Iteration times in i5 and i7
 

I'm using Prome95 for the Seventeen or Bust project. I've been running four exponents on a quad-core i7 for quite a while, I am now testing exponents in the 24.7 million range. A test on the i7 takes a little less than 7 days and iteration times are around 21-22 ms.

A few days ago I added a quad-core i5, also running four exponents. At first it was also showing iteration times of 22ms. The status shows that the tests will take about 6 days (none finished yet). (They have been running for about 2.5 days and are 30% done, so it may take 8 days.) Then yesterday the iteration times went from 22 to 41. I checked the power settings and made changes to set it for high performance, and the iteration times went back to 22. Then this morning they were back to 40-41. I checked the power settings, and they are OK. I rebooted and they are still at about 41. At 41 the tests are going to take about 11.4 days.

Is there a way to get it back to 22ms iteration times?

I have an inkling that you may be feeling the effects of thermal throttling. I assume that these are desktop computers - if so, try and get some better airflow in to the case using a fan. If a laptop, try the Throttle=xx setting in the config file to reduce the amount of CPU time Prime95 runs. If that doesn't work, then just get a cooling stand, like [URL="http://www.amazon.co.uk/Zalman-Notebook-Cooler-ZM-NC1000-Silver/dp/B000JSFT2C/ref=tag_stp_s2_edpp_url"]this Zalman cooler from Amazon.[/URL]

[QUOTE=f1pokerspeed;348124]I have an inkling that you may be feeling the effects of thermal throttling. I assume that these are desktop computers - if so, try and get some better airflow in to the case using a fan. [URL="http://www.amazon.co.uk/Zalman-Notebook-Cooler-ZM-NC1000-Silver/dp/B000JSFT2C/ref=tag_stp_s2_edpp_url"][/URL][/QUOTE]

They are both desktops. The fan on the new i5 is not making much sound, but I can feel air going in where it should and warmer air going out the back. It has plenty of ventalation. Speccy says that the CPU is running at 55C (in the green) and the motherboard at 28C (also in the green).

So it doesn't seem to be hot. The task manager says that Prime95 is running at 98-99% of the CPU.

I just shut off three of the workers to see if that makes a difference.

On the i5:

running only 1 test, the time drops to 17ms per iteration and the CPU cools off to about 43C.

Two tests: 19ms (both) and CPU 49C
Three tests, 25-27ms, CPU 53C
Four tests, 35-38ms, 55C

So I can get more throughput running two tests instead of four! Is there a way to fix this?

----------------
The i7 behaves a little like that, but not as much. Going from 4 to 1 drops the time from 21-22ms to 14-15, and I can hear the fan run slower.

The LL test is not easily multi-threaded (or something like that) and therefore does not benefit (greatly) from using more than one core per exponent.

Your problems are arising because you are memory bottlenecked (ie: bandwidth not good enough to handle the amount of data being pushed around). Perform one or two LL tests and then do ECM or P-1 on the other cores to your preference.

Regarding CPU temperatures, do you know the exact i5/i7 model? If so, go to the Intel website and check for their [B]Tcase[/B] temps - if the CPU runs within about 5-10 degrees of that (I think) it then implements the thermal throttling. This is just a check process - in case your CPU is running a little off.

Not necessarily in this case, I think.

Running all 4 threads(cores) with P95, check Task manager and see how much P95 is using. Also, you might want to see what CPU-Z says...

[QUOTE=f1pokerspeed;348135]The LL test is not easily multi-threaded (or something like that) and therefore does not benefit (greatly) from using more than one core per exponent. ...

Regarding CPU temperatures, do you know the exact i5/i7 model? If so, go to the Intel website and check for their [B]Tcase[/B] temps - [/QUOTE]

I'm running four tests - each on a different core, not multithreading the test of one exponent.

I do know the exact models of i5 and i7 - I will look up that data.

Since my last message I did some tests on both computers.
On the i7, running 2 gives about 1.9x throughput, running 3 gives 2.3x throughput, and running 4 gives 2.5x throughput.

On the i5, running 2 gives about 1.8x throughput, running 3 gives 2.0x throughput, and running 4 gives 1.7x throughput.

So on the i7 there are diminishing returns, and 4 is only about 9% better than running 3. On the i5, running 4 gives lower performance than running 2 or 3!

The task manager says that on the i5, Prime95 is getting almost 100% of the CPU. On the i7 it is getting 50%, because of the hyperthreading.

I can't get CPU-Z to run. The places I found to download it install things I don't want. And after it said that it did install, it took me to a folder where it said the installation would be finished, but the folder was blank.

These are links to both CPU data sheets, and the temperature seems to be OK

[url]http://ark.intel.com/products/65509[/url]

[url]http://ark.intel.com/products/52213[/url]

[QUOTE=f1pokerspeed;348135]If so, go to the Intel website and check for their [B]Tcase[/B] temps - if the CPU runs within about 5-10 degrees of that (I think) it then implements the thermal throttling. [/QUOTE]

Do you mean 5-10 degrees below TCase or 5-10 degrees above?

Sorry for the late reply, but:

1) below Tcase, anything above for any amount of time would (probably) kill the CPU
2) You probably downloaded the ZIP version of CPU-Z. Here's the installer (with no adware/spyware/etc: [URL="http://www.cpuid.com/medias/files/softwares/cpu-z/cpu-z_1.65-setup-en.exe"]http://www.cpuid.com/medias/files/softwares/cpu-z/cpu-z_1.65-setup-en.exe[/URL]

[quote]
I can't get CPU-Z to run. The places I found to download it install things I don't want.
[/quote]He probably tried the exe installer.... try the zip. EDIT: [URL="http://www.cpuid.com/downloads/cpu-z/1.65-en.zip"]here
[/URL]

[QUOTE=Jud McCranie;348125]They are both desktops. The fan on the new i5 is not making much sound, but I can feel air going in where it should and warmer air going out the back. It has plenty of ventalation. Speccy says that the CPU is running at 55C (in the green) and the motherboard at 28C (also in the green).

