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 73

post #721 of 1079

Wish I had that Samsung ram! Is there a way to use the TongaBiosEditor for the VID increase like I did that doesn't directly effect the vddci(which gather is the RAM voltage?) Really I'm probably OK with 1525 memory clock which I can get with zero errors on stock voltages. I'll test your bios in the AM and see if I can't get at least 1563, or better, with no errors and reach for that 1300+ mhz core. Also appreciate the 12v rail tip, dont think I've ever noticed it drop below 11.85v but I'll definitelty keep an eye on it
post #722 of 1079
Bios did not fly. Memory errors and flashes during first 5 seconds of heaven extreme. Was able to down clock the memory but still had artifacting even though I saw core voltage around 1.38-1.4, which was enough voltage for me to run 1280/1525 benches. Must be something else off? Reapplied my modded bios with a slight decrease in VID compensated with trixx I could get back to 1280 benches, but was getting crashes at 1300 and 1.38+vid
post #723 of 1079

Try just turning down the ram speed to 1600 ... and turning down the 1450 value in dpm 7 to 1400 ... and try 1200 on the core rather than 1250...

I'm on Elpida and with 1175 mvdd volts I can do 1680Mhz .. which is 215GB/s and run through Time Spy of all benchmarks without error...

It's the Fujipoly 17 watt per meter kelvin thermal conductive thermal pads I use on the memory and voltage regulators to allow the very high memory clock...

Anything less than that wouldn't work. Whats your thermal material you use? Whats the thermal conductivity rating? It's in Watt Per Meter Kelvin
Edited by chris89 - 10/9/16 at 1:05pm
post #724 of 1079
I was still getting errors at 1563 memory till I switched back to my bios version. I will open yours in the bios editor and make the suggested changes and give it another shot tonight. I am running the 6.5W/mK from Arctic, which is a big step up from the 3w/mk supplied by EK for the water block. I'm sure the 12+ W/mK would benefit me a bit, but I can't help but think temps are not my major gremlin, rather volts and clocks playing well together. Perhaps just a slight bump to ram voltage and then roughly 1.4v core would net the results we want to safely pass 1280/1563?
post #725 of 1079
Try this one... memory voltage and vddci at 1125mv each... 1400mv core base with 1185mv offset which will aid more for the high volts and clocks.

M42BMW-ModdedBIOS-R9380-PowerColor2.zip 100k .zip file

Yeah I find the 17w stuff holds better for the temps you cant monitor by software like the regulators for the ram and core... Ram regulators are at the back of the card near and around the power connectors.

So I'll make another bios for ya, no problem. Be a moment.

Cool... Also I'm surprised 1563 isnt working... probably getting too hot... this fujipoly 17w makes it so any number is nearly possible up to 1680mhz for me stable through benchmarks...

M42BMW TDP Offset locations
a655 - tdp - 9600 - 150w
a657 - configurable - 9600 - 150w
a659 - usTDC - 7d00 - 125a
a65b - usbatterylimit - 9600 - 150w
a65f - usSmallLimit - 9600 - 150w
a663 - usMaximumLimit - 9600 - 150w

a57d - usVddci - 4c04 - 1100
a581 - usMvdd - e803 - 1000

a549 - usVddcOffset - cdff - -50mv
Edited by chris89 - 10/9/16 at 4:13pm
post #726 of 1079
Nice, appreciate the effort and help man! I will give this one a go tonight after work. It is possible my RAM is getting too warm, and no way to monitor unfortunately. Worst case I'm still happy with 1525-1600 memory speeds. I've attached an image of the block I'm using, there is also a full cover backplate with contact pressure at the VRMs and GPU. If I'm feeling froggy someday in the future I'll redo the thermal pads with the high end stuff, but being as I just got this loop assembled and am getting fantastic core temps I'm not in a hurry to disassemble.

CREATOR: gd-jpeg v1.0 (using IJG JPEG v62), quality = 90
post #727 of 1079
I'm actually trying to trade my 380X for an Nvidia 760/770/970/980 haha Because even the GTX 760 waxes the 380X which clocked up which it can do far easier.

760 @ 1333mhz is 42.656 Giga Pixel's per second... 112 TMUs on the 760 for 1333 x 112 = 149.296 Giga Texel's per second.
380x @ 1235mhz is 39.520 Giga Pixel's per second... 128 TMUs on the 380x for 1235 x 128 = 158.080 Giga Texel's per second

Cool dude. See back by those two black boxes near the 6-pin power connectors? 2 smaller chips right by the black block nearest to the 6 pins is the AUX regulators.. Uncooled... The 1 regulator by the black block furthest from the 6pin power connectors is the 1 and only regulator for memory which is the cause of incapability to achieve higher memory clocks... Though the bios at 1125mv might be safe and perfectly fine to clock the ram up... I would add a heatsink to those regulators near those black blocks.

Yeah I would also buy a 2 to 2.5mm thick large sheet to cover the entire back side of the pcb to transfer all of the heat from every square millimeter of the pcb to the back plate to be cooled passively by fans blowing over the card.

Nice block man!

By the way, at max clocks and volts I'm barely breaking 60C on the core on air...

Its when you see that 60C, always figure no less than 20 to 35C greater temps on the regulators for the Core, Ram, and Auxillary power.

