-
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.
AMD Polaris architecture – GCN 4.0
- Thread starter Boomstick777
- Start date
hmm more like -30% than 980Ti, i think ppl are still overestimating performance, i really dont see it beating the 980, trading blows at best.
Well, this puts it 15% behind a 980TI, depends if you trust it or not http://www.3dmark.com/3dm11/11263084
It all matches up, even the clock speeds, submitted before Computex so firstly how do you fake it and secondly how would you know all the correct specifications?
I don't want to be that guy but looking at the Raja interview again and looking at the AoTS part and then looking at our own bench which is run by Kaap and of course we have to consider overclocking to the max but surely I am not the only one seeing that CF 480 performance as a little crap?
Vs
Average FR 97.9, GPU TitanX & TitanX @1501/2002, Normal FR 105.5, Medium FR 98.7, Heavy FR 90.7, CPU 5960X @4.0, Kaapstad
Average FR 96.4, GPU Duo @1172/600,1125/500, Normal FR 112.7, Medium FR 106.7, Heavy FR 77.6, CPU 5960X @4.9, AMDMatt
Average FR 93.6, GPU Fury X & Fury X @1170/545, Normal FR 106.4, Medium FR 95.1, Heavy FR 82.4, CPU 5960X @4.9, AMDMatt
Vs
Average FR 97.9, GPU TitanX & TitanX @1501/2002, Normal FR 105.5, Medium FR 98.7, Heavy FR 90.7, CPU 5960X @4.0, Kaapstad
Average FR 96.4, GPU Duo @1172/600,1125/500, Normal FR 112.7, Medium FR 106.7, Heavy FR 77.6, CPU 5960X @4.9, AMDMatt
Average FR 93.6, GPU Fury X & Fury X @1170/545, Normal FR 106.4, Medium FR 95.1, Heavy FR 82.4, CPU 5960X @4.9, AMDMatt
I don't want to be that guy but looking at the Raja interview again and looking at the AoTS part and then looking at our own bench which is run by Kaap and of course we have to consider overclocking to the max but surely I am not the only one seeing that CF 480 performance as a little crap?
Vs
Average FR 97.9, GPU TitanX & TitanX @1501/2002, Normal FR 105.5, Medium FR 98.7, Heavy FR 90.7, CPU 5960X @4.0, Kaapstad
Average FR 96.4, GPU Duo @1172/600,1125/500, Normal FR 112.7, Medium FR 106.7, Heavy FR 77.6, CPU 5960X @4.9, AMDMatt
Average FR 93.6, GPU Fury X & Fury X @1170/545, Normal FR 106.4, Medium FR 95.1, Heavy FR 82.4, CPU 5960X @4.9, AMDMatt
yea personally thats the part of the presentation i didnt really get, why is the gpu utilisation important to note? was it caped frame rate to like 63fps, even having crossfire barely beating single gpu at a dx12 with async, isn't what i call a good selling point for AMD, the 1080 should be 50-60% faster than RX480, from the leaks so far.
slapping 2 cards to barely get couple fps more, is either showing 60% of bad scaling or lower ppl's expectation of the performance, to like 80% slower than 1080, because at 80% thats basicaly a 970 kind of performance.
so why not show 125 fps instead, or a single card with 62.5 fps
thats probably part of the power saving features that comes with polaris, maybe cross fire uses a lot less TDP in certain situations, but still that slide doesnt make a lot of sense to me.
so why not show 125 fps instead, or a single card with 62.5 fps
thats probably part of the power saving features that comes with polaris, maybe cross fire uses a lot less TDP in certain situations, but still that slide doesnt make a lot of sense to me.
I know, your not the first to say this, i was during the presentation as soon as it came on screen.
Its either vodo maths or they are playing silly beggars and in DX12 Ashes one GPU is really that fast..
Yer, that has baffled me all day and I have hunted for other SLI and CF scores for varying cards but no luck sadly. I just can't see why they would run 2 cards and only get 50% utilization and use all of that as a selling point. Why not just run one card? That would at least give some indication of what to expect. Was the FXs or Titan Xs at full utilization?
I don't want to put a dampener on this card and I do want one for reviewing but not sure I want to get something that matches a much older and slower card in truth.
