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 74

post #731 of 1079
I shared my experience with a temp gun on the vrm while I had winter air from outside blowing on the card... Vrm still got crazy hot despite my card being able to bench 1250 core and drawing more than 200w of power.
Gaming Rig
(14 items)
 
  
CPUMotherboardGraphicsRAM
Intel 7700K ASRock Z270 K6 Gaming Fatality RX VEGA 64 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 RX VEGA 64 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
post #732 of 1079
@M42BMW

The below bios has way too much voltage... 1400mv is voltage limiting at 1175mhz... we need less volts... I even went from 190w to 240w with no relief... anything above 1375mv gets hairy... maybe 1400mv barely just barely but way too high for 1175mhz.. 1175mhz might work at something like 1325mv to 1375mv... So I'm working on it.. we have to find max clock for 1375mv... basically




M42BMW_Chill_Bios_PowerColor_R9_380.zip 100k .zip file

Start with a STOCK bios, maybe change Powerplay power limit to 75% rather than 20% to yield max 225W. Then find your max stable clocks with TRIXX... +100mv on Trixx is 1.325v... +175mv on Trixx is 1.400v... I would probably do 1501MHZ memory which is 192GB/s which is enough for your one lone memory regulator. Then just focus on the core clocks and see if you can get 1250mhz totally stable around 1375 to 1425mv... Remember stock 65288 in DPM7 is 1225mv core... So remember what +XXXmv yields what voltage... So we can make you a permanent BIOS... Try 1563Mhz on STOCK bios by TRIXX... If it can't do it on stock volts it can't do any more with more voltage... So 1501Mhz shows 1500mhz in HWInfo and 192GB/s is good solid memory performance.

We will figure this out man.. For me I can't even get 1200mhz 24/7 stable... Something around 1150 to 1175 is doable though. I really wish 1250mhz was possible but the card puts up a huge fuss at anything above 1200mhz... TDP soars and temperatures soar... Not including the core... The core is cool because the manufacturers think only the core diode is important which is not the case... The voltage regulators get way way hotter than the core all the time so it's like pulling out your hair knowing this while under load... Then the thermal parameters shut down one regulator at a time off/on/off/on so we see stutters and dips because of it.

For instance my 380x has 15 core voltage regulators... Pushing 230 watts means they handle 15 watts each... Compare that to a 15 watt gpu without a heatsink or even with one and man that thing cooks at 15 watts...

These GPU's also are doing more stuff with electricity than the past, they convert high volts to high current internally channelling hundreds of Current Amperes. Amperes generate 10x the heat of low Ampere/ high voltage... They felt this was the best way to generate higher pseudo/virtual/fair land power but it's just a flop. I love AMD and won't give up on them, I just hope they get their engineering in-line because they only need to sort out thermals. The 390X should be extremely fast, yet it only edges some cards and isnt producing the FPS it is truly capable of.

@DuceGT

Yep that's AMD for ya, never addressing the temperatures... They build amazing PCB's, capable of so much like the 290/290x/390/390x...

Yet with those 64 ROPs 512 Bit GDDR5 cards... They are just too hot to show their potential...

So Until we get the VRM temps and the entire PCB under control we will never get any sort of acceptable performance out of these cards...

On my sapphire I use the 17W fujipoly yet the VRM plate that is bonded to the alluminum heatsink is made of some sort of ultra cheap metal that has horrible thermal properties... So I still can't transfer enough of the heat unless the VRM contacts pure copper sharing with the core diode.

It's quite sad.

My theory to resolve AMD's thermal issues is if they coated the entire pcb with a highly conductive thermal epoxy, and cast a one-piece copper heatsink that is cast to contact the entire pcb... Same to the backside, full pcb expoxy to pure copper backplate (The epoxy would be applied, then heatsinks mounted then melt the epoxy and let cool so it's a perfect bond)... That way the cards would last forever and would only need to be air dusted and never ever worry about any of the temperatures. Then the performance would be allowed to soar in accordance to thermodynamics. thumb.gif
Edited by chris89 - 10/10/16 at 6:30pm
post #733 of 1079
Thanks for all your help @chris89. I will be able to put some time back into figuring these voltages/clocks out Wed & Thurs... Just a bit of tinkering tonight and probably tomorrow night after work. The only thing I find troublesome with the stock bios is it has a wide spread of voltage for max core clock making it hard to stabilize the higher clocks, just like battling vdroop and voltage instability when overclocking a CPU. I will see what happens with stock bios and trixx... It would seem that +200mv from trixx should net me the 1.4+vcore needed for 1280/1563. (Previous to reading this thread I had only ever used AB for +v, not realizing Trixx can go +200mv) If I get any results tonight I'll post. Curious to try this "chill" bios too, looks promising if the memory volts get me at least 1525mhz

