AMD Vega Hardware Reviews

But only your subsequent posts will indicate if you are not one of those (i.e. if you are not here accusing/defending specific companies).
I dare to disagree. Let me remind the (probably forgotten) AEG case: Long time reputable members were offered free hardware or money to post anti-ATi content.
 
28Mhz looks like a another iteration of the clockbug. How many CPU and GPUs are able to clock that low?
I actually did not make that up with the "by 32" clock divider for deep idle. It's official from AMD. And yes, it is consuming much less in idle compared to Vega FE, which did not yet have this feature enable when it was reviewed.
 
I dare to disagree. Let me remind the (probably forgotten) AEG case: Long time reputable members were offered free hardware or money to post anti-ATi content.

Yes, that is true. But that was his first post ever on this forum, after registering on the 14th, the exact same day reviews went out. Ironically, so far he has not contributed with any facts whatsoever, while stating "futile attempts to sway them with either facts, truth or logic". Not the best intro.
 
Last edited:
With Vega there is a new deep-sleep-state:
https://www.techpowerup.com/reviews...cture_Technical_Overview/images/slides-46.jpg

Computerbase (German) writes that doing nothing yields in ~27 Mhz, moving a window ~50 Mhz and surfing at complex websites ~100 Mhz.
The HBM2 will be downclocked to 167 Mhz, if there is a certain amount of load it will jump to 500 Mhz.
https://www.computerbase.de/2017-08...st/#abschnitt_10_vorteile_der_vegaarchitektur
It would be nice to see the system power usage under idle and common non-intensive application usage... Because a Vega 56 @undervolt could become really interesting..
 

And now the thread of discussion that prompted this response has basically disappeared, including the original post by a new user and my reply to him. I expect the remaining bits including this post to be removed shortly or moved to a new thread as well (in case the others were)?

To you, Ike Turner, I guess you forgot or ignored this one? Or your observations are only valid for the green team?

Yes, that is true. But that was his first post ever on this forum, after registering on the 14th, the exact same day reviews went out. Ironically, so far he has not contributed with any facts whatsoever, while stating "futile attempts to sway them with either facts, truth or logic". Not the best intro.
 
There's a "by 32" clock divider, so the potential idle clocks are - on the contrary - extremely low. It seems not to be enabled in all available drivers though.

viRURXU.png


That's with 17.8.2 Optional, I set the fan rpm manually to a higher level.

That's really nice. I read about the lower clock state for HBM2 but the reviews I saw had 852mhz quoted as the base clock and the idle power draw seemed to back that up. Let's hope these ultra low idle clocks stick around in future drivers.
 
That's really nice. I read about the lower clock state for HBM2 but the reviews I saw had 852mhz quoted as the base clock and the idle power draw seemed to back that up. Let's hope these ultra low idle clocks stick around in future drivers.
They are in the official/optional 17.8.2, not in all the drivers that were floating around during review phase. I guess they are here to stay now. :)
 
So one thing i just noticed,

eyefinity idle clocks
core: 35-70mhz
memory clock: 167mhz

that is a massive improvement over my 290.

here is some partial data:
this is extreme default ( 1600x900) , sry for poor formatting
Code:
settings    Heaven 4.0     Heaven 4.0  no aa    Heaven 4.0 16x tess    Heaven 4.0 16x tess no aa    H4  2x tess
default            2443             3404             2437                      3424                      2949
Default +50             2456                          2448                      3539                      3045
uv1                  
uv2                  
uv3                                                                                    2374                   3301  
oc1                  
oc2                  
oc3                  2520                          2598                      3618                      3229
mem0c1             2518                          2519                      3436                      3179
mem0c2                  
mem0c3             2753                          2699                      3756                      3355

Code:
    Stage 6 volt    Stage 7 clock    power                  
uv1    1050    1100    0                  
uv2    1030    1080    0                  
uv3    1030    1080    50                  
                               
    Stage 6 volt    Stage 6 clock    Stage 7 volt    Stage 7 clock    power    fan target      
oc1        1602        1702    50          
oc2    1100    1602    1150    1650    50    3840      
oc3    1080    1602    1110    1650    50    3840      
                               
    Stage 6 volt    Stage 6 clock    Stage 7 volt    Stage 7 clock    power    fan target    mem voltage    memory clock
memoc1        1537        1592    0    2400    auto    900
mem0c2    auto    1602    auto    1702    50    3840    auto    900
memoc3    1040    1602    1080    1650    50    3840    auto    900
memoc3    1040    1602    1080    1650    50    3840    900    900
 
Yay for you, @Lightman!

Still waiting for news of the ASUS Strix vega edition. Nothing whatsoever so far... :( Oh well, we're not even a whole week into september! So after waiting until june for the Vega unveil, and then until siggraph for another unveil, and then until august for RX edition, and now at some nebulous point this month... Can't remember last time there was a GPU launch that involved quite this much waiting!
 
So i have time for a 5 minute post.

So the conclusion i have reached is that Vega 56 is a TDP limited card 100% of the time. This limit is 200watts for the GPU. The stock voltages will always make you TDP bound. Because of this the only way to increase clock is to lower voltage. changing the power slider away from 0 seems to change clocking behavior and enable higher clocks but you will never go above 200watts for the GPU regardless of how many points you are positive. So then it becomes clocks vs under volting, but then vega seems self aware in what it can handle and only tries to keep to what clock/voltage settings you have set, it will happily go nowhere near stupid settings.

So when you flash your Card to 64 all that changes is that TDP limit, what i found is my highest clock/lowest voltage settings on 56 was consuming around 200 watt on 64 ( this just under 1600mhz). At this point i can start uping both voltage and clocks on 64 and before i had to go to work i was just under 1700mhz consuming mid 250watt on the GPU.
 
AMD marketing feeling the "heat" and attempting to cast doubt on reviews? Seems that's what Hilbert may be thinking based on some comments posted about the Guru3D Strix RX Vega 64 review.

Whenever people see something they do not like it is always the media outlet that is suspect. BTW your IP traced back awfully close to the AMD Markham Canada HQ? Is that a well educated observation to make or an incredible coincidence?

1) The disassembly photo-shoot of the card was done AFTER all performance tests had been done, ergo the card is tested in it's default state. The remark was made due to the fact i still needed to use the card for pending FCAT tests. I also make the photos like that so that people can see how TIM is applied. But I'll rewrite that a bit to make it more clear.

2) We tested in a final configuration. All other previews you refer to are just that, based on preview / non-final early sample cards, some media had quick access to a card for merely an hour or so as ASUS dropped by to show them. No-body tested with long duration temps and DBa measurements with the card properly warmed up, we did. Other future media reviews will back the temps we are seeing, unless our sample had an isolated problem of course.

3) The performance differential in-between reference and the ASUS card is 100% based on throttling, and VEGA64 sure is throttling up and down a lot causing small FPS differences.

But you know what - I hope it is an isolated issue on our side, I honestly doubt it, but I do hope it. But be glad we do not cover up our findings and post results as they are measured here in the lab.
https://forums.guru3d.com/threads/r...-vega-64-strix-8gb.416698/page-2#post-5469841
 
Guru3D took Asus STRIX article off after ASUS request. Reason being, ASUS shipped card with stock AMD Vega 64 BIOS instead of custom one. Funny how manufacturers can get basics so wrong.
 
yeah, the review did reaveal the Strix card to be underwelming without any BIOS magic . Still, It would be nice if they kept their old results too when the new review is published
 
Back
Top