Anyone with a Prolimatech Megahalems and an overclocked I7-2600k?

[Fear No Evil]

Can you tell me what your temps are under load? I'm not sure my heatsink is properly seated and my temps will go to near 80 degrees C with a single fan . Not sure if that's expected or if I need to pull it apart and try to seat it again or if I just need to add another fan or higher rpm fan. I'm currently overclocked to 4.7 ghz. Room temp is about 78 degrees F.

 

[BonzaiDuck]

Can you tell me what your temps are under load? I'm not sure my heatsink is properly seated and my temps will go to near 80 degrees C with a single fan . Not sure if that's expected or if I need to pull it apart and try to seat it again or if I just need to add another fan or higher rpm fan. I'm currently overclocked to 4.7 ghz. Room temp is about 78 degrees F.

What software are you using to load up the processor and stress it? I just left a new thread I started to raise the point that PRIME-95 is suddenly an ultimate stresser, with IBT giving sissy-low temperatures.

I"m 100% sure that my own system with the Noctua D14 is as cool and proper as it gets. But I can imagine the thermals moving toward 80C as you get to 4.7. I was "there" -- briefly the other day -- because I started fiddling with ASUS AI_Suite, punched the extreme button out of careless curiosity, and it was "runaway horses" and a BSOD, reboot, TPU "doing its thing" to sort things out. I'm determined to make that my last and only BSOD on this puppy.

I haven't seen enough second-hand results to really know what sort of temperatures to expect for the higher clock settings. All I know is that all four of my cores flip around at room-ambient values in idle, but with PRIME95 the "hot" core will go to 71C if the room-ambient is 80F or more. . . . The IBT load temperatures only get into the low 60's Celsius.

What is the effective VCORE under load with your 4.7Ghz settings?

 

[Shmee]

and how much airflow does your fan have? I use a 110 CFM scythe on my second system's TRUE black. I know some go over 250 CFM, but are loud.

 

[birthdaymonkey]

What software are you using to load up the processor and stress it? I just left a new thread I started to raise the point that PRIME-95 is suddenly an ultimate stresser, with IBT giving sissy-low temperatures.

Is there a new version of Prime95 that just came out? Adding the new(ish) linpack binaries to IBT, I've found the opposite to be true. Prime95 (64 or 32) heats up my 2500K to ~65, while IBT pushes ~75.

 

[manimal]

29-34 idles across the cores with it overclocked to 4.4. Loads mid 60s with summer ambients. I

 

[Axon]

Fear,

Just saw this. I have the setup outlined in my signature. I also get temps into the mid 80s on the hotter two cores and low 80s on the cooler cores. If I shut down HT, it's more like high 70s. I have a megahalems and a push/pull config with two scythe gentle typhoons at around 1500 RPM. Plenty of airflow in my case.

Previously, I had a 2500k and that capped out around 75c as well, so it seems that's just how hot these chips get. I see these crazy low temps around and I wonder how the hell people get them, but as long as I don't experience instability, I'm not worried.

That said, depending on how it reviews, I might give the Corsair H100 a try and throw some good aftermarket fans on it. I really like the clean look of those mini-WC loop systems.

 

[BonzaiDuck]

Is there a new version of Prime95 that just came out? Adding the new(ish) linpack binaries to IBT, I've found the opposite to be true. Prime95 (64 or 32) heats up my 2500K to ~65, while IBT pushes ~75.

Let's explore this further. My IBT was the last revision, as was the PRIME-95 revision. Absolutely sure of that. I did a double-take the other day, noting that IBT has an option to change the number of threads it runs, but it has a default "Auto" setting. PRIME-95 also "auto"-selects the number of threads and brings up eight workers. The number of threads that IBT runs on "auto" is not apparent, although activity shows in all eight activity windows of my monitoring software.

Also, I THOUGHT for SURE that I installed all the drivers and software correctly, including download-updates to the motherboard BIOS, Intel features etc. that were absolutely the most recent and current revision.

But CPU_Z -- most recent version -- doesn't seem to report the frequency/speed correctly, giving only what would be expected under idle without EIST for any bCLCK adjustments. Further, it's VID voltage reading bounces around, and appears to have an upward bias, because the ASUS monitoring software seems to be consistent and reliable by contrast.

This also holds true for CoreTemp and RealTemp -- also most recent versions -- with inaccurate/untrustworthy speed or VID reports, but temperatures consistent with my ASUS monitoring software.

