Personal computing discussed

Moderators: Flying Fox, morphine

 
jackbomb
Gerbil XP
Topic Author
Posts: 344
Joined: Tue Aug 12, 2008 10:25 pm

Re: What happened to my overclock?

Thu Dec 21, 2017 3:07 pm

Well, after a few days of using it, I still can't say that Win10 is much better now than it was at launch. Just switched back to Win7, fully expecting the instability @ 4.4GHz to return, but so far it's been stable. It just installed something like 260 updates without a hitch. I'm really wondering what caused the hanging and BSOD on that ONE day last week.
Like a good neighbor jackbomb is there.
 
Kougar
Minister of Gerbil Affairs
Posts: 2278
Joined: Tue Dec 02, 2008 2:12 am
Location: Texas

Re: What happened to my overclock?

Fri Dec 22, 2017 11:13 pm

Could be guessing forever. I'd just run 24 hours of blend Prime95 and see if all the threads were error free . I've had enough systems that where that 0.01% instability was such an issue that just running it overnight isn't enough in my opinion. If it can pass that then the root cause is someplace else.
 
mcarson09
Gerbil
Posts: 31
Joined: Sat Aug 01, 2015 5:34 pm

Re: What happened to my overclock?

Sun Dec 24, 2017 2:38 am

Kougar wrote:
Could be guessing forever. I'd just run 24 hours of blend Prime95 and see if all the threads were error free . I've had enough systems that where that 0.01% instability was such an issue that just running it overnight isn't enough in my opinion. If it can pass that then the root cause is someplace else.


A Prime95 run in any form (blend/small FFTs/large FFTs) doesn't match a linpack stable system in my book. OCCT is also a better stablizer than prime too. If one wants to run handbrake with an OC and not have errors in their encodings use Linpack to judge system stability.
Ignorance is not an excuse.
 
jackbomb
Gerbil XP
Topic Author
Posts: 344
Joined: Tue Aug 12, 2008 10:25 pm

Re: What happened to my overclock?

Wed Dec 27, 2017 12:59 am

Well, it started freezing again today. Just like the first time, it froze while playing a video file in MPC-HC. Win7 does indeed seem a little more sensitive to overclocking than Win10.
So 4.3GHz at 1.15v it is, then.
Like a good neighbor jackbomb is there.
 
Kougar
Minister of Gerbil Affairs
Posts: 2278
Joined: Tue Dec 02, 2008 2:12 am
Location: Texas

Re: What happened to my overclock?

Fri Dec 29, 2017 3:04 am

mcarson09 wrote:
A Prime95 run in any form (blend/small FFTs/large FFTs) doesn't match a linpack stable system in my book. OCCT is also a better stablizer than prime too. If one wants to run handbrake with an OC and not have errors in their encodings use Linpack to judge system stability.


It's been awhile since I've used Linpack, but blend runs have been much more effective for me at finding trace memory system instabilities, whether it's the CPU memory controller, CPU cache, or RAM settings. Every haswell system I've built has been sensitive about memory settings after any OCing.

Linepack focused too much on cpu-core testing and wouldn't find memory system errors nearly as quickly. Particularly on one occasion, whereupon after many months when my 4770K magically began to go unstable with 1600mhz 1.35v DDR3, but would test stable if I bumped the RAM to 1.5v. Crucial claimed the kit was Haswell rated for 1.35v and that was what the first XMP profile defaulted to.
 
synthtel2
Gold subscriber
Gerbil Elite
Posts: 858
Joined: Mon Nov 16, 2015 10:30 am

Re: What happened to my overclock?

Fri Dec 29, 2017 4:35 pm

IME, P95/Linpack/[whatever] are all going to be within 20mV of each other, and which one's in front is down to the CPU uarch and luck (but P95 seems to be one of the most consistently stressful ones). Most of the time (when you want to leave a decent margin anyway and aren't pushing ludicrous voltage), picking one decent stress test and leaving a 30mV margin seems a fine way to go. If the OC is aggressive enough that a 30mV margin is unacceptable, relying on a single stress test is probably a bad idea.

P95 does have a couple of handy advantages. In addition to what Kougar said, it's a more scalable thermal load than most; with FFT sizes of 2M+, it resembles the high end of normal workloads, rather than being thermally off in the stratosphere somewhere like Linpack. If you're thermally limited and trust the CPU's throttling to handle 0.01% worst-case scenarios like Linpack, this more accurate accounting of the typical stress on the CPU can result in a much higher OC. IME, it isn't noticably worse at finding instability in this low-temp mode.
 
Ummagumma
Gerbil
Posts: 40
Joined: Fri May 27, 2016 9:18 pm

Re: What happened to my overclock?

Fri Dec 29, 2017 6:20 pm

Waco wrote:
mcarson09 wrote:
Waco wrote:
Servers won't, but he's talking about desktop boards unless I'm mistaken.


Certain 2 Processors have an overclocking feature. The asus z10pe-d8 ws and z10pe-d16 ws being amoung those boards. Supermicro has what is calls a hyper-speed setting which is indirect control OCing.

I don't consider those server boards, and further, they don't accept non-Xeon CPUs unless I'm mistaken.

Supermicro is a blacklisted vendor at many institutions for good reason.


Can you enlighten all of us on what you are really saying about Supermicro? What is that "good reason", or is that something you can't talk about in public?

My own experience with Supermicro specs goes like this:
    - If you follow Supermicro's published list of supported CPUs, you can have confidence they will work (if the CPU isn't "broken" in some way). Stray from that list and you are on your own; no guarantees.
    - If you follow Supermicro's published list of supported RAM, you can have confidence they will work (if the RAM isn't "broken" in some way). If you stray "slightly" from their list, like same RAM chip manufacturer & "format but different brand of RAM "stick" vendor, then your RAM will probably still work. If you go "way off the mark", then you are on your own; no guarantees.
    - The "supported" operating systems list for any given board will work. Pick something not on that list and you are on your own; no guarantees.

I agree with other poster's comments on Supermicro server boards; they do not appear to overclock (at least the boards I have seen). I have not used any of their "desktop" boards.

I think people buy "server class" boards for stability, longevity of support from Supermicro (board models tend to be around longer than "desktop" boards), and ECC support (when that is needed).

If you overclock or like to experiment with different combinations of hardware and possibly software, then Supermicro "server class" boards really aren't for you.
I used to do networking & network security for a living. Now I just do it for fun, but I still take it seriously.

Who is online

Users browsing this forum: No registered users and 1 guest