Overclock.net › Forums › Graphics Cards › AMD/ATI › ASUS R9 290X DCIIOC CrossfireX overclocking
New Posts  All Forums:Forum Nav:

ASUS R9 290X DCIIOC CrossfireX overclocking

post #1 of 20
Thread Starter 
Howdy,

Info:
OS: Windows 10 64bit
Driver: Crimson 16.4.2
CPU: Intel i7-6700K @4.6GHz 1.28V (auto)
GPU: ASUS R9 290X DCIIOC x2
RAM: Corsair Vengeance LPX 16 GB DDR4 Red 3200 MHz
PSU: Corsair AX 1200i, 1200W
Custom waterloop for CPU and both GPUs
OC program: Sapphire Trixx 5.2.1

Benchmarks:
Stock: 1000/1250
OC: 1180/1560 +100mV

Valley Extreme HD, stock: 4622, core temp 55C, VRM temp 65C (35C and 29C idle)
Valley Extreme HD OC: 5027, core temp 61C, VRM temp 73C

Heaven Extreme +1920x1080 + Fullscreen, stock: 2655, core temp 53C, VRM temp 58 C
Heaven Extreme +1920x1080 + Fullscreen, OC: 3007, core temp 61C, VRM temp 74C

Fire Strike, Stock: 17668, core temp 55C, VRM temp 51C
Fire Strike, OC: 19688, core temp 58C, VRM temp 68C

I'm trying to find a stable overclock for my cards, but I'm having some trouble. They are watercooled, so I have no thermal problems. The most annoying and time consuming problem I have is that when an overclock fails and the display driver crashes it requires a complete reinstall...... This has happened about 20 times now and I'm getting really annoyed at this problem.

After a driver crash I reboot as usual, the the BIOS post comes, then Windows 10 boot, and when I'm supposed to se the login screen it's all black instead. The only bypass for this that I've found is safe mode boot and then run DDU (Display Driver Uninstaller 15.7.5.5) to remove the driver and then reinstall. This happened with Crimson 16.4.1 and the last Catalyst driver also. Every time I reinstall I do a complete removal first and then a custom installation where I check everything except the Evolved Gaming App.

When I overclock I try to find this so-called golden ratio, but I am uncertain of how I know when I've found it. Most overclocks are stable with the ratio 1.28-1.33, up to a point that is. The best overclock I've managed is 1200/1600 +50% power limit and +150mV. Temperatures are about 65C on core and 80C on VRM (except with FurMark stress test, there VRM goes beyond 100C, so I stop the test). This works fine in Unigine Heaven 4.0, Valley 1.0, Fire Strike (Extreme and Ultra included), but not so much in games. BF3/BF4 are semi-stable. They have random crashes after 1-2 hours gameplay generally. Rise of the Tomb Raider crashes when I load a save. The built in benchmark fails to even load. Dark Souls 3 is completely fine though.

These cards are factory clocked at 1050/1350, but my games aren't stable unless I have them at stock clock, 1000/1250.

I use Sapphire Trixx because it is the only program (that I have found) that allows me to have +200mV on the GPU. I have disabled ULPS and synchronized CrossFire cards in settings.

I want to continue overclocking to get the most out of these cards, but the constant crashes and reinstallations are too exhausting and annoying.

Questions:

1) Is there anything I am doing wrong that causes the driver to crash so hard that it is unusabe afterwards? Does anyone else experience this?

2) Does anyone have any tips on how to stabilize and overclock? I have found that increasing the voltage helps, but only to a certain degree. It might stabilize the overclock for a benchmark, but games seem unstable no matter what.

3) How can I be certain that I've found the golden ratio?

4) I read somewhere that these cards need more power than what can be suplied by a single PCI-E cable, so they suggested using two cables for each graphics card instead of one split one for each. Will this make a difference? My cards only have 8-pin + 6-pin.
post #2 of 20
Quote:
Originally Posted by Jokkis View Post

Howdy,

Info:
OS: Windows 10 64bit
Driver: Crimson 16.4.2
CPU: Intel i7-6700K @4.6GHz 1.28V (auto)
GPU: ASUS R9 290X DCIIOC x2
RAM: Corsair Vengeance LPX 16 GB DDR4 Red 3200 MHz
PSU: Corsair AX 1200i, 1200W
Custom waterloop for CPU and both GPUs
OC program: Sapphire Trixx 5.2.1

Benchmarks:
Stock: 1000/1250
OC: 1180/1560 +100mV

Valley Extreme HD, stock: 4622, core temp 55C, VRM temp 65C (35C and 29C idle)
Valley Extreme HD OC: 5027, core temp 61C, VRM temp 73C