Someone posted an illuminating explanation of "Intel Management Engine" the other day, noting that if it was the wrong revision or wasn't installed, both over-clocking or proper monitoring would suffer or just be impossible. But I'm sure I got the latest revision. I can check again. But why would the ASUS AI-Suite monitor seem to be spot-on? would it not also require IME?

 

[birthdaymonkey]

Let's explore this further. My IBT was the last revision, as was the PRIME-95 revision. Absolutely sure of that. I did a double-take the other day, noting that IBT has an option to change the number of threads it runs, but it has a default "Auto" setting. PRIME-95 also "auto"-selects the number of threads and brings up eight workers. The number of threads that IBT runs on "auto" is not apparent, although activity shows in all eight activity windows of my monitoring software.

I had to update IBT manually by adding the new Linpack files that enabled AVX. This made a huge difference in the amount of GFLOPS that the program reported ~57 - ~110, and a significant change in the temps too. Of course, my CPU doesn't make use of hyperthreading, so that might be the difference. IIRC, hyperthreading actually slows down performance in IBT, so you might try disabling it for maximum CPU torture.

EDIT: A bit of Googling reveals that IBT 2.51 has the new Linpack libraries, but 2.5 did not. I think you might need SP1 to make it work properly. The general consensus is that for SB, AVX Linpack is the ultimate heat producer, so something seems to be fishy if you're getting higher temps with Prime95.

This thread for more info: http://www.overclock.net/intel-cpus/1007442-gflops-question.html

 

[BonzaiDuck]

EDIT: A bit of Googling reveals that IBT 2.51 has the new Linpack libraries, but 2.5 did not. I think you might need SP1 to make it work properly. The general consensus is that for SB, AVX Linpack is the ultimate heat producer, so something seems to be fishy if you're getting higher temps with Prime95.

This thread for more info: http://www.overclock.net/intel-cpus/1007442-gflops-question.html

Responding to your observations, I opened IBT just now, which reports in "About" that it's indeed version 2.51.

Also, for the choice of stress intensity -- "High," "Very High" "Maximum" -- I've only run the second-highest so far. It seems that the difference is the amount of memory incorporated into the test. I'll keep looking at it, and we'll get this sorted out. Your contributions are much appreciated for that, as I myself investigate a bit further.

One more thing: You most likely refer to SP1 for Windows 7. At the moment, my 64-bit OS is VISTA SP2.

 

[birthdaymonkey]

Responding to your observations, I opened IBT just now, which reports in "About" that it's indeed version 2.51.

Also, for the choice of stress intensity -- "High," "Very High" "Maximum" -- I've only run the second-highest so far. It seems that the difference is the amount of memory incorporated into the test. I'll keep looking at it, and we'll get this sorted out. Your contributions are much appreciated for that, as I myself investigate a bit further.

One more thing: You most likely refer to SP1 for Windows 7. At the moment, my 64-bit OS is VISTA SP2.

Yeah, AVX is enabled in Win7 SP1. Maybe it doesn't work with Vista... I'm not sure... more investigation sounds like a good idea.

As for the High, Very High, Maximum, you're right. It changes the amount of memory used. It doesn't seem to affect the amount of heat or FLOPS generated.

Also, you're welcome!

 

[BonzaiDuck]

Yeah, AVX is enabled in Win7 SP1. Maybe it doesn't work with Vista... I'm not sure... more investigation sounds like a good idea.

As for the High, Very High, Maximum, you're right. It changes the amount of memory used. It doesn't seem to affect the amount of heat or FLOPS generated.

Also, you're welcome!

I do things in discrete steps with this business now . . . sometimes waiting for a day to ponder what will happen. . . . . also . . . . it's not too hard to focus on one thing at a time. I assume AVX is a feature involving graphics? Would it relate in any way to the iGPU?

That's why I may switch to W7-64-SP1, because it supports the Lucid-Virtu software. I had just sent a polite but tersely-worded query to LucidLogix, and have yet to hear from them, but the Lucid Virtu bundled in my motherboard support software is only to be found in the Win7 64-bit subdirectories of the support-disc.

Further, in their "requirements," Lucid clearly states "compatible with Win 7 only" and I think they used the word "only" -- one way or the other.

Planning to get to it later, and having installed what appears to be the Intel Graphics drivers with no adverse impact on VISTA-64 or Event-Viewer Bangs, I eventually wanted to find out in detail what and how the iGPU (enabled in BIOS but not yet for "multi-monitor for Lucid Virtu") -- provides in enhancement. Does it speed up rendering? I only got the gist of a HOW-TO review of the ASUS P8Z68 V Pro, with most or all of the review equally relevant to the entire P8Z68-V/V-Pro/Deluxe model line.

