Caporegime
- Joined
- 18 Oct 2002
- Posts
- 30,729
I personally refuse to believe that AMD have taken such a step back from Fiji, and that the stepping rumour (or similar) that we heard about earlier is true...
-
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.
- Status
I don't think it is a step back, just not a step forwards. Clock for clock it looks similar to FuryX.
Somewhere there should be another 20-30% performance. It's hard to believe that can be drivers after all this time . I know AMD is slow with driver development but this doesn't add up.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 30,729
I think we should take this for what it is, and judge Vega when the gaming cards are released. I'm pretty astounded that AMD let it get to this though, random people benching the hot and loud FE cards just to get a glipse of the performance that everyone is desperate for. You'd think they'd want to control the situation a little better...
AND officially said typical clock speed of 1382MHZ for air cooled FE. If you take that figure then they performance numbers get closer to believable. That puts FE at slightly lower FP32 performance than a 1080ti, and and care have recently always required more brute force for the same performance as NV (just comparing Maxwell and Pascal).
So if there had been no architecture improvements and Vega can only hit 1382 in gaming then things start to match up. The question then becomes why is there no architecture improvement, or if there is what else is limiting performance. And why can it only maintain 1382MHz at 300w.
That iswhat Makes this crazy,What were AMD expecting to happen?I think we should take this for what it is, and judge Vega when the gaming cards are released. I'm pretty astounded that AMD let it get to this though, random people benching the hot and loud FE cards just to get a glipse of the performance that everyone is desperate for. You'd think they'd want to control the situation a little better...
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 30,729
Nuts, just absolutely nuts.
I find it incredibly hard to believe AMD has made no architecture improvements. After all the song-and-dance, and all the slides they made about how there are various improvements (including claims like 2x IPC in certain tasks (not referring to the 2xFP16 btw), and 2.6x the triangles-per-second under best-case scenarios using the new primitive discard accelerator), HOW can it still be basically GCN1.999.
GCN was very impressive for its time, but it was time to move on at least a year ago.
And also how can it only handle 1.38 GHz average on a 14nm die shrink? GloFo's 14nm is clearly the worst of the bunch, but the 1050 Ti is built by Samsung on the same process (albeit more mature) and can still handle 2 GHz (though does have far less cores to be fair) AND has higher perf/W than the TSMC build Pascal cards (and normally larger cards have higher perf/W because the memory takes up a smaller % of the overall power the more cores a card has).
A FuryX at 1.4 GHz should match a stock 1080, assuming 1:1 clock-scaling, so these initial results make it look like it's just an overclocked FuryX with 1:1 clock-scaling. All that work, and a die-shrink, for that makes no sense.
I can only imagine/hope that this is a combination of crap drivers and clockspeed issue (is the 550W PSU plus to bad cooler causing the average clock to be low?). If they can add 10% with drivers, and manage 1600-1650 MHz stable with RX Vega and proper coolers, they might be able to snatch victory from the jaws of defeat.
Associate
Looking at a live benchmark video, Vega Frontier Edition is underwhelming and doesn't appear to bode well for RX Vega. Granted, there's always the excuse running around that it's not a 'gaming card', but then one has to ask what difference will the 'gaming card' make for RX Vega?
I don't think that RX Vega will be more than 10% better than Vega FE, so in that case the only hope is that it's priced a lot better than a 1080. Which I'm highly doubting. Seems like Vega is the Fury all over again, but worse since it's a year late relative to the intended competition (the regular 1080). At least Fury came out a few months after the 980ti.
I don't think that RX Vega will be more than 10% better than Vega FE, so in that case the only hope is that it's priced a lot better than a 1080. Which I'm highly doubting. Seems like Vega is the Fury all over again, but worse since it's a year late relative to the intended competition (the regular 1080). At least Fury came out a few months after the 980ti.
Associate
Theres a lot of debating but i still don't get it why everyone jumps the gun on tge gaming performance based on the FE card.
As i wrote a few pages back, gaming and professional cards could have >50% performance difference only because of drivers and cooling.
If the gaming drivers would be ready then they could've released the RX cards as well.
As i wrote a few pages back, gaming and professional cards could have >50% performance difference only because of drivers and cooling.
If the gaming drivers would be ready then they could've released the RX cards as well.
Associate
The primitive shaders needs the driver to have game specific update. Sort of like the game ready drivers. They will work without it, but will be loads faster when optimized.
Great so it's a slightly cheaper (presumably), hotter, GTX 1080.
Gonna keep my RX 480 if this is how it unfolds, at least it can stay under 70 degrees maxed out with no fan noise and use <100 W of power.
And probably not consume 400 W at the same time.
Last edited:
Soldato
I'm not sure why everyone is singing from the same AMD doomed hymn sheet. The FE is a card none of us are buying, it's not the gaming card and is being shown off by what appears to be slightly amateur hour builders or gamers. I'm not judging to hard until RX Vega arrives.
Year it's pretty dumb judging a GPU based on a professional version that AMD have stated isn't as fast nor as optimized for gaming as the RX Vega version is going to be. This is more like a proof of concept and money grab from AMD and probably not representative of what to expect in terms of actual performance from the RX Vega. Don't like the high power consumption though.
i would be happy with 1080 performance but this late in the day i would expect better pricing
If the gaming version delivers 1080 performance, it should be priced at 1070 (or lower) level.
We as consumers should expect price tiers to drop 1 (or more) every year.
i.e. 1080 performance should cost 1070 price ~1 year after 1080 release. Then 1060 price ~2 years after 1080 release.
Incidentally are there any semi-direct comparisons on Nvidia's side. Any gaming benchmarks of the Quadro based on the 1070/1080/1080 Ti?
It'd have to be like £300 if it's a very late GTX 1080 that uses way more power. Still early though, it's possible performance will improve with official reviews.
Caporegime
- Status