• 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.

Poll: ** The AMD VEGA Thread **

On or off the hype train?

  • (off) Train has derailed

    Votes: 207 39.2%

  • (on) Overcrowding, standing room only

    Votes: 100 18.9%

  • (never ever got on) Chinese escalator

    Votes: 221 41.9%
  • Total voters
    528
Status
Not open for further replies.
Mauller said:
This is the heatsink from Vega FE, its just as unimpressive as the reference Nvidia cooler. Aftermarket Air cooling should easily work better.

DSC03544.JPG


full teardown here.

Click to expand...

It will be interesting to see if some of the performance dips are down to the GPU hitting thermal and power limits too. If the GPU is kept cooler it should help drop power consumption AFAIK.

BTW,GN also said their estimate of the die size is around 526MM2(instead of the 564MM2 from PCPER),but AMD mentioned it would be slightly under 500MM2.
 
humbug said:
Whatever it is if we are going to go off the PcPer results and assume those are final gaming performance -10% or so it nothing like enough, that would make it 10% faster than my 1070.

Is this a $250 and 280 watts, thats not good enough, not by a very very long way, unless its really cheap to give the $250 crowed and upgrade path.
Its a Fury-X + 15%, thats shocking and with the cost to AMD of the thing surley AMD couldn't actually afford to make it and put it on the market? that would cost as much as they could sell it for.
Click to expand...

I'm not assuming that's all the RX Vega will manage, As I said it should be good for another 10% and hopefully a bit better if it's under water, But even with that it's not great because of how inconsistent it is, I had been hoping new things such as whatever it is that lowers the overhead would have made it more consistent across a larger range of games because that's where GCN has been letting itself down. But it doesn't seem to be the case from what the Frontier's shown and as stated AMD have said the current FE driver has all the latest optimisations already. And I can't see there being some miraculous improvement over the next 30 day's that they wouldn't have already done in the last 6 months.
 
akarypid said:
So to summarise where we think we're at:

  • AMD focused on Vega's compute performance first and foremost. They have been vocal about attacking the strategic cloud/deep-learning sector and have invested a lot in ROCm and technologies (HBCC, SSG, HBM) that mostly make sense there.
  • This is reflected in their driver work, where the 'compute' part is ready, hence Vega FE. AMD thought it's ok to launch the FE because it's not a 'gaming card'. As a prosumer card, as long as compute is ok, the gaming speed is not the end-all-be-all.
  • Thus AMD just certified the very-first stable drivers and bundled them for the launch. Hence Vega FE sucks in gaming.
  • RX Vega will launch with 'proper' gaming drivers and we'll see its true potential.
  • Vega FE users will get a nice gaming performance bump when those drivers are released.
Or at least that's the hope.
Click to expand...


We haven't seen any benchmarks of the released Vega doing compute workloads. Best we've seen is it is really bad at mining, but that might be the HBM2 latency. Putting that aside, The actual FP32 performance is pretty similar to 1080ti, infact give the clock speeds seen in practice it has lower compute performance. The exception should be deep learning that can use FP16, but there are no benchmarks

If you mean professional software use, the Vega Fe is far behind the P5000 which is nothing more than a 1080.

The gaming performance is actually much better than the professional software performance.

The whole thing is just misdirection by frantic AMD fans. The drivers for DX and openGL are newer than the latest crimson 17.6 drivers. There is zero evidence that features are not working properly. They might not be, but you can't dismiss that easily.

At the end of the day we have an AMD card with less actual FP32 performance using typically clock and less bandwidth than a Nvidia card (ti) when traditionally AMD needed more of both to be competitive. It's not that unexpected. To close the gap AMD would need big IPC gains but those are not always easy to get.

The bigger mystery is probably the die size and what else is on that chip. There could be a lot of features that dramatically improve performance in the future when specially coded for in new games. There could be a lot.of stuff the the HPC market that just doesn't help with games.
 
