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 94

post #931 of 1079
I thought I noticed this yesterday and today I can confirm that after making some changes to ucCKSVOffsetandDisable, the checksum wasn't always changed. Might not mean anything, but never noticed this after having made many other changes.
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 #932 of 1079
Quote:
Originally Posted by ducegt View Post

I thought I noticed this yesterday and today I can confirm that after making some changes to ucCKSVOffsetandDisable, the checksum wasn't always changed. Might not mean anything, but never noticed this after having made many other changes.

Yes, I have probles with the checksum when changing ucCKSVOffsetandDisable in hex editor and after that changing clocks in TongaBiosReader and save. Is better to change ucCKSVOffsetandDisable in hex, save with TongaBiosReader, and then change clocks and save.
Quote:
Originally Posted by AnggaSP View Post

IIRC that's the gui version of atiflash, isn't it?
I think it should be the same just a different program and cmd order but i might be wrong.

Ok thanks.

Edited:

OK now I see this:

"Benchmark tessellation load modified by AMD Catalyst driver, result invalid. Check your video driver settings".

I had changed tessellation to x2, with this warning I see in the first 1100 + ucCKSVOffsetandDisable.

I was impressed with the score so I didn't see the warning biggrin.gif.

Now I have it to amd optimized. But usually I have it at use aplication settings.

So maybe thas is causing the performance increase. I have to test.

Edited Yes it is tessellation. Sorry. doh.gif

I had been testing tessellation on sunday with tomb raider and I forgot to change it back to amd optimized or aplication settings.

Maybe Tradition was using this to have those high scores. It could be a new warning, not shown before.

Edited Nope the bug is solved since SystemInfo 4.2 in September 4, 2013 http://www.futuremark.com/support/troubleshooting#validation.anomaly.DM11_ATI_TESSELLATION_SETTINGS

But CCC drivers had profiles, and config could be changed with keyboard shortcuts.

I don't know how Tradition did it, the score is valid: http://www.3dmark.com/fs/5608366

Sorry to Tradition if is no lying.

And sorry again for making you losing time.
Edited by mynm - 11/16/16 at 9:05am
post #933 of 1079
@mynm

I've done that before and then shortly after realized the issue... I used to turn the tessellation to max 8x as it helped with Fallout4 when AMD R9 cards and Fallout were not getting along very well. On a side note I've taken the dive into using AtomBioReader with the HexEditor for bios editing and I've come up with a pretty solid performing 1150/1550 setup that seems to be long term game stable. I had a 1150mhz setup that scored 9900 firestrike graphics (highest ever for me at that clock) but it would crash after 20-30mins of gaming. Changing a few voltages/offsets I came up with the one I'll be testing stability on today. It scores a bit lower at the same clocks for some reason, but it seemed to not crash in game.

I'll add some screens and upload the bios for comparison after I confirm it being at least 1hr game stable.
post #934 of 1079
Quote:
Originally Posted by m42BMW View Post

@mynm

I've done that before and then shortly after realized the issue... I used to turn the tessellation to max 8x as it helped with Fallout4 when AMD R9 cards and Fallout were not getting along very well. On a side note I've taken the dive into using AtomBioReader with the HexEditor for bios editing and I've come up with a pretty solid performing 1150/1550 setup that seems to be long term game stable. I had a 1150mhz setup that scored 9900 firestrike graphics (highest ever for me at that clock) but it would crash after 20-30mins of gaming. Changing a few voltages/offsets I came up with the one I'll be testing stability on today. It scores a bit lower at the same clocks for some reason, but it seemed to not crash in game.

I'll add some screens and upload the bios for comparison after I confirm it being at least 1hr game stable.

Very good score thumb.gif

Now I'm testing a bios using advanced fan mode to do a boost OC via a max temp and a max fans rpms. I have tested a Edited target tempetature of 80º and a max temp (Tjmax) of 80º with max rpms of 1824rpm. I have tested it with 1200mhz and 1.337v for DPM7, 225W and 188A. I'm using as well ucCKSVOffsetandDisables to 80.

