Overclock.net › Forums › Graphics Cards › AMD/ATI › Polaris Bios Editing ( RX5xx / RX4xx )
New Posts  All Forums:Forum Nav:

Polaris Bios Editing ( RX5xx / RX4xx ) - Page 171

post #1701 of 3413
Nice man. just an fyi opening the image in paint and saving as jpeg will reduce the file to 1MB or so for upload

Oh yeah about memory errors.. kind of a pain right? Memory errors is kinda like meaningless unless your artifacting in game... if no artifacting they can be ignored.

My refurb visiontek rx 480 after using super cheap pads on everything on the pcb. Thermagon t-6100 flex on everything except gpu core.

I can clock to 1407mhz but it takes a while for the fan to kick up and in that window is when it could crash, waiting for the fan spin up... Downside to 1407mhz stock voltage or even 1416mhz 1233mv...

I think that's why 1416mhz does so well is it's just below stock peak voltage of 65288 (1.25v).

Anyway so far firestrike showed leveling out core & vrm 75C at 1407mhz 2266mhz memory... I am seeing over 200 errors but no artifacts or instability when fan is spooled.

FPS is still spiking near 90fps in parts but haven't hit over 15,800 yet... Only first run through on new card. Which has ASIC 81.6% at the moment. However this card is using more power than the other one.

I think it was the Thermeez, that stuff reduced power to about 210 watts vs T6100 thermagon literally 240 watts on HWInfo... said it hit 133fps in firestrike but I never saw it...

So anyway yeah the core VRM pad is pretty much the most important part among all the other 4-5 hot spots I added pads to. Once fan spools, clock holds at 1407mhz and core temperature keeps lowering.

The Core VRM should contact an Extraordinarily thermally conductive slightly pliable compressable material that is very hard... It needs to be rock hard material that can contour to each vrm. The rigidity strength/ hard/ yet compressible material is key to spiking beyond 16,000 points on this card.

Maybe I'll just do it already and buy Phase Change compressible Panasonic PGS for this application... www.digikey.com

Maybe hit 32,000 firestrike on crossfire idk haha

What really made the difference was using NUMA mode with C States disabled.. with C states disabled max boost cpu clock is disable ie 3066mhz.

Limited to 2933mhz max clock on Xeon X5650s but memory bandwidth is near 42GB/s compared to 15-31GB/s SMP. NUMA increased Firestrike Big time on dual xeon's.

I think the cheapest option is time consuming but worth it. Buying Ceramic Powder pure, mixing with water a tiny bit and stirring to a clay. Form each VRM square by hand and place in oven at 250f for 10 minutes. Then apply and compress.. this way no electrical conductivity safety and near rock hard compressible thermal performance. Since ceramic is a prized mineral that is highly thermally conductive under pressure and compression in a dense form.
Edited by chris89 - 3/29/17 at 2:03pm
post #1702 of 3413
Well I think I found it. Ceramic alumina thermal substrate. I suppose 3m adhesive should be placed on vrm before contact to prevent electrical conductivity. Ceramic with it's alluminum oxide performs about 25w/m k +/- i'd say as much as 1000 among the z/y axis.

Z axis is side to side width conductivity in a lateral direction. Which is sort of key it conducts not only on the lateral axis but also the horizontal axis. Which means saturating and spreading the heat to fill the whole substrate but also transfer to the heatsink contact area.

I'm sure it's 10-50x more conductive than Thermeez I used which scored 15,819 points firestrike graphics.

So anyway here it is.. 7 bucks... stuff is used on the space shuttle haha

http://www.ebay.com/itm/LOT-OF-TWO-2-HIGH-PURITY-ALUMINA-CERAMIC-SUBSTRATES-TILES-STRIPS-No-46-/301452134461?hash=item462ff2883d:g:n7kAAOSwg3FUmFv2

10 bucks for possibly more performance so it's thinner by 0.3mm or so means more performance if gap can work with that thickness that is...

http://www.ebay.com/itm/LOT-OF-FOUR-4-THIN-RECTANGULAR-ALUMINA-CERAMIC-STRIPS-SHEET-PLATE-No-504-/291493686789?hash=item43de60ae05:g:bk0AAOSweW5VgASa