CAT-THE-FIFTH said:
It will be interesting to see if some of the performance dips are down to the GPU hitting thermal and power limits too. If the GPU is kept cooler it should help drop power consumption AFAIK.

BTW,GN also said their estimate of the die size is around 526MM2(instead of the 564MM2 from PCPER),but AMD mentioned it would be slightly under 500MM2.
Click to expand...

The actual GPU will will be slightly smaller than you can measure due to the packaging
 
nashathedog said:
I'm not assuming that's all the RX Vega will manage, As I said it should be good for another 10% and hopefully a bit better if it's under water, But even with that it's not great because of how inconsistent it is, I had been hoping new things such as whatever it is that lowers the overhead would have made it more consistent across a larger range of games because that's where GCN has been letting itself down. But it doesn't seem to be the case from what the Frontier's shown and as stated AMD have said the current FE driver has all the latest optimisations already. And I can't see there being some miraculous improvement over the next 30 day's that they wouldn't have already done in the last 6 months.
Click to expand...

Maybe, be that as it may its nothing like good enough for a 500mm 280 watt GPU with HBM 2, its a $250 card at best and i doubt AMD would get good enough margins on that to put it into production.

Its another Fury-X + a little more two years after the Fury-X was already an under-performing card.

Its a high power consumption 1070, that's where PCPer resulted its performance... it has no place unless its very cheap. VERY cheap.
 
humbug said:
Maybe, be that as it may its nothing like good enough for a 500mm 280 watt GPU with HBM 2, its a $250 card at best and i doubt AMD would get good enough margins on that to put it into production.

Its another Fury-X + a little more two years after the Fury-X was already an under-performing card.

Its a high power consumption 1070, that's where PCPer resulted its performance... it has no place unless its very cheap. VERY cheap.
Click to expand...

Price is what will make or break it, I'm hoping it's reasonable but when you look at the FE price and factors like how it has HBM2 it may not be worth what they'll want for it and then for me it'll be a toss up over whether I'm willing to be overcharged in order to keep Freesync. I'm hoping it will be but the sign's aren't good so far. Another month and we'll know for sure.
 
humbug said:
They were also told the gaming performance on the RX version of the card would be much higher, again PcPer ignored that and reviewed the FE card as if its a gaming card at its final gaming performance.
Click to expand...
Not caught up with the thread yet... but have to post here. Didn't AMD say "zero to 8 percent higher"... not "much higher" perf?

Where are people (not just yourself) getting this "much higher perf with RX Vega" from? AMD didn't say that, did they?

This is probably answered by the time I catch up to the end of the thread tho, knowing my luck :p
 
D.P. said:
We haven't seen any benchmarks of the released Vega doing compute workloads. Best we've seen is it is really bad at mining,
Click to expand...

The first guy streaming did some productivity benchmarks near the start of his stream where it did ok to excellent but nothing that would define the card.
 
FoxEye said:
Not caught up with the thread yet... but have to post here. Didn't AMD say "zero to 8 percent higher"... not "much higher" perf?

Where are people (not just yourself) getting this "much higher perf with RX Vega" from? AMD didn't say that, did they?

This is probably answered by the time I catch up to the end of the thread tho, knowing my luck :p
Click to expand...
It was the PCper guy who said he thought there would be from zero up to an 8 to 10 percent increase in performance for the RX release, game dependant.
 
harney said:
That is why i gave up and started drinking GnT's
Click to expand...

no love for a simple magners with ice?

look it's safe to say there probably will be a bit better performance from rx vega, mostly down to a driver update (as amd always seem to have poor release drivers) but they'll take a few months at best judging from previous cards, that and the fact better cooling will help with throttling.

but I do think people need a reality check in thinking there's going to be some magical 30-40% performance gain.

at this point in its life, expect GTX 1080 performance and you probably won't be disappointed, best case is it does get better.
 
Last edited:
Status
Not open for further replies.
Back
Top Bottom