So it doesn't seem to be hot. The task manager says that Prime95 is running at 98-99% of the CPU.

I just shut off three of the workers to see if that makes a difference.[/QUOTE]
As again, 55C is perfectly fine for temperatures. I would be worried if it went over 75-80C imo. :smile:

On the third try I did get CPU-Z to install and not install anything I didn't want. The temp seems fine on the new i5 but it is a little hot on the old i7. I have cut both back to running three threads. On the i5, three gives better throughput than four. On the i7, the fourth one doesn't add much to the throughput, and it causes the fan to run faster because the CPU is hotter.

So what is the problem with the i5 - memory bottleneck or throttling back? The temp is not high on it. For a while it was running four exponents at 22ms per iteration. So the 3.0GHz i5 Ivy Bridge was matching the 3.4GHz i7 Sandy Bridge. Now it is quite a bit slower.

[QUOTE=Jud McCranie;348194]On the third try I did get CPU-Z to install and not install anything I didn't want. The temp seems fine on the new i5 but it is a little hot on the old i7. I have cut both back to running three threads. On the i5, three gives better throughput than four. On the i7, the fourth one doesn't add much to the throughput, and it causes the fan to run faster because the CPU is hotter.

So what is the problem with the i5 - memory bottleneck or throttling back? The temp is not high on it. For a while it was running four exponents at 22ms per iteration. So the 3.0GHz i5 Ivy Bridge was matching the 3.4GHz i7 Sandy Bridge. Now it is quite a bit slower.[/QUOTE]

Just curious. Is the clock to spec in CPU-Z and is it detecting all your memory? "Memory" tab

Is the i5 system a single memory stick? It sounds like bad memory bottlenecking, but usually at those exponents there is no problem with 2 tests and little slowdown with 3. If you have just one stick of RAM, you only have half the bandwidth available on the board. If the board offers overclocking options, you could mess with speed changes to test if heat is the problem; I once undervolted a CPU at stock speed to cure a throttling problem.

ECM and P-1 take little memory bandwidth, and are likely to pay dividends for SOB. You can run these on the cores deemed unhelpful for Prime95.
-Curtis

CPU-Z does show the correct CPU speed and it detects the memory. (Speccy does too.)

About the memory, right now I have three sticks in the i5. I knew it was coming with 4GB and can be expanded to 16GB. (I know you get better performance with pairs of memory sticks.) When I ordered it, I didn't know if the 4GB would be on one stick or two (I assumed two). I ordered 8GB (two sticks) and had them ready to install when the i5 arrived. It came with one stick of 4GB. I added the two sticks, for a total of three sticks, 12GB. I ordered another 4GB stick which should arrive in a couple of days. That may have some effect.

Incidentally, slot #2 was the one that was vacant.

It came with a stick of 12800 CL=11. The two sticks I added were 10700 CL=9. The one I ordered is another 12800.

I was trying to rearrange them so the matching pair are in corresponding slots. The manual doesn't say and the only thing on the motherboard that I can determine is that the slots are black, blue, black, blue. I assumed that the ones with the same color are the matching ones, so I did some swapping. Now #1 and #3 are the 10700, #2 is the 1280, and #4 is vacant. But the iteration times went from 29 to 30ms.

Get rid of the third stick until the fourth one comes in. I don't know what 3 sticks will do but I can only imagine you're getting single-channel bandwidth. If this is in fact the case, then your bottlenecking issues will vanish.

Intel's processors can take some serious heat. Mine is currently running in a range that would almost immediately kill an AMD. To help with thermal issues, you may want to consider adding a fan or getting a better heatsink if you're using a stock heatsink. A coolermaster hyper 212 is your best bet if it can fit in your case.

On the other hand, it doesn't sound like a thermal issue at all. Try removing the single stick first. You're right about the colours. Also, from what I've seen, motherboard manufacturers typically have more confidence in the memory slots furthest away from the CPU. This is usually for overclocking purposes, but when you're not populating all four, you might as well use those.

If your RAM is good quality, particularly the 10700, you might be able to overclock it to the same settings as your 12800. RAM also has a frequency (think of ticks per second) and the timings (CL9 usually indicates something like 9-9-9-24) are the number of ticks the memory gets to do whichever stage of reading or writing memory it is at. So, CL11 gets eleven ticks during which to do the first step, whereas CL9 only gets nine. On the flip side, the CL11 is clocked higher so the overall latency might be about the same.

The reason I bring this up is your RAM may be able to handle the faster frequency if you loosen the timings. Looser timings are typically associated with snappier response and higher frequencies with better overall bandwidth, which is what you really need.

[QUOTE=TheMawn;348475]
Intel's processors can take some serious heat. Mine is currently running in a range that would almost immediately kill an AMD.[/QUOTE]

What kind of heat are we talking about here?

[QUOTE=TheMawn;348475]Get rid of the third stick until the fourth one comes in. ...

Intel's processors can take some serious heat. ...

On the other hand, it doesn't sound like a thermal issue at all. [/QUOTE]

The new stick is due to arrive tomorrow, so I'm not going to bother to take the third one out. However, I was getting 22ms with three sticks the first day, now it is 29-30.

I don't think heat is an issue. Right now I have it on its side, and the air intake vents are on the top. Yesterday I inadvertently put a phone book on top of it - blocking the air vents. The fan sped up and got loud. Until then the fan had been very quiet. I realized what I had done and took the phone book off. Within probably a minute the fan was somewhat quieter and within a few minutes it was back to very quiet.

Actually I was getting 22ms on four workers on the i5 at first - matching my i7. Then it went to 40. I cut it back to three workers to get 29-30. I assume the iteration times it gives are accurate.