Edited by chris89 - 3/29/17 at 2:41pm
post #1703 of 3413
@chris89

sorry to bother you but i can't stop thinking about your firestrike result smile.gif. I carefully read all your posts and it is making even more misterious. According to 3dmark db your score would be the fastest run of firestrike on a computer equiped with X5650 with the lowest clocks set.

http://www.3dmark.com/search#/?mode=advanced&url=/proxycon/ajax/search/cpugpu/fs/P/2056/1092/500000?minScore=0&cpuName=Intel Xeon Processor X5650&gpuName=AMD Radeon RX 480

People with card clocked 100 more mhz do not reach such results due to limitations of the platform PCI-E 2.0 etc.

http://www.3dmark.com/search#/?mode=advanced&url=/proxycon/ajax/search/cpugpu/fs/P/2056/1092/500000?minScore=0&cpuName=Intel Xeon Processor X5650&gpuName=AMD Radeon RX 480

Even if PC is equiped with 2 CPUs it will barerly impact the graphic score within the measurment error.

I'm still waiting for validated results.

One more thing, i strongly appreciate that what you are doing for a whole community but in some situations you mislead users. One more time, VRM is not limiting your cards, it is not a reason of instability or artifacts, especially if VRM temp is below 70 degrees or around.

I understand that you have integrated cooler which is responsible for GPU and VRM at the same time, and it is your blind shot to blame VRM for some issues. I have two seperated cooling system for VRM and GPU. VRM is hot as hell no instabillity, temps around 100, it's even more than on ref coolers. GPU below 45.

Last thing, reading this and few other forums i found an information that GPU temperature may be a factor, depending of sample above some temperature GPU start to behave instabile.

@jeztur2003

there may be an issue with WattMan and 3rd party oc software like Afterburner or Trixx. Keep on monitoring OCLMembench, i have noticed that for some unknown reson you can set 2100 mhz on mems but in OCLMembench you will find results characteristic for stock clocks. This may be the answer why some time you have errors some times not. Another thing is the temperature of IC chips, i found a correlation between temp and amount of errors, cooler = less.
Little facker
(13 items)
 
  
CPUMotherboardGraphicsRAM
i5-6600K @ 4.6 Asrock Fatal1ty Z170 Gaming-ITX/ac Sapphire Ref RX480 8GB @ 1480/2100 DDR4 HyperX Fury 2133 MHz cl14 @ 3100 MHz cl13 
Hard DriveHard DriveCoolingOS
Samsung SM951 Seagate ST2000DM001 Custom loop Win 10 Edu 
MonitorKeyboardPowerCase
Iiyama X2483HSU Ozone Strike Battle Corsair HX520EU Corsair Obsidian 250D 
Mouse
Logitech MX518 
  hide details  
Reply
Little facker
(13 items)
 
  
CPUMotherboardGraphicsRAM
i5-6600K @ 4.6 Asrock Fatal1ty Z170 Gaming-ITX/ac Sapphire Ref RX480 8GB @ 1480/2100 DDR4 HyperX Fury 2133 MHz cl14 @ 3100 MHz cl13 
Hard DriveHard DriveCoolingOS
Samsung SM951 Seagate ST2000DM001 Custom loop Win 10 Edu 
MonitorKeyboardPowerCase
Iiyama X2483HSU Ozone Strike Battle Corsair HX520EU Corsair Obsidian 250D 
Mouse
Logitech MX518 
  hide details  
Reply
post #1704 of 3413
The main difference between my card and all others is no one has used Thermeez ever before on core VRM. I have not been able to achieve such results on any thermal material as of yet. I receive way more memory errors at 2266mhz than with using thermeez.

I'll post new results when the Ceramic thermal substrate comes in haha

Want more results? I have the rx 480 in my system right now on one xeon and 24gb ram... other xeon cut out and it's in high efficiency mode at the moment.

Main difference is I don't touch software overclocking software at all... It's all programmed into the bios like the card came from the factory at said speed.

I find huge differences between a bios with said clocks vs software though I don't use software so i'm sure that's why...

I program, crank fans to manual 75% or just open GPUz render until fans spool up and hold 1407/1420mhz... im testing 1420mhz atm 2266mhz memory.. then i quickly go in game while the thermals hold the clock...

