cancel
Showing results for 
Search instead for 
Did you mean: 

Rig Showcase Discussions

colesdav
MVP

Re: Oscar Mike

Hi,

Thanks for the information.

RE: AMD's Fine-Wine Technology ... yes new features and better performance is likely to come in with time. I agree.
Some AMD Critics are saying that "Fine-Wine" is just an excuse for failure to produce poor drivers but I do not think that is fair.
Nice to see a pair of RX480 still beat a GTX 1080 ...

Any chance you could overclock the HBM2 memory to 1000 MHz and see how your score improves?
If what I have read is correct, then  this should give you [(1000-945)/945]*100/3 = 1.94% performance improvement on Graphics Score 1.

Thanks again.

0 Kudos
colesdav
MVP

Re: Oscar Mike

And if you can get there ... 1051 MHz on the HBM2 Clock should let you hit ~ equal graphics score to the GTX 1080.

Can you please tell me what Power Mode you are running in?
Do you know the power draw?
Cheers.

0 Kudos
kxuping
Forerunner

Re: Oscar Mike

Sadly Crossfire (DX11) do not scale well on all games. I would love to see 1080/Vega64 fps with my RX580 Crossfire on all games.

0 Kudos
colesdav
MVP

Re: Oscar Mike

I had thougth AMD had been doing pretty well with Crossfir(DX11)  for the RX 480/580 on AAA titles over the past couple of years?
One single powerful card a much better idea though.

0 Kudos
kxuping
Forerunner

Re: Oscar Mike

There are games were it scale awesome. Others not so much. So its important to check how crossfire performs on the games that you want to play.

0 Kudos
colesdav
MVP

Re: Oscar Mike

Yes sure ..
Have you tried any MultiGPU DX12 games yet?
The ones I have been running have been scaling very well, and running nicely.

0 Kudos
ajlueke
Grandmaster

Re: Oscar Mike

Always good to have inspirational individuals in the community to remind us of what is actually important 

I was initially taken aback by the power and heat numbers of Vega 64, but I supposed that is just another fun engineering challenge to try and overcome.  Playing around with hardware like this can be more rewarding than actually running any software on the resultant device.  You have inspired me sir.  I think I will pay a visit to my shop this weekend and try out some of this liquid cooling for myself.

whiskey-foxtrot
Forerunner

Re: Oscar Mike

I have some cables coming next week so I'll be able to calculate the power draw (through the 2x4 power cables). Since I can't measure what's going through the PCI-e slots, I'll have to rely on various base measurements: 4 different models, same motherboard, measuring from the wall AND the 2x4 PSU cables. It will not be 100% accurate, but with multiple deltas it should be very close.

I'll be testing with 1xEVGA Kingpin 1080Ti, 1xEVGA 1080 SC3, 1xRadeon FE, and of course, the Sapphire Radeon RX 64.

whiskey-foxtrot
Forerunner

Re: Oscar Mike

This is the kind of green I'd like to see in regards to my system I know it isnt much of a difference (GPU scores) and this is just a synthetic benchmark, but I had to do it anyway. I simply increased the mem frequency a bit and turned the fans to max. I can't wait to actually work on the water blocks so I can focus more on tuning these.

vega-win.png

The actual score is posted here: Result

0 Kudos
colesdav
MVP

Re: Oscar Mike

Thanks for that.

I am looking at something else at the moment, before I respond with more detail can you please confiirm that you did not change the Power Profile at all?
Which profile were you running with previously and which one are you running with now?

Does it look like the Memory Frequency Scaling Numbers from BuildZoid/Gamers Nexus (Based on a Vega 56 Overclock) that I posted are actually applicable to the Vega 64 card?
Please note Higher HBM2 CLK seems to improve GCLK scaling as well.

I understand I might be asking you for lots of info but it might be important info to get fed back to AMD and other Users.
I would do it myself if I had a Vega 64.
If you do get the time could you please look at the following. You need a Wattmeter.

1. Push up the HBM2 CLK, drop GCLK, and get the same performance at reduced power level.
2. Undervolt the GPU Core and maintain the same HBM2 CLK, and GCLK to maintain performance but reduce power further.
3. Maintain undervolt but attempt to push GCLK up to increase performance.

If many other users are seeing good performance improvement versus HBM2 Clk increase (HBM2 power is low 20-30W for the Memory/ Memory+Controller). 

AMD might be able to do BIOS / Driver tweak later on and put the Vega 64 in a better place versus the GTX 1080/1080TI in terms of Performance / Power.

Cheers.

0 Kudos