Heaven Extreme +1920x1080 + Fullscreen, stock: 2655, core temp 53C, VRM temp 58 C
Heaven Extreme +1920x1080 + Fullscreen, OC: 3007, core temp 61C, VRM temp 74C

Fire Strike, Stock: 17668, core temp 55C, VRM temp 51C
Fire Strike, OC: 19688, core temp 58C, VRM temp 68C

I'm trying to find a stable overclock for my cards, but I'm having some trouble. They are watercooled, so I have no thermal problems. The most annoying and time consuming problem I have is that when an overclock fails and the display driver crashes it requires a complete reinstall...... This has happened about 20 times now and I'm getting really annoyed at this problem.

After a driver crash I reboot as usual, the the BIOS post comes, then Windows 10 boot, and when I'm supposed to se the login screen it's all black instead. The only bypass for this that I've found is safe mode boot and then run DDU (Display Driver Uninstaller 15.7.5.5) to remove the driver and then reinstall. This happened with Crimson 16.4.1 and the last Catalyst driver also. Every time I reinstall I do a complete removal first and then a custom installation where I check everything except the Evolved Gaming App.

When I overclock I try to find this so-called golden ratio, but I am uncertain of how I know when I've found it. Most overclocks are stable with the ratio 1.28-1.33, up to a point that is. The best overclock I've managed is 1200/1600 +50% power limit and +150mV. Temperatures are about 65C on core and 80C on VRM (except with FurMark stress test, there VRM goes beyond 100C, so I stop the test). This works fine in Unigine Heaven 4.0, Valley 1.0, Fire Strike (Extreme and Ultra included), but not so much in games. BF3/BF4 are semi-stable. They have random crashes after 1-2 hours gameplay generally. Rise of the Tomb Raider crashes when I load a save. The built in benchmark fails to even load. Dark Souls 3 is completely fine though.

These cards are factory clocked at 1050/1350, but my games aren't stable unless I have them at stock clock, 1000/1250.

I use Sapphire Trixx because it is the only program (that I have found) that allows me to have +200mV on the GPU. I have disabled ULPS and synchronized CrossFire cards in settings.

I want to continue overclocking to get the most out of these cards, but the constant crashes and reinstallations are too exhausting and annoying.

Questions:

1) Is there anything I am doing wrong that causes the driver to crash so hard that it is unusabe afterwards? Does anyone else experience this?

2) Does anyone have any tips on how to stabilize and overclock? I have found that increasing the voltage helps, but only to a certain degree. It might stabilize the overclock for a benchmark, but games seem unstable no matter what.

3) How can I be certain that I've found the golden ratio?

4) I read somewhere that these cards need more power than what can be suplied by a single PCI-E cable, so they suggested using two cables for each graphics card instead of one split one for each. Will this make a difference? My cards only have 8-pin + 6-pin.

Why do you need so much oc on the gpus? What resolution are you using?

Also, i would keep the vrms from hitting 80C at any oc even if they are rated for much higher temp. I assume they are not full waterblocks 'cause the vrms are much warmer than the core.

Try using two cables for each gpu. It could be the psu.

Stay away from Furmark. BTW, did you accept the Overdrive in Crimson?

EDIT: For oc'ing, start to find the most stable oc on the core first keeping the memory at stock. It could be the 1600 memory that is unstable. Also, with my 290s, I put the better oc'er as primary BUT my oc'ed values are based on the slower card. Meaning, I tested each one individually.
Edited by rdr09 - 4/29/16 at 11:07pm
Second Intel Rig
(16 items)
 
  
CPUMotherboardGraphicsRAM
2700 4.5/ 1.28 77 1050 16 / 1866 
Hard DriveCoolingOSMonitor
1000 212 10 64 32 1080 
PowerCase
700 912 
  hide details  
Reply
Second Intel Rig
(16 items)
 
  
CPUMotherboardGraphicsRAM
2700 4.5/ 1.28 77 1050 16 / 1866 
Hard DriveCoolingOSMonitor
1000 212 10 64 32 1080 
PowerCase
700 912 
  hide details  
Reply
post #3 of 20
Thread Starter 
Quote:
Originally Posted by rdr09 View Post

Why do you need so much oc on the gpus? What resolution are you using?

Also, i would keep the vrms from hitting 80C at any oc even if they are rated for much higher temp. I assume they are not full waterblocks 'cause the vrms are much warmer than the core.

Try using two cables for each gpu. It could be the psu.

Stay away from Furmark. BTW, did you accept the Overdrive in Crimson?

