AMD Vega Hardware Reviews

So this is my vega in witcher 3 under the most load i can create.
82h.png
[/IMG]
82h.png


edit: i have a frame target of 60 set, can make it clock higher without it
 
Last edited:
Asus has nice history of screwing up every possible custom Radeon they've tried to do, even cooling AMD chips with lower TDP than NVIDIA chip x seems to be impossible while the NVIDIA chip with higher TDP is chilling along just fine.
For what it's worth, our Strix RX 64 sample did 1050 on the memories without issues, but OCing the GPU was even bigger failure than on reference
 
So this is my vega in witcher 3 under the most load i can create.
82h.png
[/IMG]
82h.png


edit: i have a frame target of 60 set, can make it clock higher without it
How the hell do your eardrums survive 3000rpm on the fan? It must be like having a dremel next to your ear. I’m jumping through hoops here trying to consistently reduce fan noise to acceptable levels. And failing.

Seriously, unless I can find a way to reliably reduce power draw, I’m getting rid of my Vega, emotional attachment to the name or not, never mind that at low noise levels it actually doesn’t perform any better than Polaris at the same noise level.
 
How the hell do your eardrums survive 3000rpm on the fan? It must be like having a dremel next to your ear. I’m jumping through hoops here trying to consistently reduce fan noise to acceptable levels. And failing.

Seriously, unless I can find a way to reliably reduce power draw, I’m getting rid of my Vega, emotional attachment to the name or not, never mind that at low noise levels it actually doesn’t perform any better than Polaris at the same noise level.
1. noise is relative, try two ADHD children constantly poking each other.
2. there is a lot of wood between me and my case
3. i game with head set on mostly.

If a good value replacement HSF for vega comes along then i will grab it, but right now im content.


Anything that loads it to 100% and shader heavy I suppose, superposition for a quick benchmark, 1080p extreme or the 4k and 8k test. By the end of the test it's in the low 100s for a delta of ~40C from the core.

BMPPzUu.jpg

So your just rolling all stock settings?
 
Those are at stock settings, except changes to the fan and temp. settings. I don't wish to put it over 90C so often so I'm running at -15 on power limit which gives 1.3Ghz.
 
Those are at stock settings, except changes to the fan and temp. settings. I don't wish to put it over 90C so often so I'm running at -15 on power limit which gives 1.3Ghz.
K i am downloading superposition now ( going slow)

i did some playing with only changing fan speed while playing witcher 3

2200 fan - frame rate mid 40's clock 1300 - 1450
fy9.png


2400 fan - frame rate high 40s to 50 1350-1500 clock
zv3.png


2600 fan - 50 to 55 1400-1550
49d.png



3200 - 60 ( my target fps) 1500-1600
5p6.png
 
Last edited:
Those are at stock settings, except changes to the fan and temp. settings. I don't wish to put it over 90C so often so I'm running at -15 on power limit which gives 1.3Ghz.
K i just changed to -15 left all my other settings alone, i ran bang on 1300mhz like you for the entire run fan ran at about 2600.

Score was 3096 on 1080p extreme

xye.png
 
Thanks, so it looks like a 11C delta for your card at max. which is damn good. Have you removed the cooler and checked whether the chip is molded or not, otherwise which HBM is your card using, Hynix are supposedly non-molded while Samsung are.
 
Thanks, so it looks like a 11C delta for your card at max. which is damn good. Have you removed the cooler and checked whether the chip is molded or not, otherwise which HBM is your card using, Hynix are supposedly non-molded while Samsung are.
I have not taken the cooler off ( i have tossed up the idea). According to GPU-z i have Samsung memory
 
Thanks, so it looks like a 11C delta for your card at max. which is damn good. Have you removed the cooler and checked whether the chip is molded or not, otherwise which HBM is your card using, Hynix are supposedly non-molded while Samsung are.
while keeping power @ -15.
by setting hbm v @950 s6v @980 s6c 1352 , s7v 1050 s7c 1432 i was able to get my score upto 3300.
GPU power usage ~170watts.

we4.png



edit:
3443
http://gpuz.techpowerup.com/17/09/23/acw.png
172watt
hbm v @950 s6v @980 s6c 1352 , s7v 1050 s7c 1502

Thats a very big jump over just power -15%
 
Last edited:
I have not taken the cooler off ( i have tossed up the idea). According to GPU-z i have Samsung memory
GPU-Z reads that from BIOS, I had Hynix HBM before flashing to Vega 64 and now I supposedly have Samsung HBM2 :p

4zq.png

Here are my temps mining Eth after reapplying my cooler. I see no major difference, at most 1-2C better on GPU. Mine is not molded version ...
Next step is to wait for Kraken bracket and try mounting my old H70 on Vega.