Going straight in results in either horrible scores or a crash... I wait for the card to spool up to full speed like a huge turbo and then hit the gas when spooled to peak boost basically haha

Like an electrical slingshot so to speak haha
Edited by chris89 - 3/29/17 at 3:44pm
post #1705 of 3413
everything is fine but temperature on VRM do not generate fps. Ok, there is no need to keep everyone in tense, show the validated score and basically it will close the case. Where the problem is? If you will be able to show such results probably everyone will replace TIM for this nasa thing which you are using. We can even name this method by your name.

As far as i know for graphic card it doesn't matter where you set clocks, in bios or in program. It's not a motherboard, here everything is much more simple. Anyway i had 5 spare minutes and i have checked that. I set fixed clocks in bios, no difference. I knew the result before i started but i wanted to exclude such possiblity.

Show the validated result or i will have to treat this as a custom run with custom settings as i do not see any other reason for such massive difference, incomparible result, you can tell that you had 200 fps as well. Reveal the cards ...
Little facker
(13 items)
 
  
CPUMotherboardGraphicsRAM
i5-6600K @ 4.6 Asrock Fatal1ty Z170 Gaming-ITX/ac Sapphire Ref RX480 8GB @ 1480/2100 DDR4 HyperX Fury 2133 MHz cl14 @ 3100 MHz cl13 
Hard DriveHard DriveCoolingOS
Samsung SM951 Seagate ST2000DM001 Custom loop Win 10 Edu 
MonitorKeyboardPowerCase
Iiyama X2483HSU Ozone Strike Battle Corsair HX520EU Corsair Obsidian 250D 
Mouse
Logitech MX518 
  hide details  
Reply
Little facker
(13 items)
 
  
CPUMotherboardGraphicsRAM
i5-6600K @ 4.6 Asrock Fatal1ty Z170 Gaming-ITX/ac Sapphire Ref RX480 8GB @ 1480/2100 DDR4 HyperX Fury 2133 MHz cl14 @ 3100 MHz cl13 
Hard DriveHard DriveCoolingOS
Samsung SM951 Seagate ST2000DM001 Custom loop Win 10 Edu 
MonitorKeyboardPowerCase
Iiyama X2483HSU Ozone Strike Battle Corsair HX520EU Corsair Obsidian 250D 
Mouse
Logitech MX518 
  hide details  
Reply
post #1706 of 3413
Quote:
Originally Posted by chris89 View Post

Well I think I found it. Ceramic alumina thermal substrate. I suppose 3m adhesive should be placed on vrm before contact to prevent electrical conductivity. Ceramic with it's alluminum oxide performs about 25w/m k +/- i'd say as much as 1000 among the z/y axis.
whats wrong with using the coolabs 39w/m k stuff or the better stuff on the market cant remember its name but you get 60-70w/m k i have the cool labs on my cpu this stuff is ultra thin applied with a cotton bud after achol wipe /light wire wooled etc this stuff is the jobby and when bought at 15mls you get more than enough for a good few cpu's chips only issue with cool labs is no ALU coolers

https://www.amazon.co.uk/Coollaboratory-LIQUID-PRO-Liquid-Pro/dp/B001PE5XAC

the ultra is 44w/mk

or the dady

http://www.thermal-grizzly.com/en/products/26-conductonaut-en
73w/mk

all three are thinner than a sheet of paper when applied
Quote:
Originally Posted by chris89 View Post

I find huge differences between a bios with said clocks vs software though I don't use software so i'm sure that's why...

i found this in cpus also, software just sucks and causes stability issues. its ok to get a benchmark figures from there. something to aim for

you could also decrease the load on your cpu which would help the graphics preform even better by using a smaller OS or do you do that already?
granted it be marginal but its worth a shot. i noticed a 1 ghz overclock more from windows 7 to 10. remeber your OS can determine stability issues as well as graphical stability
Edited by Mgrandy - 3/29/17 at 6:58pm
my pride n joy
(18 items)
 
  
CPUMotherboardGraphicsRAM
AMD FX-9370 970 Performance/3.1 Sapphire Nitro Radeon RX 460 (4GB)  Kingston  
Hard DriveHard DriveOptical DriveCooling
wd20earx 9wg8g94 x4 ihas20  corsair h100 
OSMonitorKeyboardPower
windows 7 64bit pro 42" LG tv A4 Tech X7 G800V Gaming Keyboard AeroCool Integrator 850W ATX Power Supply 
CaseMouseMouse PadOther
NZXT phantom full white case Sumvision Nemesis The Neon 2-in-1 Gaming Mouse ... Sumvision Nemesis The Neon 2-in-1 Gaming Mouse ... Coollaboratory Liquid Pro Liquid Metal Thermal... 
OtherOther
ms lx 3000 xbox 360 pad for pc 
  hide details  
Reply
my pride n joy
(18 items)
 