EDIT: For oc'ing, start to find the most stable oc on the core first keeping the memory at stock. It could be the 1600 memory that is unstable. Also, with my 290s, I put the better oc'er as primary BUT my oc'ed values are based on the slower card. Meaning, I tested each one individually.

Hi,

I just want more power. I run all games with ease at stock, but i built this rig to overclock it. So I'm just trying to find the limit.

My resolution is only 1920x1080.

The waterblocks are full cover except for the very end: https://www.ekwb.com/shop/ek-fc-r9-290x-nickel-rev-2-0
I read somewhere that VRM can handle 125C, and knowing AMD build their GPUs to run at 95C, I am not worried about these temps.

I do have Overdrive settings in Crimson, yes, but the options are kind of limited. Only power limit along with core and memory clocks.

I will try overclocking the way you said with the core first then memory. This is what I did with my GTX 680, but for my 290Xs I've been completely distracted by this golden ratio. I will also try with 2 cables.

Thanks for your reply and tips
post #4 of 20
Quote:
Originally Posted by Jokkis View Post

Hi,

I just want more power. I run all games with ease at stock, but i built this rig to overclock it. So I'm just trying to find the limit.

My resolution is only 1920x1080.

The waterblocks are full cover except for the very end: https://www.ekwb.com/shop/ek-fc-r9-290x-nickel-rev-2-0
I read somewhere that VRM can handle 125C, and knowing AMD build their GPUs to run at 95C, I am not worried about these temps.

I do have Overdrive settings in Crimson, yes, but the options are kind of limited. Only power limit along with core and memory clocks.

I will try overclocking the way you said with the core first then memory. This is what I did with my GTX 680, but for my 290Xs I've been completely distracted by this golden ratio. I will also try with 2 cables.

Thanks for your reply and tips

Yes, i understand, watercooled you want them high oc's. At 1080, though, it will be hard to push the cards however powerful your cpu 'cause of the infamous amd cpu overhead. If your monitor will let you use VSR, then go as high as you can. 1440 would be nice.

See, you made the same mistake i did by accepting Overdrive. Whenever you oc with Trixx or Afterburner, Crimson/Overdrive will adjust with it. Not sure which one Windows follow - Crimson or Trixx. Ever since Crimson I think Windows follow the settings in Crimson not the third party app. Only way i know to undo Overdirve is reinstall Windows.

Your temps are weird for full waterblocks and, yes, those are full blocks. I say weird 'cause i have eks and the vrms are always cooler than the core. vrm1 does equal the core at high oc's but not higher. You used the stock thermal pads? There are two sizes - 0.5 and 1. I think 1 is for the vrms, iirc. If you want to oc high, you need to use Fujipolys and they are expensive. 11wm/k at least just for the vrms. the stock pads will do for the memory.

High temps could also be caused by high ambient, lack of rad space or low water flow.

_____________

See, how high you can oc first at stock voltage keeping the memory at stock but MAX out the power limit for each GPU. Test 1100, 1125, etc. Just disable crossfire, shutdown system, and uplug power to the GPU not being tested. Make sure you hook up the monitor to the gpu being tested. Use Firestrike to test. As soon as artifacts show, stop the test. Lower the oc and test again.

I use GPUZ render test as an good alternative to Furmark to test temps. For crossfire, you gonna need three instances of GPUZ. The first one to strat the render test using the "?" icon. The other two set to sensor tabs to show temps. Set the temps to show MAX.





Again, set the temps to show MAX. Use the dropdown arrow.


You have to keep all the temps as low as possible, especially for oc's meant for games. BF4 is notorious. 65C is high for stock.
Second Intel Rig
(16 items)
 
  
CPUMotherboardGraphicsRAM
2700 4.5/ 1.28 77 1050 16 / 1866 
Hard DriveCoolingOSMonitor
1000 212 10 64 32 1080 
PowerCase
700 912 
  hide details  
Reply
Second Intel Rig
(16 items)
 
  
CPUMotherboardGraphicsRAM
2700 4.5/ 1.28 77 1050 16 / 1866 
Hard DriveCoolingOSMonitor
1000 212 10 64 32 1080 
PowerCase
700 912 
  hide details  
Reply
post #5 of 20
Thread Starter 
Quote:
Originally Posted by rdr09 View Post

Yes, i understand, watercooled you want them high oc's. At 1080, though, it will be hard to push the cards however powerful your cpu 'cause of the infamous amd cpu overhead. If your monitor will let you use VSR, then go as high as you can. 1440 would be nice.