Feel the card under load safely by touch, because 40C by touch is cool but if you feel the pcb and it feels like it could burn you it's clear the temperature of components unmonitorable are running much warmer than the core. The card has temperature regulation where it shuts down 1 or more regulator at a time to keep them in a temperature range that will prevent damage.
Edited by chris89 - 10/9/16 at 4:55pm
post #728 of 1079
I wondered if those regulators were going to be an issue without fans on the card. I do have a 140mm side vent fan blowing directly over the top of the card, so it should be doing an OK job pulling heat from those exposed components. I'll look into some small heatsinks for those specific areas, don't think I have any spare laying around currently. Off work in an hour, should be free for GPU tinkering shortly after that. Will be pretty happy with 1563-1600 memory clocks and 1280-1320 core clocks for benchmarks. Will probably detune and find a stable gaming clock around 1220-1260/1550 if all goes well! Thanks again. Also, 12v rail never dropped below 11.89v during this mornings test runs, so that was good to see.
post #729 of 1079
Great job man.

The biggest problem is the lacking backplate thickness... It's also painted and aluminum which cuts thermal conductivity significantly. I would use a nice heavy thick piece of pure beautiful copper. Drill out the screw holes, buy new screws and use 4.1W/m K thermal material and it can absorb hundreds of watts of thermal energy from the back side of the pcb.

If you cut this piece to size(local machine shop) it's a safe thickness for weight and how many watts it can absorb is within the thickness and mass. Copper in it's pure form has great thermal conductivity of way up there in the hundreds of hundreds of watts per meter kelkin.


Try MC21, and cut one in half with a big pair of tin snips for the one lone memory module regulator.


This should hold the core regulators and memory modules... make sure your gap will do well with 0.5mm pads since it gives very low thermal resistance.


I use a lower thermal conductive pad on the backside of my 380x pcb to cover the entire back which gives significant reduction in temperatures of everything on the pcb. If you have the cash and want 4x lower temps than the cheap option here's this massive thermal pad to be used for the backplate among many other things.


A little cheaper stuff, just 1.1 & 1.6W/ meter kelvin...


My dream was to achieve stable 24/7 1250mhz core, maybe you could push for the same as I can't do it. This card has hot power components so big clocks aren't possible unless every inch of the pcb is cooled appropriately, then anythings possible on amd. On nvidia the power components are better in control of achieving very stable high clocks.

If you can do 1250mhz, that's 40 giga pixel's per second... The gtx 970 has 56 ROP's and call it 950mhz core clock, that's 53.2 Giga Pixel's which is 53.2 billion pixel capability.

Run the r9 380 as 1250, that's only 13.2 billion shy which is not bad for a card which had been more obtainable plus I'm not much for 3.5GB limit... gtx 980 had the 64 ROP's.. 1000mhz x 64 ROP = 64 billion...

We might dial back on the memory volts and take it easy to see what it would take to get stable 200 Giga Bytes per second from the 256 bit gddr5...

Stock is...

Maybe we should take it easy on the one regulator...
1025mv or 1050mv or 1075mv or 1100mv... To see what is possible.. I'm willing to get you there brother.

Later bro biggrin.gif
Edited by chris89 - 10/9/16 at 7:53pm
post #730 of 1079
Still no real luck with your bios. I'm thinking I need close to 1.2v @ the ram to be error free over 1550mhz, which I'm ok with that number... Just not higher, which with my bios editing skills via TongaBiosEditor I struggle to attain while getting the 1.38+ core volts needed for 1240+mhz core. It was also still getting rainbow starbursts/crashes with your bios, which seems weird because there was plenty of core volts from what I could tell. Similar core volts via TongaBiosEditor mod on my stock bios and +mv via Trixx would net me error and artifact free valley/heaven benches up to 1280/1563. I just didn't like seeing my ram up around 1.25v still. I'll experiment more tomorrow night I think, trying to better understand changes on the volt tables in TongaBiosEditor. I'm also thinking it might be a good idea to use what I've learned and set myself up with something around 1200/1563 @ ~1.3v/1.175-1.2v as gaming for now and save the big benches for when I get everything tidy in the thermal pads and sinks department.

Edit- To be honest there is alot of room at the core for temps, its not broke 50c. I can't speak for the regulators, but I'll try and get the temp gun on it next time I have things running hot. I saw 1.45v today peak when attempting 1300mhz, but it would fail the benches. I think I was triggering a temp or amp or volt fail-safe somewhere.

Edit 2- Ran another 1280 bench this morning with my modified bios, set the vcore to peak around 1.275 via bios and then added .150v via trixx leaving me peaking just over 1.4v core and not over 1.225 ram volts. This allowed 1563 ram clock and 1280 core clock and a successful firestrike pass with no memory errors. Looking like I will be able to set a 1.250v core bios netting a "safe-ish" 1.2v @ ram and allowing up to +0.2v added via Trixx for the 1280+ benches. If I knew how to hack Trixx or AB to add more than +0.1(AB)/+0.2(Trixx) core volts, I'd do it. Still have yet to break 50c core even running Valley for 20-30min at 1.35-1.4+v core. Thinking I'll need better than 1.45v core for 1300+ core mhz, or a very steady 1.425-1.45v with no dips below 1.4v. It's also looking like a stable 1250/1563 for 24/7 gaming is plausible, so long as my radiator can keep up with my CPU and GPU under long gaming sessions. Lucky for me, a radiator upgrade is on my to-do list either way.
Edited by m42BMW - 10/10/16 at 11:39am
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