And I get a 9822 score but with a max temp of 80º and 1824rpm, so it is quiet. I have to test it in games to see if fps are increased like in the bench and if it is stable.

The problem is that I see down clocks to 980mhz in games but it is more or less always bettwen 1125 1105, but with peaks to 1140, 1170, 1200.

And the main problem is that at the beginning of games I have mhz drops to 400, because fan takes so log to go to 1824rpms. I have to test if it is solved increasing fan sensitivity.

These are the fan and TPDs codes I'm using, only sapphire is using advanced fan mode so you will se 00s in some for the values:

01 64 00 E4 12 20 07 D8 53 01 00 50 19 00 00 02 96 00 E1 00 BC 00 E1 00 E1 00 4C 04 F0 0A E1 00 50

It is explained how to mod that values in fiji bios mod gide.

Edited: It is not working good. I have clock drops to 980mhz. I think that them are caused by advanced fan mode, because if fan is not cooling the max temp setted, it is droping mhz but in a heavy way.

And the main problem of this fan mode is that target temperature it is not a target temperature, it is the temperature when fan starts spinning. So I have problems with this because my bios Tmax was 100º.
So if I set a target temperature of 80º, fans dosen't start spinning up to 80º, so temperature can go up to 90º when they starts cooling good doh.gif.

So I have tested with a target tempetature of 55º and a max temp (Tjmax) of 82º with max rpms to 1824rpm. + fan sensitivity to the value stted by gupsterg.

01 64 00 3A 2F 20 07 D8 53 01 00 37 19 00 00 02 96 00 E1 00 BC 00 E1 00 E1 00 4C 04 F0 0A E1 00 52

But I have clock drops to 980mhz and it isn't stable with GTAV, with errors messages . I see that this game haves problems with it, but I don't have these problem when I'm using Afterburner to OC, I only have these problem when doing an OC vis bios.

So I think that driver maybe are blocking the OCs in some way. I think that the only way to OC without problems is via software. And the only way of changing per DPM clocks and voltages is using a tool like AMD Wattman or WattTool. Because it is is using some kind of sofware OC. I see that I have the same results for 1140mhz via AB, than with a vios OC of 1100mhz, but the AB OC is stable and the bios OC it isn't, even using the same voltages and clocks like AB.

I will test more but I think that this is impossible.

I think that only Waatman could save us biggrin.gif :
Warning: Spoiler! (Click to show)

Edited by mynm - 11/20/16 at 10:40am
post #935 of 1079
I am on water, so no fan issues for me. I am curious if anyone has tried editing the voltage table 2 without setting specific dpm voltages? Ie. Stock dpm7 table goes 1306, 1162, 1068.

My old R9 380 had a stock load VID of 1.115v, which is the average of 1162/1068 (med and low CAC from table2) my new card is 1.188v at load, which is the average of 1306/1162 (CAC high and med). I will test tonight, but I think rebuilding table with new numbers for the averages might be interesting? I was thinking of trying 1325, 1200, 1150 for a 1.238v vid.

Edit:

I wonder if it's always using High or Low CAC (depending on load/application) and using the Medium to average off of? I just know punching in DPM7 as 1225 does not always equal 1.225v at load, perhaps using the table differently will allow for different (more consistent) results?

This idea stemmed from the 1.115 and 1.188 VID on my two cards. 1.115v and 1.188v are not found anywhere, and we know its using the table to get that VID. I think now maybe we know how it is determining that VID? High CAC or Low CAC voltage averaged with the Med CAC.


Edited by m42BMW - 11/18/16 at 3:55pm
post #936 of 1079
Quote:
Originally Posted by m42BMW View Post

I am on water, so no fan issues for me. I am curious if anyone has tried editing the voltage table 2 without setting specific dpm voltages? Ie. Stock dpm7 table goes 1306, 1162, 1068.