Hmm well if it's arriving soon then yeah don't bother.

Heat: AMD's latest get pretty prissy around 60C. My Ivy Bridge is running in the mid 70's. Throttling happens at 105C. Which is good because Ivy Bridge runs 10C-15C hotter than it needs to.

[SIZE=4]My Sandy Bridge i7 is running about 73C running three workers. With four workers it was about 78C. With four workers, the fan was on a higher speed. At 73 the fan doesn't run nearly as loudly, but louder than the IB i5.[/SIZE]

Well, i don't understand it exactly, but I added the fourth stick of memory and now I'm getting 19ms on three workers on the i5.

Recap: At first I was getting 22ms on four workers. I added some software (only Firefox and Thunderbird, I think) and set some things, making sure it would not sleep or hibernate. A few hours later it went up to 40ms. Then it went back to 22. Then it went back to 40.

I checked my memory, and slot #2 was empty - #1, 2, and 4 had sticks. I rearranged the sticks, now #4 was empty but the matching pairs were not in the same color of slot. Still got 44. I rearranged the memory to put matching pairs in the same color and leave #4 empty. Still 40ms. And four workers was getting less throughput than two or three, so I cut it to three workers, and the iteration time dropped to 29. Quite a but worse than I had been getting.

So today I added the fourth stick, and now I'm getting 19ms on three workers. So the memory arrangement must have had something to do it, but it was fluctuating from 22 to 40 when I didn't make any changes. (It would be one for a while then the other - then only 40.)

So now it seems to be OK and is pretty much matching my SB i7, which runs at 3.4GHz vs. 3.0.

[QUOTE=Jud McCranie;348518][SIZE=4]My Sandy Bridge i7 is running about 73C running three workers. With four workers it was about 78C. With four workers, the fan was on a higher speed. At 73 the fan doesn't run nearly as loudly, but louder than the IB i5.[/SIZE][/QUOTE]

[QUOTE=TheMawn;348504]

Heat: AMD's latest get pretty prissy around 60C. My Ivy Bridge is running in the mid 70's. Throttling happens at 105C. Which is good because Ivy Bridge runs 10C-15C hotter than it needs to.[/QUOTE]

So 51-54 Celsius is nothing on a i7 Sandy Bridge? My quad is running four workers. It apparently automatically overclocks to 3.8 GHz (normal is 3.6) and is running at about 21 ms for 55.6 M exponents.

Mid 50's is great on Sandy Bridge, Ivy Bridge and Haswell, too, I should think. You have a lot of overclocking headroom if your CPU is a k-series and your motherboard can handle it.

My mid 70's is an i5-3570k running at 4.6GHz on four cores, four threads.

I've never overclocked. I wonder about the advisability of it - it it causes the system to miss one bit, that ruins your result.

My CPU, in its current settings, is 72-hours stable using the torture tests. 4.7GHz was also, but it required significantly more voltage than I am comfortable with at the moment, and I could not get the settings to stick no matter what I did. 4.8GHz caused all of the weirdest things to crash, like windows Aero despite it being heavily dependent on graphics.

It's all relative. Stock settings aren't 100.0000000% perfect either. Run some double check LL's also if you're worried. Or leave the settings at stock. According to an old rule of thumb, which says that you half the lifespan of your CPU by a half for every 10C hotter you operate it, your CPU should outlast mine four times.

[QUOTE=TheMawn;348632]My CPU, in its current settings, is 72-hours stable using the torture tests.[/QUOTE]

But is there a point where errors start to creep in before it gets unstable? A one-bit error can ruin a week of work. I'm thinking of the old adage "Do you want it done quickly or done right?"

[QUOTE=Jud McCranie;348579] Recap: At first I was getting 22ms on four workers. I added some software (only Firefox and Thunderbird, I think) and set some things, making sure it would not sleep or hibernate. A few hours later it went up to 40ms. Then it went back to 22. Then it went back to 40.
[/QUOTE]

With two paired memory sticks and one not paired run times would vary with how much real memory the task got from the paired sticks and how much from the unpaired stick since that would be slower. That's the most likely explanation I can think of for the varying run times.

Chris

[QUOTE=chris2be8;348699]With two paired memory sticks and one not paired run times would vary with how much real memory the task got from the paired sticks and how much from the unpaired stick since that would be slower. That's the most likely explanation I can think of for the varying run times.

Chris[/QUOTE]

Yes, but the first day it changed drastically and I did nothing to the memory. I think all I did to the computer other than starting Prime95 was install Thunderbird and Firefox and update Windows 7. (I'm not sure what order I did those things.) But maybe for some reason it was using one bank of memory and then the other.

BTW, about 18 hours after adding the fourth stick it is still getting good performance.

[QUOTE=Jud McCranie;348690]But is there a point where errors start to creep in before it gets unstable? A one-bit error can ruin a week of work. I'm thinking of the old adage "Do you want it done quickly or done right?"[/QUOTE]

The torture test runs preset tests. It runs x number of iterations of exponent y using FFT length z and knows the answer you should get. Your one-bit error is enough for the program to detect an error.

[QUOTE=TheMawn;348774]The torture test runs preset tests. It runs x number of iterations of exponent y using FFT length z and knows the answer you should get. Your one-bit error is enough for the program to detect an error.[/QUOTE]

That should do it. You say that you run the torture test for 72 hours?

I looked up some videos about overclocking i5 and i7, and it may be more involved than I want to get into.

I looked at more overclocking videos, and it looks like a lot of work. If there was software to do it more automatically, I'd probably try it.

Anyhow, before I added the fourth memory stick I was getting lower total throughput with four workers than two or three, so I switched to running three. With the other memory stick, four workers does give the best throughput. I was getting 19ms with three workers and now 22ms with four, so the fourth one increases throughput only 15%.

[QUOTE=Jud McCranie;348781]That should do it. You say that you run the torture test for 72 hours?

