Asus Motherboard issue

[sajeev3105]

I have been trying to overclock my CPU for the past 2 days but somehow I am not getting through. The issue is that once I overclock my i9 9900K , I pass intel's stress test, Aida64 runs for more than 1Hr 10 mins without any issues and bug, cinebench R20 also passes the test but the problem comes when I am trying to run Asus Realbench, none of the stress test nor the benchmark passes, in fact the system gets to blue screen may be after 5 seconds of running the test. Does anyone have any clue why this could be happening? My current voltage set is 1.29V, LLC: Level 6, All cores OC'd to 5.0GHz, AVX: set to 2. Some one told me to play around with VCCIO & VCCSA, but strange enough every time I change the values it resets itself to 1.2500.

Please kindly help and guide.

Motherboard: Asus Maximus XI Extreme (Latest Bios)
CPU: i9 9900K
RAM: 64GB Corsair Dominator Platinum @ 3200MHz

Any help will be deeply appreciated


Best Regards

Sajeev

 

[batboy]

Not always, but most of the time a blue screen when benching means RAM is set too high or needs more voltage. You didn't mention memory settings. Do you have the RAM overclocked too? You can pass all kinds of stress tests, but some benchmarks are tougher than others. Realbench definitely will torture your system. You might need to lower your overclock a bit for that benchmark.

 

[sajeev3105]

Hi there, nope I have not overclocked the ram, just running the xmp profile to make it run at 3200MHz as it is a 3200MHz ram. nothing much beyond that. Is there anything that I should be checking here? I have set the AVX offset value to 3.

Best Regards
Sajeev

 

[trents]

Are you using a stand a lone video card or the built in graphics of the CPU? Realbench uses the GPU and the CPU to render images. Cinebench R20, AIDA and XTU only use the CPU. If you are using the IGP (Integrated Graphics Processor) and not a separate video card then Realbench will be putting more stress on the processor than the other two.

What are you cooling the CPU with. When overclocked that 9900k will run hot! Have you done any temp monitoring while running the stress tests/benchmarks you are able to run? What are your maximum core temps?

Also, Realbench can be very glitchy in my experience but when it works it's an excellent real world stress test. If you remove the overclock and put the system back to stock frequency and voltage, does Realbench still blue screen?

If you are relying on AIDA and XTU for testing the stability of your overclock I would recommend running them overnight. They are not real demanding stress tests.

 

[EarthDog]

Blue screens can mean ANYTHING. Without knowing the error code, it is impossible to pinpoint just from getting a BSOD. What error is it?

What temps in stress testing?

That said, with 64GB of RAM that is a lot for the IMC to handle. If adding a bit more vcore doesnt help, try adding SA voltage... 1.3V or so and see if that helps. To test memory, run p95 BLEND and se if it passes. Small fft is mostly for cpu... blend is more memory.

 

[batboy]

Agreed, a blue screen can mean a lot of different things, but when I'm overclocked to the hilt and benching with fingers crossed, a blue screen is usually because I need more RAM voltage or I need to loosen timings, or back off the RAM OC. That said, if the OP is not overclocking the memory, then that shouldn't be an issue. Although, my G. Skill RAM is suppose to have 1.35v on SPD setting, but it only gives the RAM 1.20v. So, I always have to set it manually. Just one of many things you might want to check while troubleshooting.

 

[EarthDog]

Agreed, a blue screen can mean a lot of different things, but when I'm overclocked to the hilt and benching with fingers crossed, a blue screen is usually because I need more RAM voltage or I need to loosen timings, or back off the RAM OC. That said, if the OP is not overclocking the memory, then that shouldn't be an issue. Although, my G. Skill RAM is suppose to have 1.35v on SPD setting, but it only gives the RAM 1.20v. So, I always have to set it manually. Just one of many things you might want to check while troubleshooting.

Again, it depends... on the benchmark for one. It can BSOD because of anything. Once we get the reason, what tests he ran etc, we can narrow things down a bit. It may very well be memory, but a simple BSOD without any other information I'm not pointing fingers until I hear more. I've had plenty of BSODs that were not memory related.