At this time, I've only installed the drivers. By the way. In some other thread about the "V" board, the essential questions were about the nature of certain hardware and their drivers. I believe the OP wanted to decide whether to install this or that. My belief is that all drivers should be installed with the devices made Enabled in BIOS, which can later be selectively turned off and mostly keeping the driver out of windows' active services in the "Disabled" BIOS state.

But again -- it would be good to know what improvements I've enabled in VISTA without any Lucid Virtu features installed.

 

[BonzaiDuck]

I had to update IBT manually by adding the new Linpack files that enabled AVX. This made a huge difference in the amount of GFLOPS that the program reported ~57 - ~110, and a significant change in the temps too. Of course, my CPU doesn't make use of hyperthreading, so that might be the difference. IIRC, hyperthreading actually slows down performance in IBT, so you might try disabling it for maximum CPU torture.

EDIT: A bit of Googling reveals that IBT 2.51 has the new Linpack libraries, but 2.5 did not. I think you might need SP1 to make it work properly. The general consensus is that for SB, AVX Linpack is the ultimate heat producer, so something seems to be fishy if you're getting higher temps with Prime95.

This thread for more info: http://www.overclock.net/intel-cpus/1007442-gflops-question.html

Back to this. . . . First, I still need to review the link you gave, but there are some distractions at the moment -- so a little later today . . .

Second, I disabled HT in BIOS, and ran PRIME for at least 10 or 20 minutes, followed by a minimal 5 iterations of IBT. Now with four threads, Prime reaches 65C only here and there on the hottest core. IntelBurnTest reaches about 68C at maximum as recorded in the Min/Max lines of the dialog.

You would agree that we are talking about a very real set of mathematical relationships between voltage, temperature and loaded speed. For Extreme OC'ing, there are some exponential curves where one is trying to find a point that is already on the part the curve after turning sharply upward -- these things were explained with graphs in a 2007(?) ('08?) Fall article here at Anandtech on what was then the just-released high-end Yorkie.

While we want to push it to a comfortable chosen limit, these things are equally important to someone who does not include "extreme" in the priority list. We all want to know where those curves take sharp bends at various levels of VCORE, temperature and speed.

I'm wondering if there is really lacking any linear correspondence with how PRIME and IBT load the cores across the scenarios of HT Enabled or Disabled. But a three degree difference with high-load VCORE at 1.288V or so, bCLCK 103, Turbo Multi of 42 and speeds shy of 4.4, such a difference doesn't seem too significant.

That may all change if I increase the multplier another notch or two, try to sqeeze between 104 and 107 out of the bCLCK, and either SET VCORE higher or let "Auto" take it there.

But the inconsistencies with the shareware monitoring SW just-recognized this week, I'll have to go back and retry the settings in that range -- all over again. And be mindful that I use the ASUS monitor . . . .

 

[birthdaymonkey]

I do things in discrete steps with this business now . . . sometimes waiting for a day to ponder what will happen. . . . . also . . . . it's not too hard to focus on one thing at a time. I assume AVX is a feature involving graphics? Would it relate in any way to the iGPU?

I believe AVX is a new instruction set for SB or a modification to an existing one, akin to SSE, SSE2 (etc.). I don't think there's a lot of software out there now that will really take advantage of it yet, but it makes a big difference in Linpack.

http://en.wikipedia.org/wiki/Advanced_Vector_Extensions

How many GFLOPS is your system currently reporting for each pass of IBT?

 

[BonzaiDuck]

I believe AVX is a new instruction set for SB or a modification to an existing one, akin to SSE, SSE2 (etc.). I don't think there's a lot of software out there now that will really take advantage of it yet, but it makes a big difference in Linpack.

http://en.wikipedia.org/wiki/Advanced_Vector_Extensions

How many GFLOPS is your system currently reporting for each pass of IBT?

Thanks. I may have posted the following question on another thread, but I'm still in the dark about how the iGPU will affect my [VISTA-64] system without Virtu. As I may have said, I enabled the iGPU allocating a nominal 64MB of memory for it; no malfunctions or adverse indications; but the feature "multi-monitor for Lucid-Virtu" is disabled.

I also installed HWMonitor -- a CPUID product like CPU-Z, and find the vCORE is consistent with the ASUS monitor, temps consistent with Real/Core/Temp.

I'll have to fire up the Z68 system today and run an IBT so I can get back here to report the GFLOPS data . . . .

