It's possible, but only if Big Vega is at TSMC 16FF. Obviously it would have to be clocked a lot higher than a "slight clock increase". Is that realistic? We don't have any idea of how much GloFo is holding it back.
I see no way Vega is going to reach OCed GP104 performance.
My guess ? small vega should be even with the 1070 , and big vega should just edge out the 1080 by 5%.
But right before the release of big vega, Nvidia will refresh both the 1070 and 1080. And the new 2070 will be 2% slower than big Vega and the gtx 2080 will smoke the big vega for $200 more money.
So how they lied then ??
Just a typo. The test was shown in real time, obviously it wasn't fake, it was just carefully manipulated and misleading. Likely they had the best binned sample card they could find, and they put on the 60 FPS cap and used medium settings so that the P11 card could be underclocked and undervolted. A 750 Ti can easily do 60 FPS at medium settings. The 950 may have been selected as a particularly power-hungry one and the clocks left at default.I'd say using a 4790K with DDR4 would be the big giveaway here.
Try finding a single test where Polaris draws 54W less than a (reference) GTX 950. That would make its performance per watt leaps and bounds above Maxwell, probably edging Pascal as well. Hard to believe. Anyway, this is the Vega thread so let's save this discussion for the other (existing) threads.
Serious question: Would this video by JayzTwoCents be a valid data point for you that Polaris can be surprisingly power efficient if the voltages are kept in check?Try finding a single test where Polaris draws 54W less than a (reference) GTX 950. That would make its performance per watt leaps and bounds above Maxwell, probably edging Pascal as well. Hard to believe. Anyway, this is the Vega thread so let's save this discussion for the other (existing) threads.
It's possible, but only if Big Vega is at TSMC 16FF. Obviously it would have to be clocked a lot higher than a "slight clock increase". Is that realistic? We don't have any idea of how much GloFo is holding it back.
One does get the impression that the binning on Polaris family chips is...
https://www.youtube.com/watch?v=zWASNajSdpg
XFX RX 480 GTR fully, constantly stable at 1465 MHz core clock. New revision chip?
Shintai, ask the same question: where is full Polaris 11? All of this may indicate refresh of AMD GPUs.
No it isn't. RX 460 has 14 CU's, full Polaris 11 have 16 CU design. Polaris is in mobile. Maybe you have not payed attention?It still only ships at 1338Mhz
RX 460 is full Polaris 11 isn't it. if they got something more, they better come up with it fast. Since GP107 is 50-75% faster at 75W or below.
But that still doesn't answer the question. Pascal is everywhere in mobile, where is Polaris?
https://www.youtube.com/watch?v=zWASNajSdpg
XFX RX 480 GTR fully, constantly stable at 1465 MHz core clock. New revision chip?.
https://www.youtube.com/watch?v=zWASNajSdpg
XFX RX 480 GTR fully, constantly stable at 1465 MHz core clock. New revision chip?
Shintai, ask the same question: where is full Polaris 11? All of this may indicate refresh of AMD GPUs.
What if Glofo isn't holding AMD back? How about its just the limitations of the uarch? You know, Pascal didn't clock high due to 16FF+ alone, it clocked high due to a fully optimized uarch.
Oh I wouldn't expect Vega to clock as high as Pascal does; but 1500-1600 Mhz maybe?
No, it was running at roughly 1.05V stock and 1.18 overclocked, thus at lower than release voltage.Was it overvolted? And it can be a golden chip or even a special overvolted chip for a reviewer, I remember some AIBs getting caught over this.
There is likely a refresh considering AMD's new scheme naming with xx5 cards.
If Vega has 12 TFLOPs, according to VideoCardz, FUDZilla and WCCFtech, FP32 compute performance and 64CU design then it has to have 1.5 GHz core clock, to achieve this.And what do you base that on? It would have to clock like no other GCN had before, including 1.3 on 14LPP. They are already what, 200W cards at OC? A 1500-1600Mhz Vega card would be what, 500W+ and water cooled?
There is a point that Embedded GPUs have higher clocks/lower TDP than desktop GPUs.I think the revision should show up in gpuz, if there's a new batch we'd know soon enough on overclocking forums.
1.4-1.5Ghz on Fiji specced Vega should land it midway 1080 and Titan XP, not bad but AMD have a great chance against <500mm2 chip from nvidia for the top dog halo here and hopefully they have something better.
And presumably, the process sync between GloFo and Samsung, and desktop Polaris arch, as a result for this - just a pipecleaner for WSA, from AMD.No, it was running at roughly 1.05V stock and 1.18 overclocked, thus at lower than release voltage.
Power was ~ 100W gaming, ~ 133W [1475 Mhz] overclocked and ~90W in a heaven loop.
Design revision, an improved process, or both, which if true, bodes very well for Vega to hit high performance at competitive wattage.
If this is truly indicative of Polaris perf/watt, then all of the previous arguments as to Vega's inability to compete have been rendered worthless.
And what do you base that on? It would have to clock like no other GCN had before, including 1.3 on 14LPP. They are already what, 200W cards at OC? A 1500-1600Mhz Vega card would be what, 500W+ and water cooled?