H100 inferior to an A70? (Newbie in need)

[Rocky Graziano]

Hi there, I'm new to the forum so I thought best I'd get the pleasantries out of the way first!

I'm from the UK and after saving for a while, I recently put together the following set-up;

P8Z68V PRO
2600K
H100
16GB of Vengeance RAM
2x GTX 580s in SLI
AX1200 PSU

Now I'm here essentially as I'm having a few minor issues with relation to overclocking. A few weeks ago I managed to run my CPU at 4.7 ghz for two hours at 1.370 volts under a stress test, and I've had that as my manual overclock until recently.

For the original test I used the AID64 stress test and it didn't fail, I simply brought it to a close and tried to go for a higher clock speed.

After a few attempts at 4.9 and then 4.8 and a collection of BSODs I settled for the original 4.7 at 1.370 volts.

Now though, I'm having problems with the CPU, I've been getting BSODs while running video games, and it no longer manages the stress test at the aforementioned figures.

I bumped it up to 1.375 and it still failed after a short while, could anyone point towards why it no longer seems comfortable at that voltage and clockspeed? I'm going to run 1.380 later, but nonetheless it's frustrating that the goalposts seem to have been moved for no apparent reason.

Does anyone with more experience than myself know what is the issue there?

Also, slightly related, when I first put the rig together I ran the CPU with a Corsair A70 heatsink. At the 4.7 ghz by 1.370 volts I mentioned previously I was getting temperatures of between mid 50s to very early 60s. There was very little fluctuation, and it seemed consistent, it was a very good stable temperature.

By virtue of the A70 being unable to fit in my Cosmos S with the side panel closed, I bought a H100 to replace it.

I expected a small decrease in temperatures, but running the same AID64 stress test at now 1.375 volts (as the 1.370 fails almost immediately) and 4.7 ghz, I'm seeing temperatures that are in fact higher than the A70's and a much greater discrepancy between the highest temps and the lowest. The temps can briefly reach the low 50s and then pop up to 63 or 64 degrees, and they continue to do so during the duration of the test so the graph develops a zig zag effect. The average is probably around 59-62, but the fluctuation is puzzling.

Could anyone provide an explanation for this? What temperatures should I be expecting to see? And is it just a possibility of it being poorly fitted to the CPU? I used AS5 thermal paste and spread it out evenly with a card on the processor, although I may have used too much, I don't know - could that be a possibility?

Thank you for any help. Also any tips and help to reach a stable overclock would be greatly appreciated, as I'm a relative newbie in this respect.

Cheers.

 

[bmwbaxter]

sounds like a bad mount on the h100. temps should be much better. also are you sure your volts are 3.70 and 3.75? thats a little impossible...

 

[Rocky Graziano]

Ah, my mistake. Haha, excuse me, 1.370 degrees.

I think I'll have a quick edit.

 

[Conumdrum]

There was some recall issues on the H80 and H100, something about the pump controls etc.

 

[bmwbaxter]

do you have pll overvoltage enabled? that might solve the stability issue.

 

[Rocky Graziano]

I looked up re the recall and it appears they were relating to the manual controls for the fan's mode of speed?

I'm free of that issue.

I'm currently running a stress test at 4.7 ghz and have been doing so for approaching 50 minutes at 1.375 volts. And there have been no issues. Really puzzling, I tried this numerous times before, yesterday no less, and it failed fairly quickly. Could someone explain this?

Also, I'm not entirely sure what the pll overvoltage is, but according to the the CPUID the voltage flicks between 1.376 and 1.384 - safe to assume it's enabled?

Also here is a graph of my temps running the current stress test;



Notice how the temps are all over the place. Some cores drop to 50, while others spike up to 64. And it's consistently spiking, rising and falling . Also note that this room is around 2 degrees cooler than usual.

Can anyone explain what's going on here?

Thanks.

 

[jason4207]

Could anyone provide an explanation for this? What temperatures should I be expecting to see? And is it just a possibility of it being poorly fitted to the CPU? I used AS5 thermal paste and spread it out evenly with a card on the processor, although I may have used too much, I don't know - could that be a possibility?

The H100 CPU block should come w/ paste pre-applied. If you didn't clean it off, and also used AS5 on the CPU then you probably have way too much paste on there.

You might want to try the dot method rather than the spread method to apply paste if you attempt to remount it. Make sure to clean both the block and CPU real well w/ 91+% isopropyl alcohol to prep the surfaces before reapplying new paste.

Also make sure the pump and fans are running at full speed. Depending how you have everything hooked up, and what your BIOS fan speed options are set to you could be limiting yourself.

I also recommend using more than 1 stress test before calling it stable for 24/7 use. Some kind of quick test like IBT followed by 8+hr runs of both P95 small and P95 blend will give you a better picture of true stability.

 

[Rocky Graziano]

The H100 CPU block should come w/ paste pre-applied. If you didn't clean it off, and also used AS5 on the CPU then you probably have way too much paste on there.

