More Polaris10 and Polaris11 Specifications Revealed

[deu]

If It's 20-25% faster than a 980ti, and a 980ti can get playable fps in many titles with out AA (which none or less is needed anyway), this stands to do better, no? I mean if you are looking for 60fps+ across all titles, no single gpu can do that... just not sure what expectations these are based off of...

Again: missing the point: We are not saying 1080 isnt a good card we are just saying the people who think polaris is in competition with it is out of their minds.

you can expect the 1080Ti and VEGA refresh to be a better than the first "generation". Im not sure if you think my example deserve a generalization or if you are trolling, but to state the obvious; there is no garanty that the VEGA/1080Ti will ENSURE 60 FPS+. But you can expect them to be around 2,5 times faster than say GTX980. That will for the first time mean that almost ANY game is playable and alot with 60FPS-ish. With one single card.

 

[FMinus]

I think you and many others are getting carried away again. The odds are very good that the full chip has 2048 shaders, and cut down versions will fill the space between this and Polaris 11. All indications are that this is a 480x so it will slot in well below the 1070, just as the 380x was well below the 970. So you'll need to wait until the fall before you see the cards that compete with the 1070 and 1080.

This is a smart move by AMD, because most likely their architecture will be inferior to Nvidia's but they will have several months where they have the best and newest midrange cards, since they'll be competing against Nvidia's Maxwell. How long has it been since that was the case?

FPS/$ will be at least as good as Nvidia's 1070. I predict Polaris 10 will match or slightly exceed 390/970 performance for $250 or less. It will be great if you are in that budget range.

How do you figure that an architecture that is not castrated from compute units is inferior to an architecture that is, is beyond me. All nvidia has right now, is clock speed. Down clock a GTX 1080 to the GTX 980 clocks and the cards would be almost the same.

 

[Nkd]

How do you figure that an architecture that is not castrated from compute units is inferior to an architecture that is, is beyond me. All nvidia has right now, is clock speed. Down clock a GTX 1080 to the GTX 980 clocks and the cards would be almost the same.

People wont believe you. That is the damn truth. Maxwell was hungry for more clocks. Pascal is basically a super refinement of Maxwell. if you clocked them the same there would be no difference or may be a 5-10% performance bump at best. But people won't see that. Pascal has other refinements like VR and stuff other than that its a Super Super clocked maxwell shrunk down. But on the other side it does take significant advantage from clock speeds that maxwell couldn't do on 28nm. So you are right but its also fast so people have the right to brag about it but if you have a 980ti that hits 1500 you will be looking at very minimal performance increase unless you got 1080 purely for the purpose of overclocking it and hope it hist 2.1ghz. I wonder if Nvidia knows that lot of these chips won't hit that so they are selling the founders edition with binned chips.

 

[HumanSmoke]

unless you got 1080 purely for the purpose of overclocking it and hope it hist 2.1ghz. I wonder if Nvidia knows that lot of these chips won't hit that so they are selling the founders edition with binned chips.

The Founder's Editions are still less than middle of the pack. The vendor customs - either using the reference PCB or a custom board - probably with 2 * 8-pin or 8-pin + 6-pin, and the hybrid watercooled cards should attain the levels of a FE pretty easily since the FE is probably board limited to 225W ( 75W via slot + 150W via PCIE 8-pin)

 

[Caring1]

... I wonder if Nvidia knows that lot of these chips won't hit that so they are selling the founders edition with binned chips.

They are not using binned chips for the Founder's Edition.

 

[Nkd]

They are not using binned chips for the Founder's Edition.

May be not. But my point was pascal is pretty much gpu on crack lol. But thats always been nvidia's ammo though ever since they went with this architecture while AMD has been not heavy on clock but they have performed within range of nvidia but have much higher compute power. They went with more balanced architecture but nvidia went with brute force, we see that now. Those insane clock speeds are like Nvidia's ammo. I am sure it will be the same way this time around with polaris.

 

[the54thvoid]

May be not. But my point was pascal is pretty much gpu on crack lol. But thats always been nvidia's ammo though ever since they went with this architecture while AMD has been not heavy on clock but they have performed within range of nvidia but have much higher compute power. They went with more balanced architecture but nvidia went with brute force, we see that now. Those insane clock speeds are like Nvidia's ammo. I am sure it will be the same way this time around with polaris.

Well, if you look at Fury X with 4096 shaders and 980ti at only 2816 (even Titan X only has 3072) you would think Fiji would be romping it all the way home. It's not just the numbers that matter, it is the underlying architectural efficiency. You can't call a design brief to get faster clocks 'brute' force anymore you can cramming in more shaders.

If Pascal is the performance winner because it is blazingly fast - that is what matters. I would love to see AMD push the clocks higher because in the past they could do it well. Perhaps without HBM to hold it back Polaris will come in at a mid point but with overclocking push it above it's fighting weight.