  
CPUMotherboardGraphicsRAM
AMD FX-9370 970 Performance/3.1 Sapphire Nitro Radeon RX 460 (4GB)  Kingston  
Hard DriveHard DriveOptical DriveCooling
wd20earx 9wg8g94 x4 ihas20  corsair h100 
OSMonitorKeyboardPower
windows 7 64bit pro 42" LG tv A4 Tech X7 G800V Gaming Keyboard AeroCool Integrator 850W ATX Power Supply 
CaseMouseMouse PadOther
NZXT phantom full white case Sumvision Nemesis The Neon 2-in-1 Gaming Mouse ... Sumvision Nemesis The Neon 2-in-1 Gaming Mouse ... Coollaboratory Liquid Pro Liquid Metal Thermal... 
OtherOther
ms lx 3000 xbox 360 pad for pc 
  hide details  
Reply
post #1707 of 3413
I'll throw up some validated results soon.

I don't know about that liquid metal thermal compound. I use ceramic, amazing stuff and dirt cheap.
post #1708 of 3413
I ran tests on Thermagon t6100 cooling everything except Arctic Silver Ceramique 2 on core...

Nothing special I am not seeing anywhere near what I saw on better thermal material.

I'm on one Xeon since with my 2nd Xeon riser I can't boot the system. So I bought "Contact Cleaner" to clean the memory slots/ riser pins and hopefully getting them both chiming on all cylinders soon.

To see beyond 15,800 gfx I need at least 2,100 total frames on Test 1 & over 2,600 total frames Test 2.

On Thermagon T6100 I am 10-20fps lower on the "High End" fps spikes in Test 1 & 2. Where on Thermeez I would near 100fps spikes in parts of Test 1 & 2.

On Thermagon T6100 I see barely 90fps "Spikes" on Test 1 & 2.

Basically the more Exotic the thermal material used, would see more "Freak" fps spikes in Test 1 & 2. Those "Freakish" FPS Spikes is what I count on for max scores.




Edited by chris89 - 3/30/17 at 12:33am
post #1709 of 3413
" liquid metal thermal" is 100% pure well 99% its like mercury in a Thermometer at room temp so easy spread real thin then goes hard at 50 deg ish the stuff you posted was only 22 and 96% pure you can see how pure it is by looking in your reflection once its spread, honestly i wouldn't use anything else now. i used to use a 11 w/mk previous to this and the difference is amazing for such a thin layer but it is hard to apply. i had to use bit of elbow grease with the cotton bud just to get it to bond to the cpu. think it may have warmed it up a bit. by way of thin if you make a finger print on your cpu with this stuff it aint going to bond due to the grease been more thick than the layer your applying. thats what i mean by super thin.
Arctic Silver Ceramique 2
i wouldnt really rate that although it does have a good point its not silicon based. i find silicon are a waste of time almost always for cooling, from what i see thats max 8.2w/mk at 100% pure.

anyway amazing stuff dont know if i would put it on the vrm chips thou that might be a bit hard. dont want this stuff touching any solder but it could be possible

are you cooling with air or water?
Edited by Mgrandy - 3/30/17 at 12:56am
my pride n joy
(18 items)
 
  
CPUMotherboardGraphicsRAM
AMD FX-9370 970 Performance/3.1 Sapphire Nitro Radeon RX 460 (4GB)  Kingston  
Hard DriveHard DriveOptical DriveCooling
wd20earx 9wg8g94 x4 ihas20  corsair h100 
OSMonitorKeyboardPower
windows 7 64bit pro 42" LG tv A4 Tech X7 G800V Gaming Keyboard AeroCool Integrator 850W ATX Power Supply 
CaseMouseMouse PadOther
NZXT phantom full white case Sumvision Nemesis The Neon 2-in-1 Gaming Mouse ... Sumvision Nemesis The Neon 2-in-1 Gaming Mouse ... Coollaboratory Liquid Pro Liquid Metal Thermal... 
OtherOther
ms lx 3000 xbox 360 pad for pc 
  hide details  
Reply
my pride n joy
(18 items)
 