Also, it would be alot of fun(albeit very costly) to super-mod one of these cards for supreme thermal efficiency and see where it takes us... But by that price point we may as well just hop up to a new RX 480, gtx 1060, or better! I am quite happy that my performance numbers put my 380 close to the Rx 470's, which I was debating on buying before putting the 380 under water which happened because EK held a big sale... Block and backplate shipped for ~$100
Edited by m42BMW - 10/10/16 at 5:18pm
post #734 of 1079
Here is what I was talking about. Sorry I don't have picture proof, but as my memory serves me, the VRM would peak over 110C (according to my temp gun) before crashing in firestrike.


Gaming Rig
(14 items)
 
  
CPUMotherboardGraphicsRAM
Intel 7700K ASRock Z270 K6 Gaming Fatality RX VEGA 64 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 RX VEGA 64 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
post #735 of 1079
Yeah the uber thermal modding can be used for as long as you take care of it and it doesn't break apart...

I bought a giant sheet years ago, and still today still using all the material I bought...

So that's the up side, not like you spend once on one card and now your out of money...

It's buy this for this card and as many cards as you buy for as long as you can take care of the material.

The material is delicate as we both know.

I made this chart... Reflect on it for the R9 380 series and your ambitions...


Edited by chris89 - 10/10/16 at 6:46pm
post #736 of 1079
@duecegt 110 is smokin hot! Before I modified my motherboard VRM heatsink with an aluminum bar backplate and added a fan directly above the heatsink ( you can see it in my system pic a few pages back) they would peak over 100c when I started overclocking my fx9730... Which is why I am very happy the backplate on my 380 is designed to put pressure on the VRMs holding them tight and flat against the water block just like my modified motherboard VRMs. Hoping they aren't getting past 80 wink.gif
post #737 of 1079
@chris89 Things I learned tonight: stock bios core volts peaked at 1.150 (matched the VID) and memory volts peaked at 1.119v. I was able to reach stable 1060/1563 error free at this voltage. I needed 1.175v (+25mv via Trixx, peaked 1.188 via GPUZ/HWMonitor) for 1100/1563, 1.2v (+50mv, peaked 1.212) for 1120/1563, and +100mv (peak 1.244v)easily stabilized 1160/1563, but then I needed more than Trixx could offer at +200mv for 1200/1563.
post #738 of 1079
@M42BMW

Cool man, you will see the true 1.225v base by stock if you test for instance Sky Diver at 5120x2880... If you can pass Sky Diver at 5k then if you feel okay with it, you can probably call it a 24/7 stable clock.

Though I find that even the base voltage could hit 200 watts with 1050mhz... So your bios needs a higher watt tdp limit, to go from 150W to 240 watts, beyond peak allowable to delimit the big end on the card.

If you set the TDP to 240 watts, you find many much lower clocks actually yield higher fps. Something like 1125mhz on the lowest achievable voltage could utilize the entire 240 watt spectrum without throttle.

See when we push past this point, gains can only be averaged higher because it's constantly bouncing off the power limit. So for actual gaming a lesser clock of say 1150mhz is ideal and find the absolute minimum allowable voltage to ride that clock out at say 1275mv core.

You can't make consistent fps gains past a certain point... I ran the benchmark and find that once you pass the limit, it yields less fps.

I could do 1200mhz with a 0 offset to the core base and could do it on 1.350v, which left the vddci core base offset up there in the 1.350v range consistently.

I feel it's better to do say 1275, maintaining CD FF which is -50mv from core base. So offset would be 1225mv. That way you can consistently utilize the whole 240 watt sprectrum and give consistent fps and stability.

I'll make you a BIOS soon for 1275mv bro.

later biggrin.gif
Edited by chris89 - 10/10/16 at 10:28pm
post #739 of 1079
Quote:
Originally Posted by chris89 View Post

@Mynm

Sorry to hear your still having freezing. I have freezing if any one of the many voltages are too low. It can be stable at first, but later by just scrolling a window down or resizing a video window it will freeze. Always solved with 25 extra millivolts, though you would need to know which dpm voltage is too low that is... Could be any one of the dpm state 0 to 7...
Warning: Spoiler! (Click to show)
The only difference from Lut Entry Voltage 1 from 1100mv to 1000mv is pretty much no difference and it didn't affect the Core voltage or VDDCI... The card just runs about 1 to 2C cooler... I also like to run the absolute minimum voltage on the offset for dpm 7 among all other dpm states. I find 1125mv on the offset yields lower temperature, less power consumption and ever so slightly more stability. Though with all the reductions, including the usVddci, and usMvdd to both 1000mv yielded another 1 to 2C cooler so totaling about 4 to 5C cooler at all times and no reduction in stability, it seems more stable actually. I also like that even with the MCLK_Dependency usVddci and usMvdd at 1000mv can still hold 1563Mhz memory just fine.