I looked up some videos about overclocking i5 and i7, and it may be more involved than I want to get into.[/QUOTE]

I ran for 72 hours. You get to a point where it's kind of silly to carry on much longer. I bumped myself up a few notches in the voltage department afterward just to be safe. I did a bit of playing around with the settings a few weeks ago and I re-did some torture tests again. 72 hours stable again.

It's never 100% certain but pretty damned close. That's what double-checks are for.

[QUOTE=TheMawn;348800]... It's never 100% certain but pretty damned close. That's what double-checks are for.[/QUOTE]

I don't think Seventeen or Bust does double checks (probably does if one reports a prime). Also, I'm going to be doing other things, starting in a few days.

On my MacBook Pro (Sandy Bridge Core i7), I seem to be seeing Prime95 performance changing significantly from time to time.

For instance earlier this evening, one of the workers was showing iteration times of .040 or .041 fairly consistently for quite awhile. But then, the iteration times suddenly changed to 0.059. I decided to stop the workers temporarily before putting the system to sleep for a few seconds. After taking it out of sleep, I restarted the workers, and remarkably the iteration times were now only .029 or .030, and its still holding that speed over an hour later. By the way, the other workers showed similar changes in performance at the same times. (I have 3 workers, and throttle set to 85%.)

It's frustrating to see Prime95 being slowed down over 2x what it's capable of, and especially with no obvious reason for it.

[QUOTE=cuBerBruce;348944][SNIP]It's frustrating to see Prime95 being slowed down over 2x what it's capable of, and especially with no obvious reason for it.[/QUOTE]

Can you monitor temps on the CPU? A rough estimate might be made by feeling how hot the air coming out of it is, but of course, numbers are better. A brief period at idle and in sleep mode could let it cool down and stop thermal throttling, if that's what it was doing when it slowed down.

[QUOTE=kladner;348949]Can you monitor temps on the CPU? A rough estimate might be made by feeling how hot the air coming out of it is, but of course, numbers are better. A brief period at idle and in sleep mode could let it cool down and stop thermal throttling, if that's what it was doing when it slowed down.[/QUOTE]

Well, Temperature Monitor reports such wildly varying temperatures for the core temperatures, I pretty much only pay attention to the main heat sink temperatures. I tend to set the throttle duty cycle lower if I see "Main heat sink 3" getting over 50C. It's been in the 48-52 range this evening without making any changes to the throttle setting. A core, on the other hand, can change from say 97 to 73 from one second to the next, for instance, if Temperature Monitor is to be believed. I don't see much point in trying to monitor these wildly fluctuating values.

And the iteration times is often very consistent for long periods of times during the slow times. Would temperature throttling tend to produce inconsistent iteration times?

It's almost seems like the cpu speed is being changed with long intervals of time at a given speed, but the programs I currently have for reporting CPU speed seem to show its speed always at exactly 2.5 GHz (which is also the standard rated speed for the processor). So far I have no confidence in these programs.

[QUOTE=cuBerBruce;348950] A core, on the other hand, can change from say 97 to 73 from one second to the next, for instance, if Temperature Monitor is to be believed.[/QUOTE]

I wouldn't believe that is accurate.

Those are some strange circumstances. I don't blame you for being frustrated. If the reported core speed doesn't change, thermal slow down doesn't seem that likely. I have no idea of the response times for thermal control, either. Besides, you're not seeing fluctuations in other parameters to match the rapid changes of temp at the core level.

It might be an interesting experiment to set the on some kind of support to allow more air flow into and around it. If you don't have a notebook cooling pad, even a kitchen wire cooling rack might serve.

Perhaps someone with more Intel experience than I have would have a better idea of what might be happening.

[QUOTE=cuBerBruce;348950]Well, Temperature Monitor reports such wildly varying temperatures for the core temperatures, I pretty much only pay attention to the main heat sink temperatures. I tend to set the throttle duty cycle lower if I see "Main heat sink 3" getting over 50C. It's been in the 48-52 range this evening without making any changes to the throttle setting. A core, on the other hand, can change from say 97 to 73 from one second to the next, for instance, if Temperature Monitor is to be believed. I don't see much point in trying to monitor these wildly fluctuating values.

And the iteration times is often very consistent for long periods of times during the slow times. Would temperature throttling tend to produce inconsistent iteration times?

It's almost seems like the cpu speed is being changed with long intervals of time at a given speed, but the programs I currently have for reporting CPU speed seem to show its speed always at exactly 2.5 GHz (which is also the standard rated speed for the processor). So far I have no confidence in these programs.[/QUOTE]
I think I've seen the same rapid temperature fluctuations you describe on an i7-2620M laptop of my own. In my case, I was using Core Temp to read temperatures (different programs can give slightly different results but they're usually consistent relative to whatever they think the TJunction Max temperature is, which is how the CPU gives its readouts).

When the CPU isn't under heavy use, the temperature stays steady at an expectedly low level (45 C or so); if I then turn on Prime95 (or another such highly CPU-intensive program) on both cores (it's a hyperthreaded dual-core), the temperatures slowly rise until they reach the 100 C TJunction Max, at which point thermal throttling kicks in and the temperatures start cycling up and down between the mid-80's and the maximum. The interesting thing is that this cycle seems to happen at a [I]much[/I] faster pace than on most CPUs - typically, thermal throttling will drop the clock frequency for a few seconds to let the chip cool down, then turn it back up and allow the temperature to slowly rise again to the maximum, at which point the cycle repeats. On other computers that I've observed (both Intel and AMD - the ones I've observed the most are a C2D desktop and a Phenom II X4 laptop), this whole cycle usually takes place over the course of anywhere from 30 seconds to 20 minutes, depending on how well the CPU is doing at venting heat (given factors like dust, etc., which are a big deal especially in a laptop). However, on my i7-2620M laptop, the cycle happens very rapidly - on the order of just a couple seconds between the ~85 C low and the ~99 C high. This confused the heck out of me until I tried reducing Core Temp's update cycle to 100 ms (the lowest) instead of the default 1000 ms; I was getting nonsense readings since the cycle was happening too fast to get useful information at the default interval. At 100 ms, I could see the two cores rapidly (and independently - the two cores' cycles seemed independent as far as I could tell) adjusting their frequencies anywhere in the range of 2.2 GHz to 3.3 GHz or so (which means Turbo Boost is still in play even at these temperatures).

