Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

[lukevega]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

Well I've loaded Windows fine with just vcore changed to 1.4V and VTT to 1.36V and am about to run 5 loops of LinX at 2048MB:



I'll probably head to bed afterwards but if it doesn't BSOD then I'll post a pass/fail screenshot.

Backup will be checked tomorrow morning :)

[lukevega]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

Well that bluescreened after a matter of seconds...

Maybe LinX will run too hot to test with from now on?

I've attached the HWiNFO file just in case it shows anything useful:

Zippyshare.com - 4.0 LinX.CSV

I'll try running prime blend tomorrow and make any further changes if you have time to recommend any.

Goodnight from England!

[profJim]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

The HWiNFO CSV log file did not show anything out of the ordinary.
Within about 12 seconds of starting the log file, cpu temp rose to 80^o(C) and the individual cpu core temps rose to ~ 65^o(C).

I suggest that you try a more modest o/c to about 3.85GHz (9.5 x 405MHz) or 3.90GHz (9.5 x 411MHz).

Once the 3.90GHz setting is stable, you might need to increase each additional o/c using ~50MHz increments.

The other alternative that I do not recommend is to try using much larger voltage increases as necessary.

Sorry, you're past the point of relatively easy 200MHz cpu speed increases.

[profJim]

Re: Macrium Reflect Backup Software

I got the current Reflect software properly installed and working after I first disabled my BitDefender antivirus software.

They have removed the Verify backup function while you are creating the backup image files.

When you go to the Restore an image window, there is an option where you can verify the integrity of any of your Reflect image files.

The freeware Reflect software lets you download WinPE 3, 4, or 5 to create a bootable optical disk.

Windows Preinstallation Environment - Wikipedia has a brief description of the different WinPE versions.


The following Reflect screen-shots show several windows for my 51GB boot partition backup to my 3TB usb 3.0 external hard drive:

note: Reflect will prompt you for the INF file to be loaded if your usb 3.0 add-in card has its own, non-Intel drivers.
To make life easier, I copied the usb 3.0 INF file(s) to a new folder in the root directory of my C: partition.







It took just over 12 minutes to verify the full disk image.

[lukevega]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

I don't know if this will give any diagnostic value (where can I find a guide that tells me how to interpret the results of failures from prime95 and LinX?!?!) but I tried LinX at 3.85GHz, keeping vcore at 1.4V and VTT at 1.36V with no luck:



I have no idea how to tell what changes I should try other than your guidance Jim!

[SmOgER]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

Ok so here is the drill.

PC is rock stable at X5460 SLBBA 4Ghz (421*9.5) 1.3v vcore with RAM at ~1360Mhz 7-7-7-20.

At 4.2Ghz (447*9.5) CPU itself is rock stable with 1.4v (could drop this a little possibly), but RAM is semi-stable only at ~1130Mhz 9-9-9-24), and even then it can pass 150% coverage of HCI Memtest, but triggers memory related BSODS when there is high RAM activity (games/video editing/intensive multitasking)... vDimm and NB overvolting doesn't seem to help, tried NB as high as 1.5V (at 421FSB it's perfectly stable under 1.3v) and vDimm as high as +0.3v (1.8v DDR3).

At this point I don't even know what else I can do... Adjusting subtimings (and PL) doesn't help much. At best it can slightly reduce the frequency of BSODs by drastically reducing the RAM bandwidth. The only thing I can think of which could be causing those issues is maybe mobo is incorrectly adjusting some or one of the timings which can't be manually overridden in bios. But why would it do that only when FSB is above 421FSB...

Any ideas?

PS. Tried several memtables extracted from another ASUS LGA775 DDR3 mobos, didn't seem to make much difference.

[lukevega]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

I'm doing some small investigations to try and improve my overclock.

I booted up at 4.0GHz (9.5x422) with all the settings the same as my stable 3.8 overclock except MCH Core which I increased one notch to 1.26V.
I ran prime95 small FTTs and almost instantly, worker 2 failed.
I increased the vcore one notch at a time up to 1.45V (increasing VTT when needed to keep the distance to vcore clost to 0.05) and repeated the small FTT test each time.
Worker 2 never made it past Test 2.

I've now relaxed 'Performance Enhance' from Turbo to Standard and decreased vcore and VTT back and am increasing them again, one notch at a time to see what happens. As I'm typing, worker 2 just failed in 0 minutes at vcore= 1.38750, VTT= 1.34...

Please guys don't hesitate to tell me if I'm doing something wrong or have missed anything!