What are your voltages and clocks at the moment? On stock BIOS? To know for sure I would set the clocks far less on stock bios voltage. Basically set DPM 1 to 400mhz rather than 473mhz, then 500mhz, 600mhz, 700mhz, 800mhz, 900mhz, 1000mhz. Since 1000mhz is extremely stable at stock 65288 which is 1.225v/ 1.175v. Does that makes sense? Would you like for me to make you an ultra stable bios with a load of voltage overhead for each clock to determine what it is that is causing the issue?

As far as the screenshots go, I'm using all the dpm states. Also the high temperature there in the screenshot is seconds after heavy 3d load testing and temperatures were still coming down. I have successfully reduced decode from 20 Watts on the core with stock bios, to no more than 10 Watts... While maintaining all the states.

I modded my card a bit more. I unsoldered the DVI Ports. Which resulted in a 30 to 40% fan speed reduction to 10 to 20% reduction in core temperature. Since most of the heat was dumping inside the case because of the exhaust vent being blocked by over 70% with the DVI ports and the mounting plate design. I have a modest fan profile and now no matter how high I clock it or volt it, I see no more than 55% fan speed at 75C. Before would dump loads of heat into the case, minimal air out the exhaust and the fan would go to near 100% to hit upwards of 80C.



I'm testing now to see if it is a system RAM problem, testing with two RAM modules istead of four. I had passed 10hs of memtest with the four ram modules and I didn't see any error, but maybe it's a mobo incompatibility or a incompatibility between RAM modules. Or could be a wddm 2.1 problem, or it could be unmasking the RAM problem.
Quote:
Originally Posted by ducegt View Post

Here is what I was talking about. Sorry I don't have picture proof, but as my memory serves me, the VRM would peak over 110C (according to my temp gun) before crashing in firestrike.
Warning: Spoiler! (Click to show)


You can test to see if changing VRM switching frequency to 300kHz is lowering VRM temps.
It is at F7, default value is A and it is 400khz, 5 is 300khz.
For 380 is adding /wi6,24,f7,05 to an AB shortcut for 5, /wi6,24,f7,0a for A, /ri6,24,f7 to read the value:
Warning: Spoiler! (Click to show)
Quote:
Originally Posted by mynm View Post

Ok thanks for the explanation.

But If a 1 nano second droop is a problema, a -100mv drop seems to be a problem as well biggrin.gif.

I don't know if LLC can solve it, the problem is where is it on i2c dump.
Warning: Spoiler! (Click to show)

[ ATI I2C Device GPU #1 / B06-D24 ]

0000 FF 80 17 7F FF FF FF FF FF FF FF FF FF FF FF FF ...............
0010 00 FF FF FF FF FF FF FF FF B0 FF FF FF FF FF FF ................
0020 22 00 FF FF 00 18 A8 FF FF FF FF FF FF FF FF FF "...............
0030 FF FF FF FF FF FF FF FF 01 00 FF FF FF FF FF FF ................
0040 FF FF FF FF FF FF FF FF FF FF 64 FF FF FF FF 55 ..........d....U
0050 FF 46 00 64 FF 10 FF FF FF FF FF FF FF FF FF FF .F.d............
0060 FF FF FF FF FF FF FF FF 2C FF FF FF FF FF FF FF ........,.......
0070 FF FF FF FF FF FF FF FF 00 00 00 00 00 00 FF FF ................
0080 FF FF FF FF FF FF FF FF E9 FF FF 3F F0 00 FF FF ...........?....
0090 FF FF FF FF FF FF 64 FF FF 1A 22 03 FF FF FF FF ......d...".....
00A0 FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF ................
00B0 FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF ................
00C0 FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF ................
00D0 00 01 00 03 50 02 00 FF FF FF 01 FF FF FF FF FF ....P...........
00E0 FF 00 00 02 03 03 00 3D FF FF FF FF FF FF FF FF .......=........
00F0 FF FF FF 68 FF FF FF 0A BC 9C FF 07 FF 0F FF FF ...h............