I've noticed that my particular laptop is not the only computer to do this; other i7's that I've tried this on (particularly laptops, since they tend to unavoidably hit their thermal limits at full tilt) exhibit the same throttling behavior at their limits. It appears that the Core i-series processors are able to adjust the clock frequency with extremely low latency, and are very good at dissipating heat quickly at lower frequencies, such that this fast-cycle strategy actually works well. The end result is that the CPU is dynamically adjusting itself to its thermal dissipation capabilities, allowing you to get the most possible performance out of it even under suboptimal thermal conditions.

When I first saw this behavior, I looked up some information online about i7 (or more precisely i-series in general) throttling behavior, and apparently this is to be expected - the i-series processors are extremely resilient to thermal stress, and are designed to safely run all the way up to their 100 C TJunction Max temperature. Unfortunately I do not have the link available as it was a while back and I didn't think to save it, but it did seem to be a relatively credible source - the guy who wrote the article I found worked for some sort of company that specialized in high-performance Intel CPU cooling solutions.

I have never been able to get this laptop to vent well enough for it to [I]not[/I] throttle like this under full load (with AVX-optimized Prime95) - it's an ultraportable tablet PC, so I wouldn't expect it to be designed for heavy-duty use like this. If I tilt it on its side to allow better airflow, the fan doesn't have to strain quite as hard as otherwise, and the case seems fairly well-built for thermal stress, so I'm not particularly worried about damaging it over time like this. You may, however, want to be a little careful with your MacBook - other members of the forum have reported that the plastic on their MacBooks has started to warp over time due to long-term heat exposure.

As for the iteration timings, I can't tell you too much since, with the CPU doing this kind of rapid dynamic adjustment to its present thermal abilities, you can't really expect anything to be super-consistent; I've mostly been running work for the Conjectures 'R Us project lately, which has been dealing with a rather wide variety of FFT sizes, so I hadn't been paying much attention to the iteration timings as I'd be comparing apples and oranges. All I can say for sure is that in order to get really consistent timings, you'll need to figure out a way to cool your system efficiently enough that it has a steady temperature (i.e. stays below the 100 C limit and doesn't trigger throttling). If you can do that, you should get full, steady performance out of your CPU. However, it may not be possible since many laptops are not able to achieve this even with a good cooling pad.

One thing you can try, for the sake of experimentation, is to put an ice pack under your computer (yes, really!) - put a thin cloth between the ice pack and the computer to make sure it doesn't sweat and get water into the holes on the bottom. Usually the "gel" type ice packs that can lay flat work best. Unfortunately, this trick doesn't work super well on many modern laptops since they tend to have plastic cases without really good thermal contact between the case and the interior parts, so you just end up chilling the plastic; it might be worth a try though. I have an old Pentium II laptop that this used to work really well on...it had some kind of plastic (might have even been metal) surface on the bottom that conducted heat really well, which was bad when it was on your lap, but made the ice-pack trick work really well. After 5 minutes of being on the pack the whole thing would be cool to the touch with the fan running at lowest speed. Long story short - your mileage may vary. :smile:

Thanks (especially to mdettweiler) for the comments. mdettweiler's explanation seems to correspond with how my system appears to be behaving. I had had the impression that thermal throttling generally caused throughput to get reduced drastically. But, at least with some Core i7s, as mdettweiler describes, apparently thermal throttling does not necessarily result in bad performance. I note that even though my MacBook Pro appears to throttle like this while running Prime95, it still outperforms (by a big margin) any of my other computers running Prime95. The processor is a Core i7-2860QM.

[QUOTE=mdettweiler;348957]I have never been able to get this laptop to vent well enough for it to [I]not[/I] throttle like this under full load (with AVX-optimized Prime95)[/QUOTE]

I don't seem to need to be running Prime95 anywhere near "full load" to get this behavior. I doubt it's possible to run it at any "useful" speed and not see this.

When I started running Prime95 on this laptop, I was concerned with seeing temperatures in the 90s a lot. I decided to run only two workers and sometimes turned one (or both) off when the heat sink temps were getting higher than I preferred. A few weeks ago I added a third worker, but used the throttle setting to limit workload rather than turning workers off. In the near future, I'll probably try going to a 4th worker (and decreased throttle setting).

[QUOTE=mdettweiler;348957]You may, however, want to be a little careful with your MacBook - other members of the forum have reported that the plastic on their MacBooks has started to warp over time due to long-term heat exposure.[/QUOTE]

Apple says the case is made of aluminum. And yes, it has gotten hot enough that I have to move it away from my skin. I generally have the back of it hanging over the edge of a table/chair so it can vent better. Doing this seems to help bring down the temperature. I imagine it needs some cleaning. I've been reluctant to try to remove the cover.

- - - - - -

I note that just before going to bed last night, the iteration times of the workers jumped up again. I did the sleep trick, and the iteration times immediately dropped down again. They stayed down over 4 hours, then jumped up again. I tried the sleep trick again, and the iteration times are down again. So anyway, it seems like I have a way of dealing with the slowdown once I notice that it has occurred. I'll try cleaning it and some other experiments when I have time.