 

[bug]

May be not. But my point was pascal is pretty much gpu on crack lol. But thats always been nvidia's ammo though ever since they went with this architecture while AMD has been not heavy on clock but they have performed within range of nvidia but have much higher compute power. They went with more balanced architecture but nvidia went with brute force, we see that now. Those insane clock speeds are like Nvidia's ammo. I am sure it will be the same way this time around with polaris.

You're so wrong...
The problem was everybody was stuck on 28nm for so many years. Since you could squeeze more transistors onto a die, Nvidia simply decided to scale back on compute power (they already have other cards for this) and repurpose transistors to do actual GPU work. This is how they were able to pretty much run circles around AMD wrt efficiency.

 

[medi01]

Nvidia simply decided to scale back on compute power (they already have other cards for this) and repurpose transistors to do actual GPU work. This is how they were able to pretty much run circles around AMD wrt efficiency.

How much is that (in %)?

 

[HD64G]

How much is that (in %)?

Kepler vs Fiji's CGN is about 10% more efficient in 4K gaming if we compare 980Ti and FuryX and their clock difference and their difference in shader number. As for the newer architectures we should wait more for W1Z reviews for both camps' new GPUs first.

 

[Pinktulips7]

Enough with that AMD BS moving towards to Nvidia, This Company is going down after got slapped by intel and now Nvidia, OMG

 

[ZoneDymo]

Enough with that AMD BS moving towards to Nvidia, This Company is going down after got slapped by intel and now Nvidia, OMG

Slapped by Intel? you mean that illegal under the table price bs they did for which they got a massive fine?

 

[arbiter]

May be not. But my point was pascal is pretty much gpu on crack lol. But thats always been nvidia's ammo though ever since they went with this architecture while AMD has been not heavy on clock but they have performed within range of nvidia

Part of the reason AMD cards don't overclock well, is they are generally clocked near the max of the chip as it is to compete with nvidia who's chips seem to be clocked pretty low from what a lot of them CAN do.

Kepler vs Fiji's CGN is about 10% more efficient in 4K gaming if we compare 980Ti and FuryX and their clock difference and their difference in shader number.

Is that really cause fiji is faster or only comes down to memory bandwidth difference? i would bet its the ladder of that not so much first part.

 

[Pinktulips7]

Slapped by Intel? you mean that illegal under the table price bs they did for which they got a massive fine?

AMD was getting slapped by intel every year since C2D even though AMD Fanboy was mot happy about it, AMD still in business because of intel!!! I am dine with crappy AMD, Time to move on, Since that Idiot Raza took over AMD everything going downhill.

 

[G33k2Fr34k]

AMD was getting slapped by intel every year since C2D even though AMD Fanboy was mot happy about it, AMD still in business because of intel!!! I am dine with crappy AMD, Time to move on, Since that Idiot Raza took over AMD everything going downhill.

Well... if I understand the point you're trying to make correctly, I'd say you're partially right. AMD has been behind Intel when it comes to high performance CPUs. What they did with Bulldozer, while it looked good on paper, didn't pan out.

Bulldozer was supposed to be their counter strategy to Intel's lead in the silicon fabrication technology. So instead of competing against Intel using the same approach, that is building big cores and prioritizing single thread performance, they decided to do something different, aka bulldozer, which evidently didn't succeed. That is behind us now. AMD is releasing Zen this year. A Zen core will have double the integer resources and quadruple the floating point resources of a Bulldozer derived CPU core, in addition to high bandwidth low-latency L3 cache. Expect Zen to compete with intel's kaby Lake.

With regard to Polaris, I'm quite glad that AMD chose to release their mainstream cards first. I'm not planning on spending more than $350 CAD (~$250 USD) for a graphics card, and I think AMD's 480/480X are going to be great performers in that price category.

 

[HD64G]

Is that really cause fiji is faster or only comes down to memory bandwidth difference? i would bet its the ladder of that not so much first part.

Upgraded CGN core was the reason. And Polaris will get an even better one...

 

[deu]

Well... if I understand the point you're trying to make correctly, I'd say you're partially right. AMD has been behind Intel when it comes to high performance CPUs. What they did with Bulldozer, while it looked good on paper, didn't pan out.

Bulldozer was supposed to be their counter strategy to Intel's lead in the silicon fabrication technology. So instead of competing against Intel using the same approach, that is building big cores and prioritizing single thread performance, they decided to do something different, aka bulldozer, which evidently didn't succeed. That is behind us now. AMD is releasing Zen this year. A Zen core will have double the integer resources and quadruple the floating point resources of a Bulldozer derived CPU core, in addition to high bandwidth low-latency L3 cache. Expect Zen to compete with intel's kaby Lake.

With regards to Polaris, I'm quite glad that AMD chose to release their mainstream cards first. I'm not planning on spending more than $350 CAD (~$250 USD) for a graphics card, and I think AMD's 480/480X are going to be great performers in that price category.

Ye AMD's problem was that they where too far ahead in expecting multicoresupport (which showed to be not so easy to scale no wish to support in most applikations.) That left them with single / dualcore performance down 20-40% and when trying to out-core intel and the market does not develop that way then they loose Zen will give intel a run for their money i am sure!

