performance wise and the 6970 is basically on par with a 480 if these results are correct.
Click to expand...

I feel the same what is the point of all this hard work if you only getting a performance jump of about 15% per card 6870 to 6950 to 6970 and what happens if they need to do a 6930
 
Indeed, we are only chasing shadows until Wednesday 5am comes.

The drivers shipped with the cards are not that brilliant but by working for a software house I know that it takes a long time to manufacture stuff so the drivers must be at least 2 to 3 weeks old. I expect surprises tomorrow!!! :)
 
zerazax said:
If the 6970 is even closer to the 580 than the 5870 was to the 480, Antilles/6990 is probably absurd overkill.
Click to expand...

6990 will be slaughter, and bloodbath to 580 and anything else nvidia might pull up of their hat which isnt likely due to the design choices they made with fermi.
Hell even 6850 crossfire totally rocks.
 
flopper said:
6990 will be slaughter, and bloodbath to 580 and anything else nvidia might pull up of their hat which isnt likely due to the design choices they made with fermi.
Hell even 6850 crossfire totally rocks.
Click to expand...
..apart fromt he Micro Stutter, the waiting for updated drivers to play your newest game and power / noise and quirky in-game glitches of course.

Having just come from a 4870x2 to a GTX 580, I have come to realise that benchmarks can be uttery meaningless when it comes to a good gameplay experience.

I haven't tried SLI, but as far as my dual-GPU experience goes with ATI, you can keep your Crossfire for anything other than benchmarking.
 
Scaling in x-fire might be the only thing to save 69## because ATM it looks like a flop.
Have a buyers ready for my 470 at same price I bought them if something amazing happens by wed. But it's looking more and more like a flop performance wise.
 
You must log in or register to reply here.
Back
Top Bottom