MemTest86 does not seem to utilize Intel Turbo Boost?

[turbocrash]

The computer:
http://www.dell.com/support/home/us/.../optiplex-7010

The RAM:
http://www.crucial.com/usa/en/bls8g3d1609ds1s00
8GB DDR3 PC3-12800 • 9-9-9-24 • Unbuffered • NON-ECC • DDR3-1600 • 1.5V • 1024Meg x 64 •

Hello,
I have a Dell Optiplex 7010 Small Form Factor running Windows 7 ultimate 64 bit. I went ahead and crammed it full of CAS 9 RAM, (4x8GB=32GB) when this computer would normally take CAS 11 RAM. (nowhere in the Dell manual does it specify the CAS, rather, the Crucial web site only suggests CAS 11 for this model.) The result of loading it up with CAS 9 RAM is that this computer randomly reboots, and corrupts the hell out of Windows in the process. (after 2 or 3 of these reboots, Windows will no longer boot into anything but Safe Mode.) This random spontaneous rebooting can happen anywhere from 10 minutes to 5 hours after the drive is re-imaged with a fresh Windows install. (with average time before crash being about... eh... 10 to 15 minutes with some youtube videos playing and surfing some random web sites.)

So my first troubleshooting test was to run MemTest86 overnight, and the result was no errors found...

So I took a second known (or at least presumed) good 7010 SFF and switched the CAS 9 RAM over to this second 7010, imaged the drive with a fresh Windows image, and I got the same deal, the crashing seemed to follow the CAS 9 RAM.

Eventually I figured out that if I disable Intel Turbo Boost, I do not seem to crash with the CAS 9 RAM, although I only figured that out a few hours ago, so I will need a few weeks of testing to say *for sure* that the random rebooting is gone.

So this experience has led me to realize that MemTest86 does not appear to get Turbo Boost mode going while testing the RAM. Irrespective of who is at fault for the user's system to be crashing, be it RAM, OS, Drivers, chipset, etc, it would be wonderful if MemTest86 could 'activate' TurboBoost while testing the RAM, just to allow the user to see if the RAM is incompatible with the Turbo Boosted system. I understand that it could be technically challenging to get the Turbo Boost to kick in while running the memory test, but if it could be done, you can see where this would provide some really useful diagnostic information to the user.

I don't know if there are any test utilities out there that would stress test in a way that would cause Turbo Boost to fire up?

If anyone at Passmark sees this and thinks it's a good suggestion, can you please forward it to the appropriate persons for review!

Thanks!

[David (PassMark)]

Turbo boost only works (at the max turbo speed) if you run 1 thread at a time.
So if you are running MemTest86 on all cores, then you won't see the turbo speed.

However there is an option in MemTest86 to select a single CPU/Core. Which should allow the one test thread to run at the turbo speed.

[turbocrash]

Hi David,
I did try with just one CPU, but I didn't get a crash right away. However on your advice I'll try it again, as follows: I turned Turbo Boost back on in the BIOS and booted MemTest86 selected option 4 to start in single core mode, so it reports to be using only CPU: 0 , however, so far it's not crashing nor finding any errors, and the CPU Clk is reported as 3403 Mhz which is what this particular CPU (306a9) is normally clocked at in this particular computer. (I don't know if Turbo Boost would result in a real-time change in the reported CPU Clk?) I'll leave it running for 10 hours like this while I'm at work and see if anything happens.

In reading about Turbo Boost I suspect that the operation of MemTest86 might not actually create exactly whatever the needed conditions are for Turbo Boost to self-activate, although I don't know. I may be wrong on this but I get the impression that MemTest86 does not control system resources via ACPI, and http://en.wikipedia.org/wiki/Intel_Turbo_Boost suggests that for TurboBoost to activate, the OS would need to request via ACPI "the highest performance state of the processor," but it is Wikipedia after all so who knows how accurate or complete that is.

One possibility would be that my system is not unstable while in any of the Turbo Boost levels, rather, my system is crashing during the transitions to and from Turbo Boost and/or during the transitions to and from one level of Turbo Boost to another. To test this possibility, MemTest86 might have to repeatedly randomly ramp up and down the Turbo Boost from 'off' state to 'maximum' state and every state inbetween. Of course it might take too many years to do this while testing every single part of a memory DIMM, but doing this ramping up and down while testing a few parts of the DIMM (beginning, middle, end?) would suffice to detect any obvious failure resulting from an overall incompatibility between the DIMM and the system in question.

[David (PassMark)]

The CPU clock speed is measured as MemTest86 starts up. Both the base speed and the single core turbo speed are measured.

These are recorded in the log and also displayed in the user interface. We believe the measured turbo speeds match the specs. Which would indicate that the CPU can turbo when running MemTest86.

If it does turbo or not also depends on other factors like temperature, power consumption & the load on the CPU. So it is hard to know exactly when it 'turbos'.

Update: After getting your E-mail on this topic, I should also add that I was talking about the MemTest86 V6 release above. While it seems you were talking about the V4 release.