[ ATI I2C Device GPU #1 / B06-D24 ]

0000 FFFF 2F80 7E17 FF7F FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF
0010 6F00 FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF 4CB0 FFFF FFFF FFFF FFFF FFFF FFFF
0020 6022 0000 FFFF FFFF 0000 0018 00A8 FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF
0030 FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF 0001 0000 FFFF FFFF FFFF FFFF FFFF FFFF
0040 FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF 0064 FFFF FFFF FFFF FFFF 0055
0050 FFFF 0046 0000 0064 FFFF 0010 FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF
0060 FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF 012C FFFF FFFF FFFF FFFF FFFF FFFF FFFF
0070 FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FB00 0000 2D00 4600 5000 3B00 FFFF FFFF
0080 FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF D2E9 FFFF FFFF 013F A2F0 0000 FFFF FFFF
0090 FFFF FFFF FFFF FFFF FFFF FFFF A364 FFFF FFFF 001A 1022 8103 FFFF FFFF FFFF FFFF
00A0 FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF
00B0 FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF
00C0 FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF
00D0 E200 8E01 3400 5603 BA50 2C02 9F00 FFFF FFFF FFFF 6201 FFFF FFFF FFFF FFFF FFFF
00E0 FFFF 6800 D500 B002 A103 CA03 7E00 003D FFFF FFFF FFFF FFFF FFFF FFFF FFFF FFFF
00F0 FFFF FFFF FFFF 0368 FFFF FFFF FFFF 810A F5BC F29C FFFF 5807 FFFF 1D0F FFFF FFFF

I don't see any info about that, I only see this now thanks to buildzoid: http://forum.hwbot.org/showthread.php?t=157437 , I will see if I can find some thing on NCP81022 datasheet.


Edited: Is it F7 OSC Freq, Vrm switching frequency?, is A=10 and acordin to page 26 table is 400khz.
Quote:
Originally Posted by gupsterg View Post

Yes smile.gif , that PDF shared by buildzoid is biblical biggrin.gif . Try setting it to 05h as that should get you 300kHz, which will lower VRM temps but up efficiency and not effect OC ability. The Stilt sets Hawaii ROMs he makes to 300kHz (default is 500kHz).
Sorry not had or got that card, perhaps another owner may share info with you.
post #740 of 1079
@M42BMW

I found I get shutdowns with TDP of 240 watts, and voltage of 1350mv... So I reduced tdp to 225 watts and it does fine. So just so you know going beyond 1150mhz is not possible because the card simply cannot suck down any more watts to yield gains... Max I ever saw was 230 watts on the core and believe you me the regulators were glowing under fujipoly 17w/m k extreme pads...

@DuceGT

Yeah 110C is hot but there are 3 VRM locations... One nearest display connections and exhaust vent, and two locations at the back of the card for Memory And Auxiliary power.

Memory and Aux will soar sky high way up in that range unless you undervolt the MCLK vddci and mvdd... Want me to make you a cooler memory regulator bios to test?

I find I do fine on 1000mv mvdd and vddci at 1563mhz. For 1150mhz core, needs 1350mv dot/ artifact free... At 1275 to 1325mv its stable but dots appear on textures and characters. 1150mhz at 1350mv has the potential to utilize the full 225 watt tdp pretty well throttle free... Push past 1150mhz and throttle all day long. It can do 1157Mhz which I like a rounded theoretical throughput number of 37 giga pixels and 148 giga texels... R9 380 would be at 1157mhz .. 37 giga pixels and 129 giga texels.

I find I get amazing performance undervolted across the board... It's quite an undertaking when changing like 20 values in BIOS hex to achieve reduced temps on the Memory And Aux Regulators.

@Mynm

Hmm I too had a memory module issue! It was plaguing my rig for some time... I ran Dell Diagnostic and sure enough 5 minutes in DIMM #1 had a correctable memory error as I have Error Correcting Memory. It didn't get solved so I reduced my ram from 48 GB to 24 GB... Plus I want to run 8GB Modules rather than a whole pile of 4GB Modules... x6 8GB Modules is ideal... Reduce power consumption on less higher capacity modules. Anyway that solved this problem I had where the whole rig would freeze for no reason randomly or blue screen with Memory Management as the cause or some other thing like Thread Stuck or some such.

How much memory do you have? What's your motherboard model/ cpu/ cooler used for cpu/ thermal paste used? If it's an AMD system I always find a more stable system with Hypertransport to Northbridge ratio of 1:1 of 2000Mhz:2000Mhz... Anything beyond 2000:2000 yields far less stability. Make sure the memory CAS latency is all as the module intended so look that up and make sure the BIOS is pushing 1.500v for ddr3 and ddr4 what 1.350v?!? into the memory modules as well. What's the speed of the modules? If you mis-match the CAS Latency and modules will have been reported to fail yet the CAS is just out of touch with the module as far as the BIOS settings go...

Cool I look forward to hearing more about your system specs etc...
Edited by chris89 - 10/11/16 at 8:01am
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