[SmOgER]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

What's your RAM running at?
Does it fail on prime95 small fft just as quick? If not, your problem is most probably with RAM or NB/MCH, if that's the case try increasing MCH/NB to 1.3v and maybe give a RAM 0.05v above stock voltage.

[lukevega]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

For everything I've tried at 4.0GHz so far worker 2 is still almost instantly failing with small FTTs. RAM is at 2.1V - I think it is rated for up to 2.2-2.3V. MCH Core (NB) I have tried up to 1.3V so far...

[SmOgER]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

Well if it's small fft failing then it's not the RAM.

What's your PLL voltage at?
GTLs?

Oh and btw, that vcore numbers you are reporting should be voltages under FULL LOAD, cause bios or idle voltages are irrelevant now. As I said before, my previous C0 did 4Ghz on 1.38V (full load), so your stable voltage should probably be in 1.36-1.405v range.

[lukevega]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

Last BIOS voltages that failed:

Performance enhance = standard
MCH Core= 1.26
vcore= 1.45
Vtt= 1.4
CPU PLL= 1.61
LLC disabled

I don't see a GTL setting...

[SmOgER]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

Find out your real vcore under full load and post back. BIOS vcore target doesn't matter much, especially with LLC disabled.

PS. Those smileys in this forum are so odd I couldn't even find a single decent one, haha.

[profJim]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

This is probably Luke's current frame of mind.
When you don't provide details including ET6, Memset, HWiNFO etc. you are wasting your time and ours.
What are GTL's? They are your reference volt settings and they are also displayed in ET6.

Switching DRAM Performance Enhance from Turbo to Standard changes quite a few bios settings.

I did a quick 4.0GHz test using DRAM Performance Enhance = Standard where nearly all bios settings were set to AUTO.

My Q9550 cpu speed was set to 8.5 x 470MHz and the only other manual settings were CPU PLL and CPU Termination (aka VTT).
LLC was disabled in all test setups.

The primary memory timings were quite loose, but most results were much better than I expected.

• The MaxxMem2 memory benchmark was lower than expected but this is due to the loose primary memory settings.
• The 5120MB LinX test results were a bit low, but it passed.

I planned on running a one hour Prime95 Blend test that would include 2 test loops for Large FFT's and 2 loops for Small FFT's.
Unfortunately my system froze at 14 minutes, 43 seconds of the first (720K FFT size) test loop.

I tried quite a few tweaks at 4.0GHz but all tests failed to complete the LinX 3 loop test using 5120MB of memory.
The LinX scores averaged 53.36 GFlops for the partial tests, before the system froze.
I did not do any reference volt tweaks.

notes:

• ambient temperature was 8^o(C) warmer than normal and the cpu and core temperatures are significantly higher
• with tighter memory timings, MaxxMem2 benchmark results were noticeably higher than the almost all AUTO initial test

note: the last MaxxMem2 test is from my untested 4.20GHz setup.

[profJim]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

I don't know if this will give any diagnostic value (where can I find a guide that tells me how to interpret the results of failures from prime95 and LinX?!?!) but I tried LinX at 3.85GHz, keeping vcore at 1.4V and VTT at 1.36V with no luck:

[ATTACH]7140[/ATTACH]

I have no idea how to tell what changes I should try other than your guidance Jim!

With Prime95:

• errors with FFT length less than 128K usually means that the problem is cpu related
• errors with FFT lenght larger than 128K usually means that the problem is memory related

With LinX or IBT I don't know of any diagnostic value when the test fails.

One thing you should regularly check is to see if HWiNFO or RealTemp reports any cpu core throttling.

[profJim]

Re: Socket 771 Mod Xeon X5460 GA-P35-DS3L Overclock

Some thoughts:

• Based on my quick and dirty 4.00GHz testing, the stability issue might be related to running with primary memory timings that are too tight.
• Our Kingston HyperX T1 memory is rated for 5-5-5-15 timings at 1066MHz for a 2 x 2GB memory kit.
• This does not mean that we can be stable with a 4 x 2GB setup with rated timings; 2.10 volts and 4-4-4-12 @ 800MHz or 5-5-5-15 @ 1066MHz.
• Several years ago, many higher quality 4 x 2GB setups had to underclock the primary memory timings to around 1000MHz or slightly lower.

In my followup 4.00GHz testing, I'm going to use looser primary timings, similar to the auto-generated 5-6-6-22 timings.

I will also tweak the MCH Reference (GTL) setting, increasing it one step at time using the VTT procedure that I posted earlier.

I'm also going to run the Prime95 Small FFT test for at least 2 test loops before I run the Prime95 Blend test.

I will still run the MaxxMem2 benchmark and LinX 5120MB 3-loop test before I run the longer running Prime95 Small FFT test.