My old R9 380 had a stock load VID of 1.115v, which is the average of 1162/1068 (med and low CAC from table2) my new card is 1.188v at load, which is the average of 1306/1162 (CAC high and med). I will test tonight, but I think rebuilding table with new numbers for the averages might be interesting? I was thinking of trying 1325, 1200, 1150 for a 1.238v vid.

Edit:

I wonder if it's always using High or Low CAC (depending on load/application) and using the Medium to average off of? I just know punching in DPM7 as 1225 does not always equal 1.225v at load, perhaps using the table differently will allow for different (more consistent) results?

This idea stemmed from the 1.115 and 1.188 VID on my two cards. 1.115v and 1.188v are not found anywhere, and we know its using the table to get that VID. I think now maybe we know how it is determining that VID? High CAC or Low CAC voltage averaged with the Med CAC.


Acordig to @gupsterg http://www.overclock.net/t/1563409/software-for-r9-285-bios-edit/360#post_24991415 in fiji gpus changing CAS you can go up to 1.250V. But I have tested and it isn't working like this on tonga. I remenber doing test with CACs and getting a 1.25v max voltage peak voltage in hwinfo but I was reducing voltages, and I couldn't replicate the sema result againg.

Acording to @chris89 is like this http://www.overclock.net/t/1563409/lightbox/post/25524450/id/2873813

Acording to @Ansau is like this http://www.overclock.net/t/1563409/lightbox/post/24804612/id/2686758

I'm going to test some chris89 bioses with some changes to see if them are stables, he have done a very hard work changing all the voltages and clocks in his bioses thumb.gif, but in my tests I have low scores in firestrike with high voltages and high fans speeds with them.

If you want a lower voltage with the same clock but with the same power consumption and temps, you could try to change Load Line Slope in VoltegeObjetInfo Table: http://www.overclock.net/t/1563409/software-for-r9-285-bios-edit/820#post_25602565 , acording to my tests: http://www.overclock.net/t/1563409/software-for-r9-285-bios-edit/810#post_25600160 , http://www.overclock.net/t/1563409/software-for-r9-285-bios-edit/810#post_25600764 , http://www.overclock.net/t/1563409/software-for-r9-285-bios-edit/820#post_25602565 , it is reducing the variability of VDDC.
You can see there two values, the first acording to my tests is for VDDCI, the second is for VDDC.
I see here: http://www.overclock.net/t/1563409/software-for-r9-285-bios-edit/900#post_25645951 that VDDCI is more stable than VDDC, so I think that you only have to change the second value.

I think that LLS to 1 for VDDC is OK, and you have watercooling, so you could go to higher clocks with less voltage, with out temps problems. Don't use it in a OCed bios, I think is better testing it in a stock bios and see what voltage you need for stock clocks, because for a low voltage you could have a high power consumption.

Edited: I have tested this chris89's bios: Mynm1160mhz1322mv225w168a14percentvoltage.zip 42k .zip file

Changing on it these values: 0 64 00 E4 12 98 08 D8 53 01 00 4C 19 00 00 02 78 00 E1 00 A8 00 E1 00 E1 00 4C 04 F0 0A E1 00 64, to this: 01 64 00 3A 2F 20 07 D8 53 01 00 37 19 00 00 02 96 00 E1 00 BC 00 E1 00 E1 00 4C 04 F0 0A E1 00 52. So with the advanced fan mode I get 81º with fans to 1840rpm:

Mynm1160mhz1322mv225w168a14percentvoltageWITHFAN1.zip 42k .zip file

Now gtaV is stable thumb.gif, but it is a 1160mhz bios and I get 1040mhz in games with droops to 700 or 900mhz. This might be caused by the high voltages, so I'm going to test if lowering voltages with an offset I could get more than 1040mhz.

I have to test aswell to decrease TDP from 225w to 180w, like is doing chris89 for capping voltages. With this maybe I will don't have core clock drops to 700mhz. This is more or less what nano is doing. I will compare nano and fury bios to see differences in TDPs and fan profiles.

Edited the bios is stable with a -15mv~-18.75mv AB offset, I have 1040mhz more or less stables with drops to 1030mhz, and peaks to 1125.
With a -20% power limit the mhz peaks are more stables to 1080 1100, but it's in 1040mhz as well.