I'm guessing after throttling like it does for awhile, it decides to limit the max frequency it will run at or something like that, causing a noticeable change in performance level. Perhaps putting it to sleep resets some state machine controlling this, so it goes back to the default max frequency. I'm just guessing, though. Anybody know if there is a way to get the true running frequencies of the cpu (or individual cores, if they can run at independent speeds)? (Remember this is Mac OS X - Lion, not Windows or Linux).

[QUOTE=cuBerBruce;349015]Apple says the case is made of aluminum. And yes, it has gotten hot enough that I have to move it away from my skin. I generally have the back of it hanging over the edge of a table/chair so it can vent better. Doing this seems to help bring down the temperature. I imagine it needs some cleaning. I've been reluctant to try to remove the cover.[/quote]
You actually shouldn't need to remove the cover to properly dust out a laptop - they have their airflow optimized such that there's essentially one unified airflow path through the case. It's not like a desktop where the dust can get trapped in all manner of nooks and crannies. Blowing compressed air in through the keyboard should cause the dust to come out via the fan vents and intakes. (Beware - this will give you a big nuclear cloud of dust, so do it somewhere that it won't get all over things. And maybe put something over your nose and mouth. :smile:)

[quote]I note that just before going to bed last night, the iteration times of the workers jumped up again. I did the sleep trick, and the iteration times immediately dropped down again. They stayed down over 4 hours, then jumped up again. I tried the sleep trick again, and the iteration times are down again. So anyway, it seems like I have a way of dealing with the slowdown once I notice that it has occurred. I'll try cleaning it and some other experiments when I have time.

I'm guessing after throttling like it does for awhile, it decides to limit the max frequency it will run at or something like that, causing a noticeable change in performance level. Perhaps putting it to sleep resets some state machine controlling this, so it goes back to the default max frequency. I'm just guessing, though. Anybody know if there is a way to get the true running frequencies of the cpu (or individual cores, if they can run at independent speeds)? (Remember this is Mac OS X - Lion, not Windows or Linux).[/QUOTE]
It might not have anything to do with any states being reset (at least not directly); putting the computer to sleep gives the CPU a "power nap" of sorts. Since the CPU is completely turned off (not just idling) it can cool down rapidly, so that when it comes back up the throttling is no longer an issue. Of course, it won't be long before it's throttling once more; but it does give it a little break. This technique can be useful when a laptop is in imminent danger of overheating and is unable to "save" itself quickly enough even under idle due to built-up dust and other factors. (I've had to use it a few times on my AMD Phenom II X4 laptop after subjecting it to intense CPU+GPU usage, i.e. gaming - even when you shut off the game and leave it at idle, the GPU is still putting out too much heat for the CPU to keep up with its own heat output even at its lowest throttling frequency.)

What's interesting about your situation is that you're getting 4+ hours of increased performance after the sleep-trick "power nap". Without knowing the actual temperature readouts, I can only speculate as to why this is happening, but you may be on to something with your theory about the CPU deciding to implement a temporary frequency ceiling after running at full load for an extended period. My Phenom laptop does something similar if it gets too dusty - when you first boot it up, it'll try cycling all the way up to the full clock frequency (2.2 GHz), but when it realizes that it can never hold that frequency for any length of time it just gives up and stays at 1.6 GHz indefinitely.

As for reading the real-time frequencies from the CPU, I know most popular temperature monitoring programs on Windows can do this (Core Temp, RealTemp); but I'm not familiar with what's available on Mac. On Linux, the usual solution is to use the lm-sensors package, which provides an interface for a wide variety of readouts; you can run "sensors" from the terminal to get a basic readout, or use a variety of GUI frontend programs.

[QUOTE=mdettweiler;349025]
What's interesting about your situation is that you're getting 4+ hours of increased performance after the sleep-trick "power nap". [/QUOTE]

Yes, but will the increased performance make up for having it off for a while?

[QUOTE=Jud McCranie;349026]Yes, but will the increased performance make up for having it off for a while?[/QUOTE]
Since the downtime is only a few seconds, the time is made up for very quickly.

Laptops (and the aluminum macbooks in particular) have a very low [I]rate[/I] of cooling, usually because the CPU doesn't happen to be situated next to a vent and the heat must be moved with a heatpipe to a set of very small fins cooled by a very small fan over a very small vent which often has just a few millimeters of clearance above whatever surface it's on. Worse yet is when the vent is entirely covered by a deformable surface, such as when it is placed on [B]TOP[/B] of a [B]LAP[/B].

The thing about the macbooks now is because they vent off heat by using the metal casing (in some cases) as a MASSIVE heatsink, they have an insanely high [I]heat capacity[/I] which means it takes a good deal of time for the temperatures to change very much.

This can be a bit problematic because if the processor is running a full load at a maxed frequency, one is tempted to say "yeah it's fine look it's still way below TJMax after a minute" but it might actually take ten or twenty or a hundred minutes for the CPU to reach a equilibrium which is already well above the max temperature you want it at.

If you have a metal case, then blowing any kind of air over the case itself (setting the laptop upside down will help a lot too) will probably solve all your problems and them some.

[QUOTE=TheMawn;349047]Laptops (and the aluminum macbooks in particular) have a very low [I]rate[/I] of cooling, usually because the CPU doesn't happen to be situated next to a vent and the heat must be moved with a heatpipe to a set of very small fins cooled by a very small fan over a very small vent which often has just a few millimeters of clearance above whatever surface it's on. Worse yet is when the vent is entirely covered by a deformable surface, such as when it is placed on [B]TOP[/B] of a [B]LAP[/B].

The thing about the macbooks now is because they vent off heat by using the metal casing (in some cases) as a MASSIVE heatsink, they have an insanely high [I]heat capacity[/I] which means it takes a good deal of time for the temperatures to change very much.

This can be a bit problematic because if the processor is running a full load at a maxed frequency, one is tempted to say "yeah it's fine look it's still way below TJMax after a minute" but it might actually take ten or twenty or a hundred minutes for the CPU to reach a equilibrium which is already well above the max temperature you want it at.

