AMD Polaris 10 Samples work at 1.27 GHz

[wahdangun]

from guru 3d :

One of the major benefits for Nvidia's Pascal architecture is that they can achieve really high clock frequencies. The 16nm Pascal based GeForce GTX 1080 boosts over 1800 MHz. AMD will bake Polaris at 14nm and as some information found in the SiSoft Sandra database, they reach 1.27 GHz with Polaris 10.
The SiSoftware Sandra 2015 OpenGL benchmark entry (look it up here) was entered 3 days ago and allegedly shows an entry of a Polaris 10 sample at 1266 MHz. The chip is identified by its Hardware ID, 67DF: C7 and 67DF: C4. These are both Polaris 10 ASICs, one Pro and one XT model.
The entry denotes once again 36 compute cores (Nvidia call theirs SMs), 36 x 64 shader processors per cluster would mean 2304 active shader processors. The entry also reveals 8 GB memory being clocke at 7.6 GHz running over a 256-bit, so that would be 243 GB/s It however is unclear if AMD opted for GDDR5 or GDDR5X.
Obviously this might be just a reference sample, AMD can still increase the clock frequency at launch, with that in mind it remains to be seen what we finally going to see from the red camp.

http://www.guru3d.com/news_story/amd_polaris_10_samples_work_at_1_27_ghz.html

and btw according sisoftsandra bench actually it was faster than GTX 980TI

 

[dark zero]

I feel that AMD won't go for the speed aspect.. Why I feel that is FX vs Sandy Bridge again? One with better spped and other with better IPC?

 

[wahdangun]

I feel that AMD won't go for the speed aspect.. Why I feel that is FX vs Sandy Bridge again? One with better spped and other with better IPC?

u mean NVDIA going FX route with pascal ?? since it was showing regression in IPC compared Maxwell ??

 

[therealnickdanger]

There is still confusion as to whether this will be a desktop or laptop part, no?

I feel that AMD won't go for the speed aspect.. Why I feel that is FX vs Sandy Bridge again? One with better spped and other with better IPC?

1080 is using speed to make up for its regression in IPC. So you're thinking AMD will have IPC improvements? If AMD can pull off IPC improvements AND achieve higher clocks than Tonga/Hawaii/Fiji, it will decimate Pascal.

 

[Timmah!]

from guru 3d :

http://www.guru3d.com/news_story/amd_polaris_10_samples_work_at_1_27_ghz.html

and btw according sisoftsandra bench actually it was faster than GTX 980TI

Good. Just bring it on ASAP and make pressure on Nvidia. I dont feel like paying over the top for Nvidia GPU cause of lack of competition.

 

[VulgarDisplay]

Well this would explain nvidias founders editions scam. They need to move as many of their cards at the higher price as possible before Polaris arrives and makes them look bad.

The 380 "regular" price for the gtx1070 tipped us all off to how Polaris would perform. The interesting thing now is to see how amd prices p10. They can afford to sell for less and keep the same margins as nvidia because of their smaller die size.

Do we have another 4870 incoming?

If tiny little p10 is as fast or faster than 980ti I can't wait for Vega.

Sent from my XT1575 using Tapatalk

 

[airfathaaaaa]

and according to sisoftsandra its slower than a 390x
still too much testing going on i guess on the clocks

 

[therealnickdanger]

Well this would explain nvidias founders editions scam. They need to move as many of their cards at the higher price as possible before Polaris arrives and makes them look bad.

The overinflated pricing is just Nvidia's SOP. Regardless of AMD, Nvidia was always going to launch 1080 at high-end prices so that they can reduce the price and fill the spot with Big Pascal (1080Ti) when ready. That's what they have done for several generations now.

Polaris is not the 1080 competition as far as we know. Vega is AMD's big stick that will compete with the 1080/1080Ti/TitanP. Maybe surpass it, maybe not, no one knows.

 

[VulgarDisplay]

and according to sisoftsandra its slower than a 390x
still too much testing going on i guess on the clocks

Is the 390x faster than 980ti sisoftsandra because the new test from the op shows p10 faster than 980ti?

Sent from my XT1575 using Tapatalk

 

[airfathaaaaa]

why amd will make vega to compete with 1080?vega will be a true fury line successor its going to be a high/enthusiast level cards the rest will be on polaris

 

[airfathaaaaa]

Is the 390x faster than 980ti sisoftsandra because the new test from the op shows p10 faster than 980ti?

Sent from my XT1575 using Tapatalk

thats the problem...
no one is taking that bench seriously not even the most burned down nvidia or amd fan...talking about it is moot
its good because it gives us numbers and only that but for the rest its meh

 

[Riek]

Is the 390x faster than 980ti sisoftsandra because the new test from the op shows p10 faster than 980ti?

Sent from my XT1575 using Tapatalk

The sisoft test is just some SP/DP/QP performance.
We all know AMD is better at those due to better DP/QP performance. Doesn't tell anything on the potential performance experience on will have using such a card.

 

[guskline]

Until the Polaris is launched and subject to 3rd party review and the GTX 1070 is likewise we are just throwing around numbers and hunches.

I think most posters here, myself included, don't believe AMD will release a gpu fast enough to beat or even equal the GTX 1080. Nvidia probably has that locked up until AMD's Vega launches.