Exactly or another angle 50% utilisation = cf only burning 130 watts, (well less than full load) to match 1080.
What exactly is "vodo"?
Voodoo
Dude do you seriously believe Nvidia should share merit for a overpriced GPU.
I want polaris to bring down the price of the 1070, so we can have a fair trade off, so I can decide what to buy.
I guess the point was you get 2 GPUs that are faster than 1080 for $200 less.
As for utilization. It's low because CF scaling in AotS is low. Around 30% in most reviews. Which means 1080 is around 25% ahead of single RX480 in AotS.
If the numbers AMD is using are actually legit that is.
Last edited:
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
At some point they'll be like, oh, yeah, that was running VR so that was 62.5fps per eye, and everyone who spent £600 on a 1080 will burst into tears.
I have no idea, in reality, very odd way to do it. It seems like something weird is going on, maybe they were just trying to point out a single card is faster than we thought, maybe the card is epic, they know it and they wanted to make it look bad to bring stock down a little so AMD and execs can buy back as much as they can afford before it sky rockets.
It's one of the single strangest ways to demo performance without at least saying something like they locked it to 65fps frame rate or this was with a 200W power usage running passive or, something. Low utilisation isn't very useful in terms of an explanation for that performance. Is the card unable to get enough power for more, are they early samples with low core clocks, was that running with a very weak cpu and was completely cpu limited.
If they just put a frame rate target to say we can beat that performance at a lower price point, they should have said in incredibly clear terms that the card can do more but it's NDA and full performance is not being disclosed so we've turned the card down enough to be just faster to give a hint of the real performance.
Could just be that Koduri dropped the X2 sample as well and it's not working right any more
Soldato
- Joined
- 28 May 2007
- Posts
- 10,207
At some point they'll be like, oh, yeah, that was running VR so that was 62.5fps per eye, and everyone who spent £600 on a 1080 will burst into tears.
I have no idea, in reality, very odd way to do it. It seems like something weird is going on, maybe they were just trying to point out a single card is faster than we thought, maybe the card is epic, they know it and they wanted to make it look bad to bring stock down a little so AMD and execs can buy back as much as they can afford before it sky rockets.
It's one of the single strangest ways to demo performance without at least saying something like they locked it to 65fps frame rate or this was with a 200W power usage running passive or, something. Low utilisation isn't very useful in terms of an explanation for that performance. Is the card unable to get enough power for more, are they early samples with low core clocks, was that running with a very weak cpu and was completely cpu limited.
If they just put a frame rate target to say we can beat that performance at a lower price point, they should have said in incredibly clear terms that the card can do more but it's NDA and full performance is not being disclosed so we've turned the card down enough to be just faster to give a hint of the real performance.
Could just be that Koduri dropped the X2 sample as well and it's not working right any more
The slide says better performance+efficiency. Do you think they are trying to say that in this instance these cards were scoring higher plus using less power to do so. Obviously due to the 50 percent utilisation. It is weird indeed.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
Yup, AMD are making a 232mm^2 core available at the $200 price point, that puts a 314mm^2 core squarely in the easily, EASILY $400 price point with good profitability.
Thing is seemingly GP106 is quite a bit smaller than Polaris 10 and AMD seems to have superior DX12 architecture(in that they were way ahead with the last gen cards and Pascal hasn't done much to address that, so even with Fury level DX12 performance... it's going to be stronger in DX12), I'd fully expect Vega 11 to be a good chunk bigger than Gp104 and at a more like $350-450 price points, then Vega 10 to be $500-650 and probably bigger than GP102.
GP106 said to be 180-200mm^2 which sounds about right considering a 314mm^2 GP104, any bigger and it would be too close. With Polaris 10 at 232mm^2 I think we're looking at 350mm^2 at least for Vega 11 and if that has HBM2 it will wipe the floor with it in terms of power saving from hbm and bandwidth for 4k not a problem with more power put into the gpu than the memory than Nvidia could match. Again for AMD to go bigger than a 350-400mm^2 Vega 11, Vega 10 has to be a good chunk bigger again and once again if Nvidia don't go HBM2 on GP102, that advantage with power saved will be very difficult to overcome. Look where Fury/Nano perform for the power they use compared to other cards, HBM made a massive massive difference for Fury. Without it Fury would be 20% slower at least.