If you have a metal case, then blowing any kind of air over the case itself (setting the laptop upside down will help a lot too) will probably solve all your problems and them some.[/QUOTE]
Ah, that makes sense now! The case must be absorbing enough heat during the first 4+ hours to significantly improve performance. After that, the i7 adjusts its throttling cycle for its new (reduced) thermal dissipation capabilities, hence the increase in iteration times. Putting the computer to sleep briefly would allow the case "heatsink" to cool off rapidly and give you some fresh absorption capacity.

That's very interesting that they're using the case as a heatsink like that. I guess they figure that most people don't use their laptops that heavily for that long, so pumping the excess heat into the case allows them to skimp on more traditional cooling hardware. Incidentally, having that much external heat-sink surface area opens up a lot of possibilities for "manually" helping it along, which is not something you can readily do very well with many laptops. In addition to running air over the case helping a lot, the ice pack trick should be extremely effective (albeit temporary). A good-sized cold pack with flat contact to the bottom of the case would increase the heat absorption capacity by a ton. Even after it's melted, the pack will continue to absorb heat in the same manner as the case itself.

I would note, with regard to putting the laptop upside down, that you do want to be careful not to leave the laptop lid [I]closed[/I] while it's crunching full bore like this. Laptops dissipate a fair amount of their heat through the keyboard, and if all of that's going straight into the LCD screen it might cause long-term damage. What I usually do is crack the lid open at about a 30-45 degree angle, and stand the laptop up on its side. This will probably not work for all laptops (some can't stand up stably like that; and the power cable placement may interfere), but it's a quick and handy way to set it up to get some good airflow around the case. Another way could be to put it on a cookie sheet - though if you're planning to use the computer like this be forewarned that you'll have a "springy" effect when you try to type. :smile:

If you were going to put the laptop on a metal sheet for the heatsinking ability, I suggest that it be aluminum like a pizza pan. And just to mention it once more, a cooling rack such as you put a hot baked item on (cake, pie) would give you plenty of clearance for air. It might not even be too unstable.

[QUOTE=kladner;349084]If you were going to put the laptop on a metal sheet for the heatsinking ability, I suggest that it be aluminum like a pizza pan. And just to mention it once more, a cooling rack such as you put a hot baked item on (cake, pie) would give you plenty of clearance for air. It might not even be too unstable.[/QUOTE]
Oh, right...cooling rack, that was the word I was looking for, not cookie sheet. :rolleyes: That would make much more sense...(hence my reference to it being "springy" when you try to type on it).

I thought I would provide a little update.

As the main heatsinks had been running a bit warm, I decided to decrease the throttle setting from 85% to 75%. I think when I had raised it up to 85%, it generally increased thermal stresses in my system, and that resulted in my system showing degraded levels of performance more often and and to a higher level.

I have done some cleaning. Spraying compressed air at the keyboard didn't seem to have much affect. I took the bottom cover off. Overall, it looked fairly clean, but there was a couple of little dust bunnies by one fan and the inner side of the bottom cover was a little dusty/dirty. (I also discovered what was causing my DVD drive not to work, and fixed it. So taking the cover off was worth it.)

I also picked up a Targus Lap Chill Mat with dual fans to set the computer on.

Now the system seems to be running for long periods without going into the modes of lower levels of performance that I was seeing before. Of course, the "potential" throughput is somewhat reduced because of the lower throttle setting. (The displayed per iteration times do not take into account the rest time between iterations.)

Several days ago I started a double check assignment on my MacBook Pro. This is part of a plan to have at least one double check run on all my hardware since a successful double-check will give some validation that the hardware is working properly. This assignment is now over half done. When complete, this should give confirmation that despite the thermal throttling behavior and changes to performance that I'd been seeing, that the computer still functioned correctly. (I don't really expect any problem, but I still think this added confirmation will be reassuring.)

In summary, I think that the dropping in the performance level to a different relatively stable value was basically the hardware reacting to and dealing with thermal issues. While Prime95 appears to simply generate heat too fast for my system to dissipate adequately, it can still manage to maintain a good level of performance while going through short duration thermal cycles. But if the hardware fails to maintain the same level of performance after running awhile, then it seems to be an indication that Prime95 is generating too much heat overall and the throttle setting should probably be lowered a little.

Off topic: Early this month, I achieved a ranking of being in the top 1000 producers (overall, over last 365 days). My MacBook Pro contributes about half of my overall productivity.

[QUOTE=cuBerBruce;349298]I have done some cleaning. Spraying compressed air at the keyboard didn't seem to have much affect. I took the bottom cover off. Overall, it looked fairly clean, but there was a couple of little dust bunnies by one fan and the inner side of the bottom cover was a little dusty/dirty. (I also discovered what was causing my DVD drive not to work, and fixed it. So taking the cover off was worth it.)[/QUOTE]
Ah...I may have misled you a bit on that. Reading your post just now, I remembered that I had it backwards - you're supposed to blow the air [I]in[/I] through the vents and intakes, and the dust comes out through the keyboard area. :rolleyes: Sorry for wasting your time (and compressed air) there. It may be that your computer wasn't too dusty to begin with - the laptop I usually do that trick on picks up quite a lot of dust since it sits near the floor hooked up to a KVM switch, instead of being moved around like a typical laptop. My other laptop (which I carry around frequently) tends to produce much lest dust when I blow it out. But, if you're still looking to improve your thermal performance a bit, it might help some if there's some more dust yet in there.

This should be my final post regarding the slowdowns I was seeing on my MacBook Pro (except to respond to someone else who might keep this discussion going).

My double-check assignment completed with successful verification of the original LL test. So my MacBook Pro at least seems to run reliably despite the thermal cycling and the observed changes in performance level.