See, you made the same mistake i did by accepting Overdrive. Whenever you oc with Trixx or Afterburner, Crimson/Overdrive will adjust with it. Not sure which one Windows follow - Crimson or Trixx. Ever since Crimson I think Windows follow the settings in Crimson not the third party app. Only way i know to undo Overdirve is reinstall Windows.

Your temps are weird for full waterblocks and, yes, those are full blocks. I say weird 'cause i have eks and the vrms are always cooler than the core. vrm1 does equal the core at high oc's but not higher. You used the stock thermal pads? There are two sizes - 0.5 and 1. I think 1 is for the vrms, iirc. If you want to oc high, you need to use Fujipolys and they are expensive. 11wm/k at least just for the vrms. the stock pads will do for the memory.

High temps could also be caused by high ambient, lack of rad space or low water flow.

_____________

See, how high you can oc first at stock voltage keeping the memory at stock but MAX out the power limit for each GPU. Test 1100, 1125, etc. Just disable crossfire, shutdown system, and uplug power to the GPU not being tested. Make sure you hook up the monitor to the gpu being tested. Use Firestrike to test. As soon as artifacts show, stop the test. Lower the oc and test again.

I use GPUZ render test as an good alternative to Furmark to test temps. For crossfire, you gonna need three instances of GPUZ. The first one to strat the render test using the "?" icon. The other two set to sensor tabs to show temps. Set the temps to show MAX.





Again, set the temps to show MAX. Use the dropdown arrow.


You have to keep all the temps as low as possible, especially for oc's meant for games. BF4 is notorious. 65C is high for stock.

Hi,

I'll have to order some new thermal pads then. Just checked, and the ones I have are EK brand with only 3-5W/mK. The best I can order from my supplier are Thermal Grizzly Minus Pad with 8W/mK. You are right, they cost way more. I might even have used 0.5mm also, thinking thinner is better, but I am certain I put them exactly where they should be placed. I do have backplates with a thermal pad on the back of the core and the VRM also.

I just remembered that when I do GPU benchmarks, only the VRM on GPU1 heats up that much. After one Valley run now VRM1 reached 59C and VRM2 only 42C at stock core/memory speed. Is this because only VRM1 is used when in crossfire, or because I correctly applied the thermal pads on GPU2?

I have 2 3x120mm radiators in serial for my loop, each with 3 static pressure fans blowing air up and out, 2 140mm air flow intake fans, 1 140mm air flow exhaust fan and 3 120mm fans beneath my mobo to keep good air circulation in my case (Thermaltake Core X2). The pump is fairly powerful: https://www.ekwb.com/shop/ek-xres-140-d5-vario-incl-pump

I can't recall seeing Crimson Overdrive as a selectable option while installing the display driver. Is it the AMD Radeon Settings one? If not, where do I unselect this? Do you install the big driver or just the minimal setup one?

I found the GPU-Z rendering test. Why hide the button like that??? Never knew it had a built in test.

Thanks again for your reply and tips.
post #6 of 20
Messing with the loop always raises the risk of breaking something. Do it right this time and be careful. Read the ek instruction. If you don't have one it can be downloaded in the ek site.

Might as well get MX-4 for the core. CLU if you are bold enuf. There is a video showing how to apply it. Tape around the core using carpenter's tape or electrical tape before applying.

I haven't replaced my pads since assembly. Got my first one at launch. Temps haven't change much but i also went from 3 120 rads to 240/360 rads. Search some more for a better thermal.

EDIT:

https://www.ekwb.com/shop/EK-IM/EK-IM-3831109869031.pdf

acrelic or acetal - same.

Here is Overdrive before accepting and using its features . . .



Once you accept the only way i know of reversing it is to reinstall Windows.
Edited by rdr09 - 5/3/16 at 12:29am
Second Intel Rig
(16 items)
 
  
CPUMotherboardGraphicsRAM
2700 4.5/ 1.28 77 1050 16 / 1866 
Hard DriveCoolingOSMonitor
1000 212 10 64 32 1080 
PowerCase
700 912 
  hide details  
Reply
Second Intel Rig
(16 items)
 
  
CPUMotherboardGraphicsRAM
2700 4.5/ 1.28 77 1050 16 / 1866 
Hard DriveCoolingOSMonitor
1000 212 10 64 32 1080 
PowerCase
700 912 
  hide details  
Reply
post #7 of 20
Thread Starter 
Hi again,

I reinstalled Windows 10 and made sure to only install the crimson driver 16.4.2. I unchecked everything that I could during the installation process. I see now that I misunderstood what you meant by the overdrive feature. I thought it was during the installation, not the tab in Crimson that you sent a screenshot of. But I have no signs of reduced GPU power with the bare minimum of the driver.