 

[EarthDog]

Instructions Per Clock is what you are talking about.

 

[Pinktulips7]

Well... if I understand the point you're trying to make correctly, I'd say you're partially right. AMD has been behind Intel when it comes to high performance CPUs. What they did with Bulldozer, while it looked good on paper, didn't pan out.

Bulldozer was supposed to be their counter strategy to Intel's lead in the silicon fabrication technology. So instead of competing against Intel using the same approach, that is building big cores and prioritizing single thread performance, they decided to do something different, aka bulldozer, which evidently didn't succeed. That is behind us now. AMD is releasing Zen this year. A Zen core will have double the integer resources and quadruple the floating point resources of a Bulldozer derived CPU core, in addition to high bandwidth low-latency L3 cache. Expect Zen to compete with intel's kaby Lake.

With regard to Polaris, I'm quite glad that AMD chose to release their mainstream cards first. I'm not planning on spending more than $350 CAD (~$250 USD) for a graphics card, and I think AMD's 480/480X are going to be great performers in that price category.

I am using AMD (Graphics) since ATI 9700 Pro!!!! What a awesome Card ATI released that Nvidia was shitting in their Pants then AMD bought/Took over ATI and Everybody's knows what happened after that. AMD needs to stop rebranding their Cards every year!!!!!!!!!!!!!!!

 

[rruff]

Ye AMD's problem was that they where too far ahead in expecting multicoresupport (which showed to be not so easy to scale no wish to support in most applikations.)

Don't you have that backwards? They used more cores to make up for their poor single thread speed.

Based on AMD's recent track record, I think it would be incredibly optimistic to expect their new designs to be on par with Intel and Nvidia.

 

[HumanSmoke]

I am using AMD (Graphics) since ATI 9700 Pro!!!! What a awesome Card ATI released that Nvidia was shitting in their Pants then AMD bought/Took over ATI and Everybody's knows what happened after that. AMD needs to stop rebranding their Cards every year!!!!!!!!!!!!!!!

To be fair, ATI's star was already fading by the time AMD acquired them (which made the $5.4bn paid all the more perplexing). Their top end (9800 Pro/XT, then X800/X850) served them well, but they were rapidly losing market share across the board and Nvidia was killing them in mobile discrete. Even with R600 getting a collective handjob from the press well ahead of launch, and guys like Charlie doing the FUD thing claiming Nvidia's G80 wouldn't have a unified shader architecture, ATI's position in the market was taking a bit hit. Funnily enough, ATI's revenue- even during their relatively dire last year, was twice that of present day AMD.

 

[arbiter]

Upgraded CGN core was the reason. And Polaris will get an even better one...

wrong, the massive Memory bandwidth it had was what gave it the advantage at 4k not cause it was faster. Its been pretty well documented by review sites over the last year.

Don't you have that backwards? They used more cores to make up for their poor single thread speed.

Based on AMD's recent track record, I think it would be incredibly optimistic to expect their new designs to be on par with Intel and Nvidia.

Hopeing to be on even terms with intel with 1 more cpu is little to optimistic. Best to shoot for say their 8 core able to directly match intel's 6 core to start which would be a bit of a jump as it is.

 

[medi01]

BS moving towards nVidia

That somes it up perfectly. Moving BS:

 

[refillable]

Of course Polaris 10 won't compete with Pascal. It never was supposed to. It should compete with the R9 390 in a big way, and if AMD's new insulation technology along with other improvements allow for MUCH higher overclocks, then 32CUs might be all you need to reach near-980ti levels of performance. NEAR-980ti though. Let's not get ahead of ourselves.

Yes, sure, it's probably going to be a great card, but that'll leave a huge performance gap, for a long time (before Vega comes out) between AMD's newest and NVIDIA's newest, which may not be good for, me at least, a PC Builder. 32 CU is not confirmed but unless they did some miracle engineering, 32 CU, at best is probably just under Grenada (390), pure speculation, though.

Anyways, BS is usual for any company, NVIDIA and AMD are notorious for doing them.

Oh yes, one thing to say, I really hated, I mean REALLY hated that they decided with the new name "Fury" (Same goes for Titan). Very confusing, I wish they'll do better on naming scheme department, maybe calling Polaris 480, 470 and Vega 490 and maybe 495. Or heck, Polaris can be 470 and 460, while 480 and 490 for Vega.

 

[HumanSmoke]

That somes it up perfectly. Moving BS:

Nice try with the trolling. The presentation and the slide are specifically Deep Learning orientated. Mixed precision - use of FP16 for object detection and recognition allied with a more comprehensive GPU to GPU/GPU to CPU interconnect. That is why the second half of the slide show the representation of four GPUs, because P100 - the focus of the presentation - has four NVLink interfaces.
I'm pretty sure only trolls and AMD fanboys try to make "Pascal 10X Maxwell" applicable across the board, rather than a the sole example it is intended for.