Pictures from the job:
 
Last edited:
K i just changed to -15 left all my other settings alone, i ran bang on 1300mhz like you for the entire run fan ran at about 2600.

Score was 3096 on 1080p extreme

xye.png

I've ran Superposition with Default Wattman settings apart form PowerTune -15%. Card ran at 1400-1440MHz all the time and fan kept 2000-2400RPM. Resulting temps were in 66-68C GPU, 84-90C HotSpot and 72C HBM2.

mkfnm.png




EDIT:

Stock clocks with tweaked GPU voltages and HBM2 overclock to 1050MHz. Power +50%, Fan limited to 3970RPM.

2017-09-23 13:14:36 , 1169.0 , 800.0 , 39.0 , 48.0 , 40.0 , 1784 , 29 , 22 , 107.0 , 4207 , 91 , 1.0000 ,
2017-09-23 13:14:37 , 1061.0 , 800.0 , 38.0 , 39.0 , 38.0 , 1815 , 30 , 0 , 17.0 , 4207 , 91 , 1.0000 ,
2017-09-23 13:14:38 , 1339.0 , 800.0 , 40.0 , 51.0 , 40.0 , 1845 , 30 , 99 , 138.0 , 4234 , 91 , 0.9688 ,
2017-09-23 13:14:39 , 1561.0 , 1050.0 , 43.0 , 66.0 , 43.0 , 1870 , 30 , 99 , 258.0 , 4242 , 91 , 1.0875 ,
2017-09-23 13:14:40 , 1603.0 , 1050.0 , 45.0 , 68.0 , 44.0 , 1909 , 31 , 98 , 258.0 , 4242 , 91 , 1.0938 ,
2017-09-23 13:14:41 , 1608.0 , 1050.0 , 46.0 , 69.0 , 46.0 , 1948 , 32 , 99 , 261.0 , 4242 , 91 , 1.0875 ,
2017-09-23 13:14:42 , 1607.0 , 1050.0 , 48.0 , 72.0 , 47.0 , 1988 , 33 , 99 , 260.0 , 4242 , 91 , 1.0875 ,
2017-09-23 13:14:43 , 1606.0 , 1050.0 , 49.0 , 74.0 , 48.0 , 2024 , 33 , 99 , 264.0 , 4242 , 91 , 1.0875 ,

2017-09-23 13:15:44 , 1581.0 , 1050.0 , 73.0 , 101.0 , 74.0 , 3970 , 76 , 99 , 292.0 , 4242 , 90 , 1.0813 ,
2017-09-23 13:15:45 , 1581.0 , 1050.0 , 73.0 , 101.0 , 74.0 , 3987 , 76 , 99 , 287.0 , 4242 , 90 , 1.0750 ,
2017-09-23 13:15:46 , 1582.0 , 1050.0 , 73.0 , 100.0 , 73.0 , 3986 , 76 , 99 , 285.0 , 4245 , 90 , 1.0875 ,


2017-09-23 13:16:02 , 1583.0 , 1050.0 , 74.0 , 101.0 , 75.0 , 3983 , 76 , 99 , 281.0 , 4245 , 90 , 1.0813 ,
2017-09-23 13:16:03 , 1583.0 , 1050.0 , 74.0 , 102.0 , 76.0 , 3973 , 76 , 100 , 281.0 , 4245 , 90 , 1.0813 ,
2017-09-23 13:16:04 , 1583.0 , 1050.0 , 74.0 , 101.0 , 76.0 , 3970 , 76 , 99 , 282.0 , 4245 , 90 , 1.0750 ,
2017-09-23 13:16:05 , 1583.0 , 1050.0 , 75.0 , 100.0 , 75.0 , 3974 , 76 , 99 , 277.0 , 4245 , 90 , 1.0875 ,


2017-09-23 13:16:21 , 1581.0 , 1050.0 , 76.0 , 103.0 , 77.0 , 3970 , 76 , 99 , 286.0 , 4245 , 92 , 1.0813 ,
2017-09-23 13:16:22 , 1573.0 , 1050.0 , 75.0 , 102.0 , 77.0 , 3972 , 76 , 99 , 280.0 , 4249 , 93 , 1.0750 ,
2017-09-23 13:16:23 , 1580.0 , 1050.0 , 75.0 , 103.0 , 77.0 , 3965 , 76 , 99 , 283.0 , 4250 , 93 , 1.0813 ,