I no longer have problems with hard crashes that require safe mode boot, DDU cleaning and then reinstallation of the display driver. Now the display driver is able to recover after a crash. My Valley score has increased to 5284 on Extreme HD preset on a slightly lower clock.

I still have not got any higher stable overclocks, but I have yet to use 2 cables per GPU. Also, I ordered some Fujipoly thermal pads for both the VRM1 and VRM2.
I got the best kind I could find: http://www.amazon.com/Fujipoly-Extreme-System-Builder-Thermal/dp/B00MQ6AJU2?ie=UTF8&psc=1&redirect=true&ref_=oh_aui_detailpage_o02_s00
I checked the manual for the EK waterblock and got the right thickness. I also got some bigger pads of equal quality for the memory chips as well. I will post an update when I get and apply them.

I also got some good thermal paste for the CPU and GPU chips, Thermal Grizzly Aeronaut Thermal Grease Paste: http://www.amazon.com/Thermal-Grizzly-Aeronaut-Grease-Paste/dp/B00ZJSZM4C?ie=UTF8&psc=1&redirect=true&ref_=oh_aui_detailpage_o00_s00

I see now that I messed up and got the Aeronaut version and not the Kryonaut one (which is the one better than MX-4), but I have no heat issues with either of these chips so I'm not that worried. It's still god thermal paste. The problem was the VRM, and now I'll have the best.

Thank you for helping me. Overclocking is so much better now that I dont have to reinstall the driver after every single crash. I'll post an update when I've tried 2 cables and new thermal pads&paste.
post #8 of 20
Lovely. That strip of Fuji is mainly for VRM1, which really gets hot when oc'ing. VRM2 (one near the ports for monitors) can use some too if you have spare. Make sure you prioritize VRM1. Never used that grease but looks better than MX-4. Use the X method when applying the grease.




With the amount of rad space you have you should not see 60C in your loop. Oc'ing keep your temps at or below that temp at load. If your rads are upfront and on top . . . have the fans blowing into the case (sucking in). Follow the procedure of tightening the screws in the ek manual.
Second Intel Rig
(16 items)
 
  
CPUMotherboardGraphicsRAM
2700 4.5/ 1.28 77 1050 16 / 1866 
Hard DriveCoolingOSMonitor
1000 212 10 64 32 1080 
PowerCase
700 912 
  hide details  
Reply
Second Intel Rig
(16 items)
 
  
CPUMotherboardGraphicsRAM
2700 4.5/ 1.28 77 1050 16 / 1866 
Hard DriveCoolingOSMonitor
1000 212 10 64 32 1080 
PowerCase
700 912 
  hide details  
Reply
post #9 of 20
Thread Starter 
Hi again,

Last post :=). I finally switched all thermal pads. The pads previously on VRM1 were too thin (0.5mm) so I switched to 1.0mm. Now VRM1 temps are lower than the core. Not as cool as your cards, but it never goes above 60C, even with +150mV. I don't know what black magic you've used on your card to keep temps so low, but I feel my temps are acceptable, especially considering how hot AMD cards are designed to run.

Thanks for all your help. Now I'll get back to pushing that overclock to the limit:)
post #10 of 20
Quote:
Originally Posted by Jokkis View Post

Hi again,

Last post :=). I finally switched all thermal pads. The pads previously on VRM1 were too thin (0.5mm) so I switched to 1.0mm. Now VRM1 temps are lower than the core. Not as cool as your cards, but it never goes above 60C, even with +150mV. I don't know what black magic you've used on your card to keep temps so low, but I feel my temps are acceptable, especially considering how hot AMD cards are designed to run.

Thanks for all your help. Now I'll get back to pushing that overclock to the limit:)

Nice. No magic involve. Mine are stock. 150mV is a lot of volts, thus the higher temp. Maybe a larger rad space and lower ambient will bring down temp to 60 for the core. Or, just lower the oc.
Second Intel Rig
(16 items)
 
  
CPUMotherboardGraphicsRAM
2700 4.5/ 1.28 77 1050 16 / 1866 
Hard DriveCoolingOSMonitor
1000 212 10 64 32 1080 
PowerCase
700 912 
  hide details  
Reply
Second Intel Rig
(16 items)
 
  
CPUMotherboardGraphicsRAM
2700 4.5/ 1.28 77 1050 16 / 1866 
Hard DriveCoolingOSMonitor
1000 212 10 64 32 1080 
PowerCase
700 912 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: AMD/ATI
Overclock.net › Forums › Graphics Cards › AMD/ATI › ASUS R9 290X DCIIOC CrossfireX overclocking