  
CPUMotherboardGraphicsRAM
AMD FX-9370 970 Performance/3.1 Sapphire Nitro Radeon RX 460 (4GB)  Kingston  
Hard DriveHard DriveOptical DriveCooling
wd20earx 9wg8g94 x4 ihas20  corsair h100 
OSMonitorKeyboardPower
windows 7 64bit pro 42" LG tv A4 Tech X7 G800V Gaming Keyboard AeroCool Integrator 850W ATX Power Supply 
CaseMouseMouse PadOther
NZXT phantom full white case Sumvision Nemesis The Neon 2-in-1 Gaming Mouse ... Sumvision Nemesis The Neon 2-in-1 Gaming Mouse ... Coollaboratory Liquid Pro Liquid Metal Thermal... 
OtherOther
ms lx 3000 xbox 360 pad for pc 
  hide details  
Reply
post #1710 of 3413
Quote:
Originally Posted by chris89 View Post

Trust me 99 errors, isn't a lot but I guess if we or you would much rather see 0 errors haha try this bios

By way way reason I say that isn't a lot since I've seen way over 1,000,000 errors to 1,000,000,0000...

So 99 isn't much haha

I wonder if your ASIC goes up with 1033mv memory? here are these two... 275GB/s memory 2149Mhz

Jeztur2003_1320mhz_2149Mhz_86C_88C_86C.zip 110k .zip file
Jeztur2003_1320mhz_2149Mhz_1033mv_86C_88C_86C.zip 110k .zip file

Your Total Crossfire Throughput of both cards is...

84 billion 480 million pixel's every second
380 billion 160 million texel's every second
&
550GB/s combined GPU memory Bandwidth... Not bad

If you bought reference 480's from Visiontek and ran 1407Mhz core & 2266Mhz memory

90 billion 48 million pixel's every second
405 billion 216 million texel's every second
&
580GB/s combined GPU memory Bandwidth

That's about 6% haha not much but it makes a difference for sure haha

Good morning Chris. So I did some gaming last night with my brother and had about 5 crashes where game would freeze and I'd have to power cycle. I didn't get logs but I can tell you the core never got over 76 and temp on the rest of my rig were well within operating temps. Unfortunately it takes awhile to reproduce this about 30 minutes or more but I have noticed the more strain on the cards the quicker it happens so I'm going to log and run some 4k cause with the crash I can't see the real time logging to see what went wrong. My guess would be memory speed since core is not getting hot I'll try to get a log on here asap. This is with the 1000mv bios. BTW thanks for the tip to downsize the screenshots here's a beauty from our game session last night.

Also on a single bench of rise of the tomb raider mgpu 1080p ultra which usually runs smooth i got a major dip in the second scene with the waterfall over half fps dip. I ran hwinfo just for that bench here's the results.

Edited by jeztur2003 - 3/30/17 at 4:21am
Double Dragon
(16 items)
 
  
CPUMotherboardGraphicsGraphics
AMD FX-8370 970 PRO GAMING/AURA Radeon (TM) RX 480 Graphics Radeon (TM) RX 480 Graphics 
RAMRAMRAMRAM
Crucial Technology  Crucial Technology  Crucial Technology  Crucial Technology  
CoolingOSMonitorKeyboard
Corsair H110i Windows 10 X64 HP 27es Corsair K55 RGB 
PowerCaseMouseMouse Pad
Rosewill Hive 1000 Watt Bronze Certified Phateks P400 Tempered Glass Red/Black Corsair Harpoon RGB Corsair MM800 RGB 
  hide details  
Reply
Double Dragon
(16 items)
 
  
CPUMotherboardGraphicsGraphics
AMD FX-8370 970 PRO GAMING/AURA Radeon (TM) RX 480 Graphics Radeon (TM) RX 480 Graphics 
RAMRAMRAMRAM
Crucial Technology  Crucial Technology  Crucial Technology  Crucial Technology  
CoolingOSMonitorKeyboard
Corsair H110i Windows 10 X64 HP 27es Corsair K55 RGB 
PowerCaseMouseMouse Pad
Rosewill Hive 1000 Watt Bronze Certified Phateks P400 Tempered Glass Red/Black Corsair Harpoon RGB Corsair MM800 RGB 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: AMD/ATI
Overclock.net › Forums › Graphics Cards › AMD/ATI › Polaris Bios Editing ( RX5xx / RX4xx )