You might want to try the dot method rather than the spread method to apply paste if you attempt to remount it. Make sure to clean both the block and CPU real well w/ 91+% isopropyl alcohol to prep the surfaces before reapplying new paste.

Also make sure the pump and fans are running at full speed. Depending how you have everything hooked up, and what your BIOS fan speed options are set to you could be limiting yourself.

I also recommend using more than 1 stress test before calling it stable for 24/7 use. Some kind of quick test like IBT followed by 8+hr runs of both P95 small and P95 blend will give you a better picture of true stability.

Thanks for the response.

I cleaned off the stock paste and applied the AS5, I only mentioned the possibility of there being too much paste applied as a potential theory as to why the temps I'm getting are no different, if not worse than the A70's.

The paste was applied evenly, and smoothly, I could have used too much, but I don't know if I did, just a theory as to why I'm getting the temps I am - do you know what temps I should be expecting?

I'm running the fans on medium mode in a push/exhaust set-up, so does anyone know if my temps are correct, or if they are higher than they should be?

I'll be pretty bummed out if I replaced a heatsink with one 3 times as expensive and there was no performance increase, hell even a marginal decrease it appears!

 

[Rocky Graziano]

I've just stopped the AID 64 stress test after 3 hours and it ran fine. Puzzling as it failed yesterday.

But another query I have is why is the CPUID currently displaying the vcore as 1.384v now I'm at idle and the clock speed is at its idle 1.6 ghz?

The vcore I set in the BIOS was actually 1.375 as well.

Is this an error, or is the CPUID correct? And if so, how do I change it to drop to a lower voltage while running in idle?

Thanks for any help.

 

[bmwbaxter]

I've just stopped the AID 64 stress test after 3 hours and it ran fine. Puzzling as it failed yesterday.

But another query I have is why is the CPUID currently displaying the vcore as 1.384v now I'm at idle and the clock speed is at its idle 1.6 ghz?

The vcore I set in the BIOS was actually 1.375 as well.

Is this an error, or is the CPUID correct? And if so, how do I change it to drop to a lower voltage while running in idle?

Thanks for any help.

try restarting cpuz

 

[Serph]

The H100 comes with Shinet-su grade paste, why would you clean it off and use AS5?

 

[jason4207]

Thanks for the response.

I cleaned off the stock paste and applied the AS5, I only mentioned the possibility of there being too much paste applied as a potential theory as to why the temps I'm getting are no different, if not worse than the A70's.

The paste was applied evenly, and smoothly, I could have used too much, but I don't know if I did, just a theory as to why I'm getting the temps I am - do you know what temps I should be expecting?

I'm running the fans on medium mode in a push/exhaust set-up, so does anyone know if my temps are correct, or if they are higher than they should be?

I'll be pretty bummed out if I replaced a heatsink with one 3 times as expensive and there was no performance increase, hell even a marginal decrease it appears!

The spread method usually results in too much paste ime. Try remounting and using the dot method. Let the pressure of the CPU block spread the paste for you. You avoid potential air bubbles this way.

I'm not sure what temps you should be getting, but if you're not getting better results that the A70 then something isn't right.

Are you sure your fans and pump are getting 12v from the mobo? Did you check the fan settings in the BIOS? Or are the fans and pump connected directly to your PSU?

Also, how is your case cooling? If it's constricted then better CPU cooling can be negated by a lack of fresh cool air.

I've just stopped the AID 64 stress test after 3 hours and it ran fine. Puzzling as it failed yesterday.

But another query I have is why is the CPUID currently displaying the vcore as 1.384v now I'm at idle and the clock speed is at its idle 1.6 ghz?

The vcore I set in the BIOS was actually 1.375 as well.

Is this an error, or is the CPUID correct? And if so, how do I change it to drop to a lower voltage while running in idle?

Thanks for any help.

If you set the voltage manually you might not be able to get the voltage to drop. That's how it is on my older Q9550 rig. C-States don't work when voltage is set manually, but speed-step does.

If you use the offset method to set your voltage you can get the voltage to change w/ the clockspeed.

The H100 comes with Shinet-su grade paste, why would you clean it off and use AS5?

+1, you should have left the stock paste on there...it's good stuff, and applied perfectly already.

 

[Rocky Graziano]

The room was cool when I ran the test, chilly even, and there was no improvement on the A70.

The fans are all connected properly as is the pump (in terms of motherboard recognition and physically), and the case has very good circulation. In fact, when I was running the test earlier I had neither of the gtx 580s inside the case.

It must be the paste or the fitting.

To answer the previous questions, I removed the original paste as I made a mess of mounting on the first attempt, so I was left with no choice - was that paste really much better than AS5 anyway? As far as I knew, there wasn't much out there more efficient then AS5, no?

Just to note, 10 minutes ago I added the 580s to the rig and started up the same identical stress test - it failed almost immediately. The ****? Can anyone help me resolve this?

Also, I'd appreciate it if someone could talk me through how to keep my idle vcore low and to only bump up only when necessary.