Edited With -15mv~-18.75mv trixx offset, -10PL and fans to 70% (to loud for my) it is not stable, and I see core clock drops to 1040 and 980.

Edited: Seems that Advanced fan mode is decreasing clock to the next lower DPM if the clock can't hold the max temp, so that it is going to 1040mhz. If I change DPM4 1040 to 1105 it is at 1105, but I see drops to lower clocks.

I'm going to test to set voltages and clock like EVV and AB are changing them. They have a max clock of 1030mhz, a max vddc of 1.225v, and a max vddci of 1.175. So voltage step are 6.25, and mhz steps are 5. VDDCI is not increased by the voltage offset, so will let it to no more than 1.175.
So I will increase voltages and clocks in these steps to see if them are stables and then use a voltage offset to see if I could lower them.

Edited: This is impossible, voltages ar going to high like this... I will set 1.225 for clocks higher than 1030 and the I will add an offset, and use AB clocks for the lower DPMs. Or better to use 1140 stable voltages for 1100.

Edited I have tested a bios with AB 1100 clocks and AB 1140 voltages a little bit adjusted and it seems to be stable in gtav with + half an hour test. And I have the same score and temps than with an 1140 AB OC. So I need +45mv to get the same clocks than with AB.

308A050011001250voltageslike1140225W188AFantype0.zip 42k .zip file
Warning: Spoiler! (Click to show)

I'm going to see if I could lower voltages.

Edited: seems to be stable with a 1.100 vddci instead of 1.175.
Edited by mynm - 11/20/16 at 2:58pm
post #937 of 1079
@mynm

Impressive work! I may have made a small breakthrough on my new card, actually benched stable with scaled results up to 1220mhz. I havent kept up on screenshots, but I can say I'm adjusting TDP to 225, TDC to 190, LuT to 1150, vddci offset of -100, mvdd at 1050, dpm7 1300(good for 1180/1563mhz). This lets me bench to 1220 @ ~ 1.375v using AB for +mv, working on more. I'll be working more on it Wed this week I think.

@chris89

Any luck with that copper project?
post #938 of 1079
Quote:
Originally Posted by m42BMW View Post

@mynm

Impressive work! I may have made a small breakthrough on my new card, actually benched stable with scaled results up to 1220mhz. I havent kept up on screenshots, but I can say I'm adjusting TDP to 225, TDC to 190, LuT to 1150, vddci offset of -100, mvdd at 1050, dpm7 1300(good for 1180/1563mhz). This lets me bench to 1220 @ ~ 1.375v using AB for +mv, working on more. I'll be working more on it Wed this week I think.

@chris89

Any luck with that copper project?

Thanks.

I have tested this bios lowering lower DPM voltages and VDDCI, but I get an error with an hour gaming: 308A050011001250VDDCI1100voltageslike1140225W188AF.zip 42k .zip file

So I'm going to test if it is stable with vddci to 1.1v only, and them I will try to test to lower the lower DPM voltages.

It is weird that nobody is saying that a ~ +45mv voltage is needed for the same clocks for bios OC than with AB, and that you are getting the same scores than an ~ +40mhz AB OC. confused.gif