2017-09-23 13:17:04 , 1579.0 , 1050.0 , 77.0 , 104.0 , 78.0 , 3969 , 76 , 99 , 289.0 , 4250 , 94 , 1.0813 ,
2017-09-23 13:17:05 , 1579.0 , 1050.0 , 76.0 , 103.0 , 78.0 , 3975 , 76 , 99 , 286.0 , 4250 , 94 , 1.0813 ,
2017-09-23 13:17:06 , 1579.0 , 1050.0 , 77.0 , 104.0 , 78.0 , 3958 , 76 , 99 , 289.0 , 4250 , 94 , 1.0750 ,


2017-09-23 13:17:26 , 1581.0 , 1050.0 , 76.0 , 103.0 , 78.0 , 3942 , 75 , 99 , 278.0 , 4218 , 95 , 1.1000 ,
2017-09-23 13:17:27 , 1581.0 , 1050.0 , 76.0 , 104.0 , 78.0 , 3963 , 75 , 99 , 287.0 , 4218 , 95 , 1.0813 ,
2017-09-23 13:17:28 , 1580.0 , 1050.0 , 76.0 , 104.0 , 78.0 , 3964 , 76 , 99 , 285.0 , 4218 , 95 , 1.0813 ,

2017-09-23 13:17:32 , 1579.0 , 1050.0 , 77.0 , 105.0 , 78.0 , 3990 , 76 , 99 , 289.0 , 4218 , 95 , 1.0813 ,
2017-09-23 13:17:33 , 1561.0 , 1050.0 , 76.0 , 105.0 , 77.0 , 3975 , 76 , 99 , 288.0 , 4218 , 95 , 1.0813 ,
2017-09-23 13:17:34 , 1552.0 , 1050.0 , 76.0 , 104.0 , 77.0 , 3974 , 76 , 99 , 276.0 , 4218 , 95 , 1.0938 ,
2017-09-23 13:17:35 , 601.0 , 167.0 , 69.0 , 72.0 , 71.0 , 3967 , 75 , 0 , 12.0 , 314 , 33 , 0.9000 ,
2017-09-23 13:17:36 , 140.0 , 167.0 , 67.0 , 69.0 , 69.0 , 3960 , 75 , 0 , 8.0 , 314 , 33 , 0.9000 ,


1z66yb9.png
 
Last edited:
I've ran Superposition with Default Wattman settings apart form PowerTune -15%. Card ran at 1400-1440MHz all the time and fan kept 2000-2400RPM. Resulting temps were in 66-68C GPU, 84-90C HotSpot and 72C HBM2.

mkfnm.png




EDIT:

Stock clocks with tweaked GPU voltages and HBM2 overclock to 1050MHz. Power +50%, Fan limited to 3970RPM.




1z66yb9.png
On just -15 power what does your gpu only power consumption come to? Trying to understand such a large variance. Seems pretty crazy just to be from the CPU goop. Maybe I can still push the voltages higher with -15 and still keep power in the 170 range.
 
On just -15 power what does your gpu only power consumption come to? Trying to understand such a large variance. Seems pretty crazy just to be from the CPU goop. Maybe I can still push the voltages higher with -15 and still keep power in the 170 range.

I've just tested it and it runs at 170W / ~1410MHz / 1.0000 vGPU.
 
All this Vega talk meanwhile I'm still here waiting to buy a humble 470 for less than 300 dollars...(more about 150-130) but life in hard for Gamer these days I suppose. :(
 
Maybe they actually had a fault card?
Literally nothing indicates that they did. It ran faster than reference cards at stock speed, the card simply didn't overclock or undervolt well. That's not indicative of being faulty.

Orly. jpg
Yarly.jpg! Still looking at buying one of these boards, maybe even two if I can possibly swing it, because despite its flaws, vega is way faster than my current card, and also the most modern GPU released so far, and I rather like ATIAMD anyway. I've owned more red team 3D graphics boards than any other so far (2 PowerVR; 1 Chromatic Mpact; 0 3dfx; 5 NV; 9 ATIAMD).

So you can take your ideas of bias and stick it where the sun don't shine, mkay? :devilish: I'll criticize anything I feel deserves it without worrying how hurt your butt will be in the process.
 
Thanks for the testing guys, I have reseated cooler, reapplied TIM and I'm not going to get better temps. for the hotspot unless I go for better cooling. I'm seeing people do 20-30C delta on hotspot even with the better cooling(liquid/morpheus) so I don't think there's anything terribly wrong, it's just too bad that the card has guaranteed 20% performance there for the taking.
 
Back
Top