• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Is 7zip a valid CPU stability test?

E

evilsofa

[H]F Junkie
Joined
Jan 1, 2007
Messages
10,077
I'm trying to figure out if my i5-2500K is being throttled by my Intel DP67DE B3 motherboard.

In this thread, (Link broken because HardOCP doesn't like that site) it is reported that "The OCing bottleneck for this board is the 'naked' MOSFETs that provide voltage for the CPU. If they get too hot, BAM you're down to the non-turbo multiplier of x33."

So I switched from using a CoolerMaster Hyper 212+ cooler to using the stock Intel cooler that came with the 2500K, because apparently downdraft coolers will cool the naked MOSFETs better. I set the BIOS up per his instructions later in that thread. At x43, I get a BSOD at startup; at x42, Intel Burn Test will run 18 times before it throttles to x33 (sure enough, that's the MOSFETs getting too hot); at x41, IBT ran 50 times without throttling down once, with a max CPU temp of 89C, which is a bit hot.

The behavior I see that is puzzling me is that when I use 7zip to compress or decompress large files (I've been using a 4GB .mkv as a test file), CPU-Z reports the CPU throttling down to x16 repeatedly; it'll spend some time at whatever multiplier I set, then it will spend some time at x16, then go back up again. I never see x33. It's clearly doing the task MUCH slower when it's throttled to x16. This happens even at x34. Is this 7zip just not giving the CPU enough work to do?

When googling, I saw passing mention in a HardOCP thread of "the 7-zip benchmark" being more difficult than IBT, but that seems strange to me because IBT loads all four cores while 7-zip only loads one.

I'm doing all this to make a decision about whether or not to ditch the Intel board, get an ASUS 1155 board and be able to use the CM Hyper 212+ cooler again for much lower temps and maybe an OC that's stable higher than x41.
 
Last edited:
The down draft is not going to be cooling them that much when you think about how much extra heat the CPU is dealing with and going into the socket and surrounding parts (MOSFETs), and at those temps, the BSOD and throttling at higher multis is going to be CPU temp related, as you are hitting almost 90C at the lowest stable OC. If you had a nice big cooler that was a down draft it might be worth using, but you are going backwards here going to the stock HS. If you want to cool them, get some HS's to put on them
 
Over the years I have tried most known stability tests and have found OCCT to be the best and catch the most errors others wont.
 
I will put the Hyper 212+ back on and see what kind of clearances I have on the MOSFETs for heatsinks, and how stability and temps change now that I know how OCing this board works. I'll also try disabling speedstep before and after, and OCCT before and after.

After that, I'll decide if I want to muck about with buying some mosfet heatsinks (if they will even fit), or just buy an ASUS board for $78 or so that will hopefully just go to x44 without all the fiddling with tiny things.
 
Intel Burn Test or LinX will seriously test your system.

Prime95 is decent.

However, I've found that OCCT's load is closer to what I need for rendering, compression etc... AIDA64 is also pretty decent.
 
ImperfectLink said:
Intel Burn Test or LinX will seriously test your system.

Prime95 is decent.

However, I've found that OCCT's load is closer to what I need for rendering, compression etc... AIDA64 is also pretty decent.
Click to expand...

IBT is good for quick testing and max heat, OCCT uses linpack and something similar to Prime for its CPU testing, its quick also but both will miss allot of things that Prime will catch, Prime stresses everything quite well and I have seen many OC's that pass 20+ passes of IBT and OCCT fail before 5 hours of Prime.

To me, I use IBT/OCCT right off the bat to test if the OC will fail, if it passes those which often take an hour max I will move onto a 24 hour Prime run and make small tweaks as needed to the OC if it's not Prime stable.
 
Okay, I got the Hyper 212+ back on, and the case is now closed and in normal upright position. I now get very consistent behavior. CPU temps are great, never exceeding 57C while running IBT. But it throttles to x33 after about 10 runs of IBT, no matter what multiplier I set - even x34.

As far as I can tell, I can't overclock without enabling Speedstep. Disabling Speedstep disables the part of the BIOS where you OC; enabling the part of BIOS where you OC enables Speedstep.