The OP mentioned he did not overclock the RAM but enabled the XMP profile (with that much RAM it could need more though).

RE: Your RAM, that sounds backwards. Your GSkill RAM likely has 1.35V on XMP, but JEDEC should be rated at 1.2V and speed is 2133 - this is the platform's base spec with that CPU and how it boots without XMP profiles enabled typically (for compatibility reasons). You can see this in CPUz. I have similar sticks...

 

[batboy]

Yep, my CPU-Z says about the same exact thing. Never understood why, just always used manual vdimm after I noticed.

 

[wingman99]

I would test the memory with running XMP without overclocking the processor first to see if the rig will BSOD.

 

[sajeev3105]

Are you using a stand a lone video card or the built in graphics of the CPU? Realbench uses the GPU and the CPU to render images. Cinebench R20, AIDA and XTU only use the CPU. If you are using the IGP (Integrated Graphics Processor) and not a separate video card then Realbench will be putting more stress on the processor than the other two.

What are you cooling the CPU with. When overclocked that 9900k will run hot! Have you done any temp monitoring while running the stress tests/benchmarks you are able to run? What are your maximum core temps?

Also, Realbench can be very glitchy in my experience but when it works it's an excellent real world stress test. If you remove the overclock and put the system back to stock frequency and voltage, does Realbench still blue screen?

If you are relying on AIDA and XTU for testing the stability of your overclock I would recommend running them overnight. They are not real demanding stress tests.

Hi, I am using Radeon VII as my GPU. I have a custom liquid cooling and as of now my system is on an open bench, so I am sure heat is not an issue, ambient temps are around 18-19 Degrees Celsius.

I did do temp monitoring, while running the OCCT for about 1 HR, Intel Burn Test on (Maximum), Aida 64.. Very rarely did the temps even cross 80, though it touched 90 for a brief second or two on one of the cores but rest were nearly averaging between 72-78the whole time.

I stopped getting the blue screen on R20 as soon as I got the AVX offset value to 2 and everything was just working perfectly fine. with no issues whatsoever.

Realbench works when on stock, its just that it doesn't work with my values even though all other stress test pass with no issues whatsoever...


Rgds
Sajeev

- - - Auto-Merged Double Post - - -

Blue screens can mean ANYTHING. Without knowing the error code, it is impossible to pinpoint just from getting a BSOD. What error is it?

What temps in stress testing?

That said, with 64GB of RAM that is a lot for the IMC to handle. If adding a bit more vcore doesnt help, try adding SA voltage... 1.3V or so and see if that helps. To test memory, run p95 BLEND and se if it passes. Small fft is mostly for cpu... blend is more memory.

The Blue screen error is something like WINDOWS UNKNOWN ERROR, system tries to recover and directly restarts to windows.

As said above my temps are well within limits as the system is on custom water coolings and ambient temps are 18-19 Degrees Celsius.

By Vcore do you mean the Cpu core/cache voltage? I'm already at 1.29, will try and see if SA Voltage is the issue though, it's on 1.25 as of now..


Best Regards
Sajeev

 

[trents]

Hi, I am using Radeon VII as my GPU. I have a custom liquid cooling and as of now my system is on an open bench, so I am sure heat is not an issue, ambient temps are around 18-19 Degrees Celsius.

I did do temp monitoring, while running the OCCT for about 1 HR, Intel Burn Test on (Maximum), Aida 64.. Very rarely did the temps even cross 80, though it touched 90 for a brief second or two on one of the cores but rest were nearly averaging between 72-78the whole time.

I stopped getting the blue screen on R20 as soon as I got the AVX offset value to 2 and everything was just working perfectly fine. with no issues whatsoever.

Realbench works when on stock, its just that it doesn't work with my values even though all other stress test pass with no issues whatsoever...


