Overclock.net › Forums › Graphics Cards › AMD/ATI › Software for R9 285 bios edit
New Posts  All Forums:Forum Nav:

Software for R9 285 bios edit - Page 85

post #841 of 1077
@m42BMW Nice yeah me too but I'm a believer in single gpu setups rather than sucking down extra power. I used to run 3x gtx 470s in sli and ran everything 1080p ultra @ 150fps but power consumption was way up there.

The 380x uses a lot of power honestly but it's okay. It's very powerful for what it is. I run everything ultra 1080p @ way over 60fps & I can't tell an fps difference between 2560x1440 & 1920x1080 as they seem to run the same. Though I really like 3840x2160, and can even run ultra at 4k at 30fps ish usually never less than 30, typically more. Until the ram gets taxed and fps drops. If it had 8gb of ram it would be perfect.

A single rx 480 is like 4x faster than the 380x in reality because of the 14 nanometer process and lower power consumption. It can clock like mad, 1600Mhz or more possibly totally stable. So two of them even at stock speeds has been seen to out run the gtx 1080 with it's what 128 ROP's? The rx 480 is very special, its performance is amazing. Temps are high but with some modding, I'm sure a massive reduction would be in order of -20C at least.
post #842 of 1077
@chris89 I have tested the 1160mhz and I get 8300 graphics score less than stock for my 380. I will compare the bioses with my stock bios in games to see fps diferences, maybe they are performing better in games than in benches.Thanks :thumb:I will test the 180 and 225W bioses.
post #843 of 1077
@mynm

Great bud. Looking forward to the results... throw up a GPUZ, HWinfo screenshot of the results of analysis.

Even know my T7500 has 108 amperes over the 6x 12 volt rails, it has a hard time with high powered GPU's. Idle is low and load droop is significant. At 1230Mhz @ 225 watts @ 168 ampere, the result is still "Green Screen Freeze Ups & Shut Downs". Backing down to 215 watts, 161 ampere it is able to maintain continuous stability. Yet with 215 watts, it rarely sees 1230Mhz and hovers in the 1180 to 1210mhz range. Only when it's hooked up to my Kingwin 1000 watt power supply can it handle anything I throw at it well over 250 watts of tdp. Though these cards have poor internal power rail stability, any chance we can minimize the +12 VRM Input idle & droop? Do we know what PowerPlay Power Limit leakage means?

Could it be within the GPIO parameters? Thanks again for your help and good luck with the BIOS. thumb.gif
Edited by chris89 - 10/25/16 at 4:12pm
post #844 of 1077
@chris89

http://www.anandtech.com/show/1611/3

This seems relevant to the leakage settings. I would assume it is combating or compensating for leakage in some way, which from what I've read leakage gets worse as we add voltage, so perhaps we will need to adjust these settings at some point?
post #845 of 1077
@m42BMW thumb.gif

Thanks for the info M42Bmw. It appears as the nanometer process reduces the resistance rises as the clocks rise because of the very small process. Correct me if I am mistaken.

What AMD PowerTune uses is what's called Low/ High CAC Leakage. Now to be sure of what the CAC stands for, according the article my guess is "Current Adaptive Continuity". Since it must adapt to the leakage as the voltage/ clock rises it must have a low and high limit of continuity. Basically which means the measurement of continuous flow of energy or current. As power and voltage rise among clock rate, the leakage becomes so great it could render a non-functional card. As I have tested this a little, tinkering results is no post (Although I was throwing unrealistic numbers at it just to see the result). It may be worth tinkering with a bit more if results are desirable.

LoadLinePSI as well is a value of 14, have you messed with this? @mynm



I tested LoadLinePSI from 16, to 24 and it didn't do really anything that I could tell. I also test low/high CAC leakage and it appeared the lower you go reduces wattage and temperatures and decreases stability or increased power and became more unstable. So its a waste of time to adjust those settings I would say.

Here's some R6 Siege benchmarks @ 1160mhz just like the one I made for Mynm.

1080p ultra


4k low


4k ultra


rx480 4k ultra

Edited by chris89 - 10/26/16 at 2:04am
post #846 of 1077
I don't have tested Low/ High CAC Leakages, I only have tested to change it to the values that 280 gpus have, but I didn't see any thing. I don't know what they are doing.

LoadLinePSI is changing the variability of the voltage. I have tested it to 1E (111 LL slope 90.3% and 10 = initial programmed offset (default) ) and I can't load firestrike at stock clocks because of the low min voltages.