Since my previous post, I decided to raise Prime95's throttle setting from 75 to 80 (but still less than 85 that was the setting when the issue of very significant slowing down was observed). The displayed iteration times remained essentially the same for the next 45+ hours. Then a definite but relatively minor slowdown occurred on all workers. After noticing the slowdown about 4 hours later, and being the heatsink temperatures were relatively low, I set the computer into sleep mode for a couple seconds, after which Prime95 performance went back to the normal level. It has stayed at that level now for another day and half.

The performance slowdowns have seemed to be related to coping with heat that is generated. The lack of big slowdowns over the last few days seem to be due to the lower throttle setting that means Prime95 rests more between iterations, thus generating less heat per unit of time. The chill mat and generally more mild ambient temperatures over the last few days probably have also helped.

It might be nice to further understand exactly what mechanism is causing this slowdown, and also why after a brief period of sleep mode, the system doesn't tend to go quickly go back to this slowdown mode when Prime95 resumes.

But anyway, it appears that measures to improve heat dissipation, and using Prime95's throttle feature to limit heat generation (while also obviously limiting the possible performance that can be reached) does seem to help a lot in preventing these slowdowns from occurring. So I feel that I have the issue effectively under control.

[QUOTE=cuBerBruce;349899]Since my previous post, I decided to raise Prime95's throttle setting from 75 to 80 (but still less than 85 that was the setting when the issue of very significant slowing down was observed). The displayed iteration times remained essentially the same for the next 45+ hours. Then a definite but relatively minor slowdown occurred on all workers. After noticing the slowdown about 4 hours later, and being the heatsink temperatures were relatively low, I set the computer into sleep mode for a couple seconds, after which Prime95 performance went back to the normal level. It has stayed at that level now for another day and half.

The performance slowdowns have seemed to be related to coping with heat that is generated. The lack of big slowdowns over the last few days seem to be due to the lower throttle setting that means Prime95 rests more between iterations, thus generating less heat per unit of time. The chill mat and generally more mild ambient temperatures over the last few days probably have also helped.

It might be nice to further understand exactly what mechanism is causing this slowdown, and also why after a brief period of sleep mode, the system doesn't tend to go quickly go back to this slowdown mode when Prime95 resumes.
[/QUOTE]
I think you can chalk it up to the effect TheMawn described back in post #46 - the MacBook's large aluminum case is acting as a heatsink with a very large thermal capacity, and thus is "soaking up" heat when the system first comes on (that is, the non-slowed-down mode); after a while it "fills up", and the CPU can only dissipate heat as fast as it actually vents off the case, so it dials back performance to match.

It sounds like the chill mat has helped a [I]lot[/I] - as TheMawn had also said earlier, blowing air over that case-acting-as-very-large-heatsink boosts its dissipation capabilities quite a bit. Any time you have that kind of heatsink surface area to work with, airflow makes a huge difference. Hence, why now the system can stay in its "higher performance mode" for 45+ hours instead of 4+ as before - a very significant improvement.

As for the brief period of sleep mode letting it "reset", it seems that the case is able to quickly dissipate much of that "stored" heat when the computer is off (or in sleep mode, which is basically the same in terms of heat). I've noticed a similar thing with my desktop computer; if I simply turn off Prime95 (or similar program) and leave it running idle for a while, or even restart the computer without turning it off, the CPU temp never drops below a certain point, but if I actually turn it off, even briefly, it can cool down to a lower level and idle at a lower temperature before I start up Prime95 again. Again, just like with your laptop, this is likely due to the computer's heatsink (in this case a real heatsink on top of the CPU, not a large metal case, but the principle is the same) being able to much more rapidly dissipate its "stored heat" while there's no new heat being pumped in.

[QUOTE=mdettweiler;349902]I think you can chalk it up to the effect TheMawn described back in post #46 - the MacBook's large aluminum case is acting as a heatsink with a very large thermal capacity, and thus is "soaking up" heat when the system first comes on (that is, the non-slowed-down mode); after a while it "fills up", and the CPU can only dissipate heat as fast as it actually vents off the case, so it dials back performance to match.

It sounds like the chill mat has helped a [I]lot[/I] - as TheMawn had also said earlier, blowing air over that case-acting-as-very-large-heatsink boosts its dissipation capabilities quite a bit. Any time you have that kind of heatsink surface area to work with, airflow makes a huge difference. Hence, why now the system can stay in its "higher performance mode" for 45+ hours instead of 4+ as before - a very significant improvement.

As for the brief period of sleep mode letting it "reset", it seems that the case is able to quickly dissipate much of that "stored" heat when the computer is off (or in sleep mode, which is basically the same in terms of heat). I've noticed a similar thing with my desktop computer; if I simply turn off Prime95 (or similar program) and leave it running idle for a while, or even restart the computer without turning it off, the CPU temp never drops below a certain point, but if I actually turn it off, even briefly, it can cool down to a lower level and idle at a lower temperature before I start up Prime95 again. Again, just like with your laptop, this is likely due to the computer's heatsink (in this case a real heatsink on top of the CPU, not a large metal case, but the principle is the same) being able to much more rapidly dissipate its "stored heat" while there's no new heat being pumped in.[/QUOTE]

So far, I don't have any hard data to conclude how much of the stability is due to decreased throttle setting versus the use of the chill mat. I definitely do think the chill mat is helping to keep heatsink temperatures down, so you may be right in that it's making a big difference in keeping the Prime95 performance stable.

Thanks again (and to TheMawn for the earlier post) for your thoughts on this. I don't consider myself to be any sort of expert on thermal issues. I'll keep your posts in mind as I continue to monitor the performance I get out of Prime95. I'll soon experiment with adding a 4th worker while reducing the throttle setting to 60% to start with. With the same number of workers as actual cores, maybe this will help distribute the heat within the CPU better. But with 3 workers, I assume the OS has been bouncing the workers around the four available cores to get better distribution of the heat.