So, I guess I have a motherboard that's designed to not be able to run the 2500K at spec. Should I grab a refurb ASUS P8P67-M Pro for $73, or is there a better choice?
 
Sounds like you might be mixing things up, I don't think it is throttling, as that has to do with CPU temps, if the MOSFETs over heat etc the system will become unstable and hang/BSOD. What it sounds like is speedstep acting like normal and reducing the multi when not under load. Open CPU-Z and CoreTemp, when running IBT and see if the multi drops when the cores are at 100% load, if the multi is dropping when not under 100% load then that is speedstep doing its job.
 
What I do is start CPU-Z and RealTemp, and then run Intel Burn Test at Standard for 20 runs.

With the Hyper 212+, CPU temps never exceed 57C even with all four cores at full load.

The multi is dropping to x33 after about 10 runs of IBT, while all four cores are under 100% load.
 
First off, I will chime in with others in saying that there almost no conditions where the standard intel HSF is going to perform better than the 212, as the amount of air moved by the stock HSF is just sad.

What voltage are you running at while overclocked? Can we maybe get some screenshots of when you're running IBT, of temps/cpu-z at idle, during the first runs where you say its acting correctly, and then once the multi drops?

Skimming through the thread you linked (and not searching more into it), if that's truly the case, it sounds like a pain in the ass board, and I'd run away from it if you're serious about oc'ing at all. That being said, if you're intent on keeping it, then the couple people posting in that thread seemed to think that the multi-drop was just another symptom for stability, so maybe you're just not giving it enough volts to stabilize? But then again, it appears as if too much voltage is whats causing you to have the multi-drop in the first place, so its practically a catch-22. Bottom line, it sounds like a quirky mb and one you'll have to find people with hands-on experience with that specific board to get a possible solution.
 
Last edited:
I've been running with a 1.1625V "Processor Voltage Override", because that is what the guy in the first thread I linked to used. When I turned off the voltage override, nothing changed.

Looking over this thread, now that I understand things better, it looks like ultimately the answer is "Intel does not allow unlimited time Turbo multiplier over clocking on their non-extreme mother boards" and there's nothing I can do about it but get another motherboard.
 
Not sure about the specifics of that board, but in general, if the voltage is really being set to 1.16 the chip isn't going very far. I want to say that the stock load voltage is ~1.2v, so if CPU-z is showing 1.16 with that override setting when you're burning it in, that chip wouldn't overclock much on any mb. To get anywhere decent you're going to end up between 1.3-1.4 as a rough guideline for the CPU itself. I apologize if I'm misunderstanding something you said.

Edit: and yes, a cursory search about that board seems to indicate its just not worth trying to use it as a oc'ing platform as it treats the "turbo" multipliers as a transitory and short term boost. I'd return it if at possible just for the sake of ease.
 
Last edited:
I've done what I should have done a year ago and ordered an Asus P8P67 Pro (Rev 3.1).
 
Good luck with it. I've seen a few people have some complaints on the Asus p8p67 line, but most tend to be pretty happy. I personally used that same exact board and couldn't have had an easier time overclocking - was one of the better boards I've tinkered with in recent memory.
 
Received the P8P67 Pro Rev 3.1 and have it installed now. After a bunch of blundering around and trying to manually overclock it, I just used the high performance setting on the EZmode UEFI. It overclocked the CPU multiplier to x42 and raised the BCLK to 103Hz. According to CPU-Z and RealTemp, under Prime95 load it clocks up to 4327 Mhz with core voltage between 1.288v and 1.304v and stays there with temps at 68C. I feel no need to go higher than that, because I noticed the CPU fan gets noisy when the ASUS AI Suite tries to overclock it to 5.1 GHz (which was a silly setting because the CPU hit TJMax under full load).

And to finally answer the question in the thread title, the CPU still briefly clocks down to x16 many times while using 7-Zip. Apparently 7-Zip is failing to keep the CPU fed with work, so it is not a valid test.
 
You must log in or register to reply here.
Back
Top