Edited: And for my the stock fan profile isn't enough to be at less than 80º, with fan to 68%~2050rpms, is not enough. So gpu is getting loud.
Edited by mynm - 11/21/16 at 10:40am
post #939 of 1079
Great job guys and mynm among all others. Hope you guys are happy with your bios and clocks. I found ultimately if you want stability on this card, you have to stick with factory voltage/ clock specifications. Ie 17.5%... If you want headache free stability that is...Anything less than 17.5% won't last and give great big headaches. I wasn't going to be happy with my 380x if it couldn't pull 1250Mhz. Which it couldn't because the thermal engineering involved to pull this was a bit of a headache. I did find the proper voltage to pull 1250Mhz though but it needed a plate similar to the RX480 blower to achieve this. Because Sapphire cut major corners in my opinion, the card couldn't hardly pull anything more than stock clocks. Although I did take the world record on 380x OpenCL compute with graphite on the core regulators with a week old gpu... As I tinkered with the heatsink more it all went to heck. If sapphire used the AMD Factory plate then a massive door of opportunity would open for them. I could have and may still purchase the copper to allow 1250Mhz. Until then my 380x is no-more for now until it is repaired. Baked it at 450F for 18 minutes and unlike usually, the solder melted at well below 450F and pretty much took out all the caps and a memory module. I touched the PCB too soon after nearly 20min and the module slid off the copper points sadly. So I would revise the temperature to probably no more than 12 minutes at around 350 to 375F (With 3 to 4 layers of tin foil over plastic and caps)(I used 1 weak layer). Baking it was also idiotically unneeded as I later foolishly found out. I had blue lines, thinking the gpu had physical issues I idiotically baked it when all along it was a faulty HDMI cable.. doh.gif

Anyway if you guys plan on buying a Polaris RX 480, I do find the factory blower to be ideal. As they engineered a cooling plate similar to the old GTX 200 series, among GTX 400 series, and other high end 512 bit variants like the 290X. With this plate, will save loads of headache and all easy clocking because the plate is positioned to cool all hot points by thermal pad. It lacks the pads for the memory regulators near the 6-pin, which I would add thermal material for to keep them cool. Although the blower has a so-so thermal ability core contact heatsink, the plate allows to add a Arctic Accelero III or IV and still leave the plate to distribute the heat of all components most effectively. Replacing the pads with fujipoly 17w/m K or 1mm thick Panasonic PGS would allow to clock sky-high. These Polaris GPU's can pull 30 percent or higher clock-ability with this plate. I am very excited myself to test out the Polaris 8GB 480 GPU, replacing the headache 380x. I plan on achieving 1563Mhz which gives desirable pixel and texel numbers to the eye by GPU-Z... 32 x 1563 = 50 Billion 16 Million Pixel's per second. Among 144 x 1563 = 225 Billion 72 Million Texel's per second. To note that difference from 1266Mhz to 1563Mhz is 23.45% increase overall which is a lot. It should be very capable of achieving 1563Mhz very stable (More clock for benchmarks if your nuts like me and many others haha), while dialing in the proper TDP to let it not spend all day at 1563Mhz but hop around according to thermal threshold limits (225W TDP Should be right on). I like the plate so much, I may even send it off to a Plating Company that can sandblast it down to whatever material it is... I'd say weaker than alluminum ie probably 20 to 200 watts per meter kelvin. Have the plating company plate, the plate in Beautiful Shiny Gorgeous Copper with it's 500 watts per meter kelvin haha... I've Always wanted to try this... I really like a nice shiny smooth piece of Copper haha

http://www.tomshardware.com/reviews/amd-radeon-rx-480-polaris-10,4616-3.html


Edited by chris89 - 11/23/16 at 2:50pm
post #940 of 1079
@chris89 Thanks:thumb:


I think that this bios with 1100/1500 clocks, 1.25V vddc and 1.1V vddci, is stable. 308A050011001250VDDCI1100voltageslike1140225W188AF.zip 42k .zip file
I will test it more.

About Load Line Slope I think that maybe I'm wrong thinking that is reducing the voltage variability. Maybe all the LLS have the same average voltage so that all will have the same stable voltage.
But you have to decreace offset because of the different initial drooped voltage.

I have tested the timings these bioses: http://nerdralph.blogspot.com.es/2016/09/advanced-tonga-bios-editing.html he say that I could got to 1625mhz in memory. I have tested and I see memory error and artifact, but I see that firestrike graphics score is increased from 9450 to 9625. Warning: Spoiler! (Click to show)

So that biggrin.gif, i'm going to test to add at VoltageObjectInfo table an svid2 table for mvdd with a +25mv offset to see if it's working, like in vddci and vddc smile.gif. Edited itsn't working, I get a black screen after windows logon.
Edited by mynm - 11/25/16 at 11:42am
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