Rgds
Sajeev

- - - Auto-Merged Double Post - - -





The Blue screen error is something like WINDOWS UNKNOWN ERROR, system tries to recover and directly restarts to windows.

As said above my temps are well within limits as the system is on custom water coolings and ambient temps are 18-19 Degrees Celsius.

By Vcore do you mean the Cpu core/cache voltage? I'm already at 1.29, will try and see if SA Voltage is the issue though, it's on 1.25 as of now..


Best Regards
Sajeev

If you can pass all these stress tests when either on stock values or with an AVX offset of 2 then there is nothing wrong with the hardware but this points instead to an overclock that is either too aggressive or whose settings need adjustment. I would try giving your RAM and your IMC (Integrated Memory Controller) a bump in voltage.

Also, check your CPU cache multiplier and make sure it's multiplier is not too high. Some motherboards will overclock the cache along with the CPU cores and cause instability because the cache cannot overclock as high as the cores. In that case, give the cache a fixed multiplier value equal to stock or slightly more.

 

[sajeev3105]

Just wanted to know if this could actually be ram related as I had bought two sets of these rams and not the one which has 4 dimes in a pack? Though both these sets comprising of 2 Dimms each have the same clock speed and frequency and are of the same series and brand..

Rgds
Sajeev

 

[trents]

I don't think there is a problem with adding a second kit of RAM as long as they are the same as the ones originally installed.

But realize this:
1. Four sticks of RAM places more stress on the IMC than does two sticks, even when they add up to the same total amount.
2. 64 gb of RAM is a lot. The higher the total amount of RAM is, the more stress it puts on the IMC. You might try experimenting with removing two sticks and see if you can pass some of those stress tests which you were failing before when overclocked. If so, then that points to a need for more voltage to be given to the IMC and possibly to the RAM itself.

 

[wingman99]

Just wanted to know if this could actually be ram related as I had bought two sets of these rams and not the one which has 4 dimes in a pack? Though both these sets comprising of 2 Dimms each have the same clock speed and frequency and are of the same series and brand..

Rgds
Sajeev

Have you tried with the processor running default settings and only changing the memory to run XMP? If the memory passes testing the ram is not the problem, it is the processor overclock.

 

[sajeev3105]

If you can pass all these stress tests when either on stock values or with an AVX offset of 2 then there is nothing wrong with the hardware but this points instead to an overclock that is either too aggressive or whose settings need adjustment. I would try giving your RAM and your IMC (Integrated Memory Controller) a bump in voltage.

Also, check your CPU cache multiplier and make sure it's multiplier is not too high. Some motherboards will overclock the cache along with the CPU cores and cause instability because the cache cannot overclock as high as the cores. In that case, give the cache a fixed multiplier value equal to stock or slightly more.

How far do you think I should bump the voltage on RAM or how much do you think is actually within safe limits? It is currently at 1.35. Never done this so I am not sure about the values, also a little scared off bricking the ram, thus an input from your side will be highly helpful. Same with IMC too.

Just FYI my fixed multiplier is on 50 and cache max and min is set to 43.


Rgds
Sajeev

 

[EarthDog]

Sajeev.... I wouldnt go over 1.35 vcore as your temps likely will be too high (if you peaked at 90 at 1.29V already). For system agent, I'd trump to 1.3v or so.

Yep, my CPU-Z says about the same exact thing. Never understood why, just always used manual vdimm after I noticed.

View attachment 204911

xmp changes the timings and voltage automatically. No idea why you are setting anything manually because of 1.2v thing.. It's how Intel has worked, even AMD, for generations. Boots at JEDEC and needs xmp set which changes voltage/speed/timings automatically.

Anyway a bit OT...so I digress.

 

[Johan45]

Have you tried just adding some core/cache voltage? You have headroom there as long as you keep it cool. Try 1.31 V or a bit more if needed

 

[DaPoets]

What's your VRM temp hitting?

 

[wingman99]

My VRM temperature runs 117c with a i7 9700k.

 

[Johan45]

I highly doubt his board is having VRM issues, his is up to the job!