More and more, it looks like the battleground is GTX 1070 at $379/$449FE vs what ever AMD will release in the next month or so at whatever price (fill in the blanks).

Perhaps AMD will pull a rabbit out of it's hat but I doubt it. You sure don't want another "overclocker's dream" moment!

IF the higher end Polaris (numbering aside) can be very close in performance AND price to the GTX 1070 then it's GAME ON especially with the reports that the entire GCN architecture has been improved.

 

[guskline]

why amd will make vega to compete with 1080?vega will be a true fury line successor its going to be a high/enthusiast level cards the rest will be on polaris

It's looking more and more like Vega will be timed to release when HBM2 is ramped to adequate production levels.

 

[tviceman]

It is not an apples to apples comparison between Polaris clock speeds and Pascal clock speeds. Wait for review benchmarks, that is all that matters.

 

[Mahigan]

Under DX12, small Vega will decimate the GTX 1080.
Big Vega will beat big Pascal. Polaris will cover the rest of the market.

Under DX11, harder to tell because AMD haven't released a lot of info regarding their IPC improvements (single threaded performance). If the improvements are significant then we could very well see AMD checkmate NVIDIA market-wide.

 

[railven]

Under DX12, small Vega will decimate the GTX 1080.
Big Vega will beat big Pascal. Polaris will cover the rest of the market.

Under DX11, harder to tell because AMD haven't released a lot of info regarding their IPC improvements (single threaded performance). If the improvements are significant then we could very well see AMD checkmate NVIDIA market-wide.

But will P10 be 90% of GTX 1080 at DX12? That what I want to know.

 

[master_shake_]

amd is really conservative with clock speeds.

look at tahiti.

925 stock people were getting ~1100 on stock voltage.

they were so low that they released a "new" card just to say they were faster than the newest from the other guys.

 

[krumme]

Do we get any nv 1080/1070 mobile variants and if when?
Same for p10/11?
I am in for a new laptop and want something new and shiny. Had nv for prior 3 gen but would like amd this time because of dx12 and long term usage. We keep the laptops for 5 years as they are handed down so to speak.

 

[DownTheSky]

If this news was @ 1.4 or 1.5ghz I'd have something to look forward to.

Plus if you watch the SP performance, it's 5Mpix/s vs 9Mpix/s in favor of 980Ti.

Polaris : Score 1674.64Mpix/s

Single-float GP Compute 5312.60Mpix/s
Double-float GP Compute 527.88Mpix/s
Quad-float GP Compute 29.17Mpix/s

980Ti : Score 1256.49Mpix/s

Single-float GP Compute 9005.12Mpix/s
Double-float GP Compute 175.32Mpix/s
Quad-float GP Compute 7.92Mpix/s

 

[Flapdrol1337]

Under DX12, small Vega will decimate the GTX 1080.
Big Vega will beat big Pascal. Polaris will cover the rest of the market.

Under DX11, harder to tell because AMD haven't released a lot of info regarding their IPC improvements (single threaded performance). If the improvements are significant then we could very well see AMD checkmate NVIDIA market-wide.

If it's really running at "only" 1.27GHz I don't think we'll have a market-wide AMD checkmate.

 

[parvadomus]

If it's really running at "only" 1.27GHz I don't think we'll have a market-wide AMD checkmate.

Too much talk, but nobody knows Polaris capabilities, maybe it can give Hawaii performance at a much lower theoretical TFLOPs and bandwidth. Nobody knows...
It might be as fast or fasther than a 980TI at only 1.27Ghz but who knows?

 

[Paul98]

I was a bit worried that they were having problems getting higher speed. It's good to see this

 

[thesmokingman]

amd is really conservative with clock speeds.

look at tahiti.

925 stock people were getting ~1100 on stock voltage.

they were so low that they released a "new" card just to say they were faster than the newest from the other guys.

It depends. Tahiti and Hawaii did clock rather well. Look at these clocks on water on a very cold night for a reference 7970. Fury however was a real 180 about face in overclocking. I'm crossing fingers on this Polaris clock speed leak. Roughly 1.3 presumably at stock is a good sign. :thumbsup:

http://www.3dmark.com/3dm/4282003

 

[RussianSensation]

I was a bit worried that they were having problems getting higher speed. It's good to see this

Those clocks are still a bit underwhelming. My R9 390 hits 1150mhz on stock voltage. After seeing NV hit 1.7Ghz with ease and 2-2.1Ghz overclocking headroom on FE cards, 1.27Ghz is disappointing from a much smaller chip. I suppose it's possible they are gunning for a much lower 120-130W TDP, even lower than 1070/1080. In this case the only way for P10 to come close to a stock 150W 1070 would be if it has incredible IPC. I think they might aim prices low where $449 FE 1070 will be miles high. I doubt there will be many AIB 1070 cards at $379. If AMD has P10 Pro with 390 performance for $199-229 and P10 XT with 390X+5% at $249-279, the 1070 in comparison is priced so much higher, they won't even target the same buyers.

This strategy is still very risky since NV is going to launch GP106 sooner or later, before Vega drops. NV will close the low end and mainstream Pascal gap far quicker than AMD will close the $350+ gap with Vega. That's one big risk for AMD.