Re: GTX1080 review thread
Looking over at OcUK,it seems a GTX980TI due to its low boost clock for reference cards,might actually not be massively slower!!!
http://i1056.photobucket.com/albums/...psgiorvjlm.png
http://i1056.photobucket.com/albums/...psl2ujtylq.png
Re: GTX1080 review thread
Liking these reviews but will wait a bit just in case AMD do something exciting.
Re: GTX1080 review thread
Don't think AMD will be competing with that on out and out performance; they're going for mainstream products this time round and looking to push performance and efficiency improvements in the volume market.
it's interesting actually, it's the first time I can remember the two companies specifically going for different markets with their first launch in a generation. NVidia have gone in relatively high, AMD are (purportedly) going in mainstream. Be interesting to see where they line up performance-wise, but it could well be that whether you get nvidia or AMD this time around will depend solely on how much you want to spend (mainstream/mid-range: AMD, enthusiast: Nvidia). AMD are targeting the volume market so hopefully that'll look good for their revenue in the next couple of quarters...
Re: GTX1080 review thread
Quote:
Originally Posted by
scaryjim
Don't think AMD will be competing with that on out and out performance; they're going for mainstream products this time round and looking to push performance and efficiency improvements in the volume market.
it's interesting actually, it's the first time I can remember the two companies specifically going for different markets with their first launch in a generation. NVidia have gone in relatively high, AMD are (purportedly) going in mainstream. Be interesting to see where they line up performance-wise, but it could well be that whether you get nvidia or AMD this time around will depend solely on how much you want to spend (mainstream/mid-range: AMD, enthusiast: Nvidia). AMD are targeting the volume market so hopefully that'll look good for their revenue in the next couple of quarters...
Well,AMD is going with a denser node and it seems the GP104 has 10% less transistors than a GM200. Polaris 10 might actually have a similar amount of transistors when compared to the GP104.
In the end I suspect performance will be more dependent on how high AMD can clock Polaris 10.
Re: GTX1080 review thread
Looks like the latest version of Nvidia Boost really benefits from lower core temperatures:
http://www.techpowerup.com/reviews/N...X_1080/29.html
Re: GTX1080 review thread
I wonder why the 1070 embargo wasn't also lifted today? Halo effect hype I guess?
Interestingly though it looks like the 1070 is quite significantly cut down from the 1080 with 3/4 the cores combined with lower clock speeds and slower memory: http://videocardz.com/60127/nvidia-g...920-cuda-cores
For comparison, 980Ti was 22 vs 24 on the Titan X, plus the 980Ti had generally higher clocks.
It could be a yield thing but it could leave quite a gap between the two and at that rate it looks like the 1070 will be performing at or a bit below the 980Ti mark. Fair enough if the price is low enough. It will be interesting to see what happens with power consumption though - GPU die harvests often draw similar power to the fully enabled dies, or in some cases even more for some reason. Unless they do a good job of gating off the unused cores the efficiency could take a bit of a battering.
Re: GTX1080 review thread
They're going for enthusiast 'volume' with the 1070 I think - massive cut down and conservative clocks (= reduced power) because they need to produce a lot of them to meet demand. Selling it as a $400 980ti won't be too hard either. Polaris 10 looks like it'll end up just a step below, which is what people would have paid for it even if it had been the same speed, so AMD just increase their margins comparatively.
Re: GTX1080 review thread
Oh I'm not saying it's bad, I'm just mulling over some of the numbers.
WRT power consumption, don't just take my word for it, have a look at 970 vs 980, 980Ti vs Titan X, 290 vs 290X, etc - maybe they're not doing a good job of gating the disabled cores (which might change with this generation, hopefully) but power consumption doesn't seem to change much for the harvested chips.
http://hexus.net/tech/reviews/graphi...80-ti/?page=10
http://hexus.net/tech/reviews/graphi...0-ftw/?page=15
Re: GTX1080 review thread
Quote:
Originally Posted by
watercooled
Be careful not to compare overclocked cards to reference w.r.t. power consumption!
Clearest comparison there is the 290 to 290X which as Anandtech note, doesn't have enough of a cut to significantly impact power:
Quote:
based on our performance data we estimate the 290 to have a nearly identical TDP to the 290X, leading us to keep it at an unofficial 300W. Lower-tier parts typically trade performance for power consumption, but that will not be the case here
Also consider effect of voltage binning - which will have been higher for 28nm chips since the spread for finfet is supposedly tighter.
Re: GTX1080 review thread
Yeah I was aware of the non-stock 970. There's a stock one in these charts, it's really not that different: https://www.techpowerup.com/reviews/...Gaming/25.html
Nvidia's reference cards do tend to be a fair bit more efficient than the AIB cards for some reason though, which is important to remember when you're not buying a reference card, or when reference cards basically aren't available like for the lower end stuff.
There's a stock 980Ti/Titan X in the charts too - same story. In fact the 980Ti actually draws more power despite less cores and less VRAM.
The 290/X are pretty close in performance (and AMD are probably just being more realistic with the TDPs to be honest, in fact looking at some more AMD cards like 7950 vs 7970, they're appear to be doing a better job of power gating?), but like I say just compare any harvests vs fully enabled dies and for most GPUs it seems there's not a great deal of difference, if any, for the most part. I'm not just picking on one or two cards, just reeling off a few examples. I don't recall such a significant cut-back for a harvest for some years so I can't think of many good comparison points in that regard.
Re: GTX1080 review thread
This is interesting:
http://videocardz.com/60151/nvidia-g...rmarket-cooler
PCGamesHardware tried the GTX1080 with a better non-reference cooler and said the reference board is power limited!!
Re: GTX1080 review thread
I suppose it's a sensible way of preventing people from pushing silly power through it and cooking the VRM?
Re: GTX1080 review thread
Quote:
Originally Posted by
CAT-THE-FIFTH
Not entirely unsurprising for the reference board, there was info circulating from the event that the GPU wouldn't exceed 180w, though we've seen that it will also use PCIe power. I personally was expecting from the outset that overclockers would have to wait for partners, in particular gigabyte for higher powered overclockers. What's more interesting is what was said about reviewers all struggling to reach the demonstrated 2GHz, which suggests the display sample is either cherry picked (no surprise there ofcourse) or was using a custom board with more power available.
So not entirely unexpected, but its sure to leave the people buying founders for the premium cooler etc etc rather disappointed if they were expecting to overclock.
Re: GTX1080 review thread
Quote:
Originally Posted by
watercooled
I suppose it's a sensible way of preventing people from pushing silly power through it and cooking the VRM?
Yep, and sounds quite similar to the AMD powertune type limits.
Re: GTX1080 review thread
Quote:
Originally Posted by
kalniel
Yep, and sounds quite similar to the AMD powertune type limits.
Nvidia said the card would be great for over clocking - the Founders Edition really looks more and more subpar.