Overclock.net › Forums › Graphics Cards › AMD/ATI › [Official] Polaris Owners Club
New Posts  All Forums:Forum Nav:

[Official] Polaris Owners Club - Page 48

post #471 of 4151
KarathKasun - Yeah. If I up the "memory" voltage to +100 and I can do 2220 without memory errors in 3dmark. Before I'd still get memory errors when it down volts. It's practically worthless though because I hit power limit faster and my score is lower or the same. 2210 is error free on auto so I leave it there. You mention valley scores of 2500. How are you doing that? That's my score at 1435 and 2220 on a rx480?

edit: I don't need as much extra voltage to be error free in valley at 2220. Also the "Memory" is probably the voltage for the memory controller. the VRAM chips on this card are rated at 1.5 volts. Someone else said early on in the BIOS mod thread that we don't have direct control over vram voltage and that we probably wouldn't ever.

GecKoTDF - I checked my 3dmark firestrike ultra temps at my current oc of 1435 and I get 47 max. There's a difference, but like I said temp isn't what's holding either of us back so your solution is much more cost effective







Unigine Valley Benchmark 1.0







FPS:
59.9
Score:
2505
Min FPS:
30.7
Max FPS:
115.6

System






Platform:
Windows 8 (build 9200) 64bit
CPU model:
Intel(R) Core(TM) i7-4790K CPU @ 4.00GHz (3998MHz) x4
GPU model:
Radeon (TM) RX 480 Graphics 16.200.1035.1001 (4095MB) x1

Settings






Render:
Direct3D11
Mode:
1920x1080 8xAA fullscreen
Preset
Extreme HD

Powered by UNIGINE Engine


Edited by greytoad - 8/12/16 at 8:02am
post #472 of 4151
@greytoad - Totally agree! - you are using 1435 mhz GPU for 27/7 and what about mem clock and voltages? - I see a few errors memory in HWiNFO what really don't understand what that mean, i mean... understand there is "errors" but... i don't see glitches or something like that, what mean that?
post #473 of 4151
I'll be honest I'm not sure what kinda of fault HWINFO is monitoring, but I have seen frame rate consistency issues and lower benchmark scores with a higher error rate.

edit: 1435 @ 1231mv VID and 2210 mem Power usage spikes to 280 watts in valley at times. A little worrisome as we're way out of spec. on pci bus.
Edited by greytoad - 8/12/16 at 8:15am
post #474 of 4151
Quote:
Originally Posted by greytoad View Post

I'll be honest I'm not sure what kinda of fault HWINFO is monitoring, but I have seen frame rate consistency issues and lower benchmark scores with a higher error rate.

Make sense probably are calculation errors, the GPU has a compensate mode to fix this error like RAM memory have, and this make a delay in the processing, because need to stop, fix and continue - I will check this if i see any points difference in benchmark of any tangible difference in performance.-
By the way, how much mem freq you live in your OC?
post #475 of 4151
2210 to 2220 2210 is my no worries error free.
post #476 of 4151
@greytoad you are in the right, finish to check on the fly with unigine heaven, with 2195 mhz don't have any errors, with 2200 mhz start to have error little one, 1 every 5 seconds or 3 may be, so decide to be in 2195 mhz of mem freq at stock voltage of mem - Good information about.-
post #477 of 4151
Your probably correct about the HWiNFO. Makes sense it would be counting fault retries on the memory controller. There's no ECC as far as I know on these chips so diving into how the memory controller finds the fault would be interesting.
post #478 of 4151
I find something about the Correction in GDRR5 its like a hash system

The specific CRC function used in GDDR5 can detect 1-bit and 2-bit errors with 100% accuracy, (...) Should an error be found, the GDDR5 controller will request a retransmission of the faulty data burst, and it will keep doing this until the data burst finally goes through correctly

http://www.anandtech.com/show/2841/12

I think HWiNFO see this request retransmissions and show you that, that because when you go more faster the sensitive memory start to have error calculations, and request again the data, that produce a low performance - When you downclock your Mem this error disappear and you start to have again a better performance
post #479 of 4151
IDK, was running 1350/1950 for 2475. I'll rerun everything and recheck my saved results when I get home in about 8 hours.

At least I'm pretty sure it was 2475.
μRyzen
(12 items)
 
Mini Box
(4 items)
 
 
CPUMotherboardGraphicsRAM
Ryzen R5 1400 MSI B350M Gaming Pro Zotac GTX 670 4GB G.SKILL FORTIS Series 8GB (2 x 4GB) 
Hard DriveCoolingOSOS
WD Green 3tb Wraith Stealth Windows 10 Debian 8.7 
MonitorKeyboardPowerMouse
ViewSonic VX-2257-8 Chinese backlit mechanical Kingwin 850w Chinese laser optical 
CPUMotherboardGraphicsRAM
Athlon 5350 Asus AM1I-A EVGA GTX 750 Ti SC 2x4GB DDR 3 1333 
  hide details  
Reply
μRyzen
(12 items)
 
Mini Box
(4 items)
 
 
CPUMotherboardGraphicsRAM
Ryzen R5 1400 MSI B350M Gaming Pro Zotac GTX 670 4GB G.SKILL FORTIS Series 8GB (2 x 4GB) 
Hard DriveCoolingOSOS
WD Green 3tb Wraith Stealth Windows 10 Debian 8.7 
MonitorKeyboardPowerMouse
ViewSonic VX-2257-8 Chinese backlit mechanical Kingwin 850w Chinese laser optical 
CPUMotherboardGraphicsRAM
Athlon 5350 Asus AM1I-A EVGA GTX 750 Ti SC 2x4GB DDR 3 1333 
  hide details  
Reply
post #480 of 4151
In case your interested the 280 watt spike was the highest I've found in valley. After about thirty min of running it, average was more in the 220 to 230 range spikes. No burning smell and 24pin is not hot so I'm good. Thank you for the info on the gddr5. Sounds like the move to compute has increased the gpu's ability to maintain calculation accuracy. If they were just drawing pretty pictures, a few pixels here and there, a few hundred errors... wouldn't matter much.

The actual run was at 1435 and 2210: I couldn't take a screenshot during a full screen run and my mouse was on the average vddc for the run.

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: AMD/ATI
Overclock.net › Forums › Graphics Cards › AMD/ATI › [Official] Polaris Owners Club