Thanks.

 

[RJARRRPCGP]

I would use IBT, like jason4207 posted or LinX.

 

[Serph]

Shinetsu is ranked 2nd? I believe with Indigo Extreme being the top. The H100 comes with Dow Corning TIM which is comparable with the Shinetsu.

http://forum.corsair.com/v3/showthread.php?t=97551

If you can, I'd get some mx-3 and reapply the paste as AS5
isn't that great of a performer anymore.

 

[Rocky Graziano]

I was literally just posting a response earlier where I was jtyping about how I'd again started up the stress test and that it'd been running for 25 minutes without any problems and then... you guessed it.

Why is not playing ball now that I've added the 580s? Coincidence, or was the oc never stable in the first place, despite running for 3 hours earlier today?

Could anyone give me some help in that respect?

Tips and pointers to reach a stable oc?

Also, when I was fiddling around the case earlier, I accidentally caught and trapped one of the 580's fans (while it was running) while I was attempting to tidy a few cables, and now one of the card's fans runs slightly slower than the other.

Considering the card is a Lightning Extreme, and I've spunked a considerable amount of cash on it - what should I do?

And I know, what a dumbass...

*Sigh*

 

[jason4207]

AS5 was the bees-knees...6yrs ago. That's why you see lots of recommendations for it on web searches. It's still great stuff, but for the same $ or less you can do better now.

Can you verify what rpm readings you are getting from the pump and fans in your BIOS?

Adding the 580's will definitely add more heat to the case, and thus the H100 will be pushing warmer air thru the rad. It won't be as effective at cooling your CPU, so you'll have to increase fan speed, or try more vcore to compensate.

Try doing a test we are more familiar with. I don't think most people here are familiar w/ AID64 including myself.

I would work on getting it stable w/ manual vcore for now, and once you get that then we can try the offset method which will give you lower vcore at idle.

Make sure you have the latest BIOS for your mobo, and then also enable PLL Overvoltage in the BIOS. You should be able to find the setting if you dig around.

Sorry I can't be more specific, but my time w/ Sandy Bridge has only been a few hours on a friends rig several months ago.

Hopefully others can chime in w/ more specific advise and correct any errors in my SB knowledge.

 

[jason4207]

I was literally just posting a response earlier where I was jtyping about how I'd again started up the stress test and that it'd been running for 25 minutes without any problems and then... you guessed it.

Why is not playing ball now that I've added the 580s? Coincidence, or was the oc never stable in the first place, despite running for 3 hours earlier today?

Could anyone give me some help in that respect?

Tips and pointers to reach a stable oc?

Also, when I was fiddling around the case earlier, I accidentally caught and trapped one of the 580's fans (while it was running) while I was attempting to tidy a few cables, and now one of the card's fans runs slightly slower than the other.

Considering the card is a Lightning Extreme, and I've spunked a considerable amount of cash on it - what should I do?

And I know, what a dumbass...

*Sigh*

If you didn't break a fan blade then the fan should be fine. Look at the fan speed in eVGA Precision or MSI Afterburner to verify the rpms.

Don't reach in the case while it's running! I'm guilty of this myself and always kick myself later. I've broken fan blades, seen sparks, and cut a deep gash in a finger (some fan blades are strong!).

 

[Rocky Graziano]

Thanks again mate.

Funny you should mention the review stuff, I had no idea of the paste's age, saw a few good recommendations, and thought - that'll do! ;p

Never realised how long it's been around for. That'll teach me.

With regards to the heat generated from the 580s being pushed through the radiator, surely that shouldn't cause a BSOD if the temps remain reasonable, which they did. I monitored them closely on the last stress test which lasted 25 minutes and they were around mid 60s - that can't be responsible for the crash can it?

Btw, I'm positive the rpm of the fans on the radiator is normal, on all settings, I can't stand the noise of 'performance' so I'm holding out on medium.

Anyway, knowing the voltage I had to put through the processor to reach 4.7 volts - do you think it'd be better just to bite the bullet and settle for 4.6?

And I had a fiddle attempting to oc with offset just now and somehow managed to shoot the vcore up to just over 1.5 volts under the load of the stress test, despite adjusting the offset in the BIOS to make the final vcore 1.34. Not ideal! Lol.

And re the 580, I can see the fan is spinning a little slower than the one adjacent to it, particularly as the card's LED light makes it easier to recognise in comparison with the other. I'm sure the rpm on the afterburner will say they're running normally but I can see one of the fans isn't, and its probably due to some physical damage in the head of the fan - which I presume the software wouldn't be able to recognise.

Could be a weakened spring, screw I've no idea but it's running more feebly than it should be. Visibly so.

Could that effect cooling / performance in any significant way?

I'm a bit gutted that I've damaged it considering it's brand new and not least considering the cost!

Thanks for your help though mate, I'm learning, slowly but surely, nonetheless help is needed and greatly appreciated.

Cheers

 

[Serph]

If you can supply couple screenies of your bios, it'll help in determining the cause of the issue.