 

[BonzaiDuck]

OK . . . back again, gents . . . and for BirthdayMonkey, the stats:

So far, four iterations of IBT.

GFLOPS Min: 62.5
GFLOPS Max: 63.34

Keep in mind, this is 4.34Ghz, HT turned OFF.

The temperatures:

Min core at load: 55C
Max core at load: 62C

This seems below the four-core results from RealTemp the other day, same circumstances. Today's room-ambient: 77F [July elec bill will be unfortunate with this heatwave . . . ] Seems the last run the other day was at 78F.

Something weird has happened, or the RealTemp monitoring was off from what it should have been, because I recall a high-core of around 68 or 69F. 1/2C in room-ambient hasn't anything to do with it. These new numbers are all from HWMonitor.

The HWMonitor voltage info is identical to the ASUS monitor voltage info, and I think the ASUS monitor thermal info will also be consistent with HWM.

Now . . . . if I'm missing something here for lacking an instruction set enhancement to the OS choice of VISTA, then I'll look forward to the FED-EX arrival of the Win-7-SP1. Gotta get crackin'.

What are the GFLOPs reported for Win 7, HT "OFF" in an overclock range between 4.3 and 4. . . . whatever ? I KNOW they can be over 100 for more extreme OC speeds in Win 7. And that's per something I read in the overclock.net forum thread . . .

 

[birthdaymonkey]

OK . . . back again, gents . . . and for BirthdayMonkey, the stats:

So far, four iterations of IBT.

GFLOPS Min: 62.5
GFLOPS Max: 63.34


What are the GFLOPs reported for Win 7, HT "OFF" in an overclock range between 4.3 and 4. . . . whatever ? I KNOW they can be over 100 for more extreme OC speeds in Win 7. And that's per something I read in the overclock.net forum thread . . .

Just ran IBT on my 2500K. At 4.2GHz it gets 104 GFLOPS. Before I had the new Linpack files installed I got ~60. I'd say the issue is either Vista or 2.51 doesn't actually come with the updated libraries and they have to be installed manually.

It probably doesn't really matter in real world performance though... It's just a highly synthetic benchmark.

 

[BonzaiDuck]

Just ran IBT on my 2500K. At 4.2GHz it gets 104 GFLOPS. Before I had the new Linpack files installed I got ~60. I'd say the issue is either Vista or 2.51 doesn't actually come with the updated libraries and they have to be installed manually.

It probably doesn't really matter in real world performance though... It's just a highly synthetic benchmark.

What I could do is look to some other LinPack-affiliated utilities. the 2.51 version was fairly recent. Maybe OCCT, or LinX if I can find it. But if I wait until early in the week, I'll be installing the Win-7-64 OS anyway. Only problem with that, I was really hoping to do it on a different hard disk -- an uncertain arrival.

MS is still supporting VISTA SP2. You wonder what it would take for them to add recognition of the additional instruction set. But there were other reasons I decided to throw up my hands and "move up" now. . . .

 

[BonzaiDuck]

Just ran IBT on my 2500K. At 4.2GHz it gets 104 GFLOPS. Before I had the new Linpack files installed I got ~60. I'd say the issue is either Vista or 2.51 doesn't actually come with the updated libraries and they have to be installed manually.

It probably doesn't really matter in real world performance though... It's just a highly synthetic benchmark.

OK . . . here's some more skinny. The re-view: I installed my Windows 7 in what seems to be around 10 minutes. Got all the drivers installed, everything set up as with the previous VISTA. System now recognized the Intel Graphics iGPU and drivers.

Meanwhile, at 4.53 Ghz, I've trimmed the VCORE (still on "Auto") down to about 1.28 to 1.29V -- pegging a maximum of 1.30 to 1.31V through an hour of stress-testing at between 78 and 80+F room ambient.

LinX doesn't show any increase in the GFlops -- the software needs revision. But IntelBurnTest -- the lates version of June (I think?) shows about 122 GFlops over the 60+ I had with the VISTA installation.

This also adds a tad to the heat, but not in a linear way. And -- yes -- Hyper-Threading seems to "slow down" the IBT, because it shows closer to 100+ GFlops. Even hyperthreading with PRIME95 is cooler on temperatures than the HT plus IntelBurnTest.

Howsoever that may be, I think I can reach 4.6 Ghz and still keep the temperatures from peaking beyond a 72C to 75C range. And I still haven't found the minimum voltage for 4.53, but I'm close . . . . I'm close . . .