I see that 380x gets 9200 points at graphic score with 1040/1500: http://www.legitreviews.com/sapphire-radeon-r9-380x-nitro-video-card-review_175481/12 , I get only 8700 with 1040/1500, a ~5.4% less. I get the same 9200 with 1100 or 1140mhz depending of the drivers, so I need 100mhz to match the 380x, so is imposible to my to get your scores. I don't have R6 Siege, I prefer to compare with gta v or firestrike, to see if I get a ~5.4% less than you.
post #847 of 1077
@mynm

Cool dude yeah I noticed not much of any change from LoadLinePSI up to value 24 from 14 and I started at 16 and jumped to 24, minimum change. I have screenshot comparisons. LoadLinePSI results in no change that I can see. Though CACLeakage low/high does affect power amount of delivery, lesser values result in less no change in VRM +12v and only affects the VRM power amount of delivery. A low value of say low 900, high 3000 from 1100/2800 results is less watts and a reduction in temperatures with very poor stability. Higher levels say 1500/3500 results in more heat and more VRM wattage with very poor stability as well. So it's clear 1100/2800 is ideal and default LoadLinePSI of value 14 is ideal as well.

On R6 Siege I pulled 30.1fps average first score(4k ultra) among 31fps ish on the others with huge power delivery and VRM whining @ 1230mhz. At 1160Mhz it uses far less power than 1230mhz and was about 1fps average less with around a 75 watt power reduction.

The fastest single r9 380x score on firestrike with PCIe 3.0 was 9,424 points and 50.39fps test 1, and 44.33fps on test 2. http://www.3dmark.com/fs/9298572

My 380x at stock speeds at PCIe 2.0 is I think around 7,500 I believe. Pulling over 9,000 points is very tough to achieve. Pulling 9,389 points for me was tough as PCIe 3.0 gives more fps per Megahert. I did however manage to pull 50.52fps on test 1, and 44.35 fps on test 2. Though my combined was 17.20 and the guy with 9,424 points recieved 17.15fps on the combined score. The 9,424 points score received a Physics CPU score of 48.42 and my 2x Intel Xeon's pulled just 46.51fps. My CPU's hit 3060Mhz per core turbo boost on all 24 threads, and the 9,424 point score had an i7 6700K @ 4800Mhz. If I popped in some X5675's @ 3333Mhz per thread it would take the lead but I run these X5650's because they are fine for now, cheap and just 95 watts per cpu also very cool and stable, around 50C full speed boosted on all 24 threads.






r6 seige unknown mhz 1080p ultra


r6 seige unknown mhz 1440p ultra

Edited by chris89 - 10/26/16 at 9:31am
post #848 of 1077



Just a squidge to go to surpass your graphics score! Once I get my card benching 1200+ again anyway... Literally gain no performance past 1200 anymore, and can't clock past 1250 anyway. Lower score is my current 1190/1563 game stable. High score is 1280/1563. Both were with monitor apps running and no real optimization measures taken for "max" score.

Should have the new components ready to rock within the next two weeks smile.gif
Edited by m42BMW - 10/26/16 at 1:12pm
post #849 of 1077
@chris89 I have tested your 1160 225w bios and I have freezes and whit the 1160 180W I get an 8802 in firestrike graphics score.

But I have tested CACLeakage low/high to 900/3000 and I can go up to 1160mhz with a max voltage of 1.206V with LLS1 (12.9%), bios 225W, bios 188A, and fan to 75% (too loud for my). Before I get artifacts with more than that voltage.
Warning: Spoiler! (Click to show)

I don't know if it is a coincidence, or if it is some tingh wrong, maybe the score is to low. I have to test it more. I know that fiji gpus have CACLeakage low/high to 0, maybe they are capping some thing, and with 0 is un capped, I will test. The problem is that is impossible to know how to change that values.

Nice scores m42BMW
post #850 of 1077
mynm. I just did Firestrike at 9358 graphics score with 1133/1435. Looks like your card drew 25w more VRM power out than mine despite your card peaked 3C less and almost half of max fan RPM. Looks like you've got it tweaked good for your needs.
Edited by ducegt - 10/26/16 at 3:38pm
Gaming Rig
(14 items)
 
  
CPUMotherboardGraphicsRAM
Intel 7700K ASRock Z270 K6 Gaming Fatality R9 285 TridentZ 3600 15-15-35 
Hard DriveCoolingOSMonitor
Samsung EVO 850 Corsair H115i Windows 10 Nixeus NX-Vue24a 
KeyboardPowerCaseMouse
Corsair K70 Rapidfire  Corsair RM850X Corsair 600Q Logitech G303 
Mouse PadAudio
Steelseries Qck+ Integrated  
  hide details  
Reply
Gaming Rig
(14 items)
 
  
CPUMotherboardGraphicsRAM
Intel 7700K ASRock Z270 K6 Gaming Fatality R9 285 TridentZ 3600 15-15-35 
Hard DriveCoolingOSMonitor
Samsung EVO 850 Corsair H115i Windows 10 Nixeus NX-Vue24a 
KeyboardPowerCaseMouse
Corsair K70 Rapidfire  Corsair RM850X Corsair 600Q Logitech G303 
Mouse PadAudio
Steelseries Qck+ Integrated  
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: AMD/ATI
Overclock.net › Forums › Graphics Cards › AMD/ATI › Software for R9 285 bios edit