W520 Quad-core sustained turbo frequencies?

I've been considering buying the W520 and I have a question for those of you who are doing things like video encoding or other tasks that make sustained (e.g. more than 10 minutes) full use of all CPU cores. Discrete GPU use is not necessary.

I would appreciate if people could tell me what frequencies they are seeing when all CPU cores are fully loaded for a sustained period of time with something like Prime95, video encoding, 3D rendering, etc. Please also state which CPU you are testing with.

I've seen reviews and comments stating that most every 15" quad-core laptop (including the W520) throttles when fully loading CPU+GPU. This is understandable as the 15" form factor has a limit to how much heat it can dissipate. The comments pertaining to only CPU loading tend to be vague (e.g. "No major throttling issues"), which could mean a lot of things.

Thanks!

 

Just so people know, I've found a post on Lenovo Forums that answers this question. Many thanks to huberh. Mark_Lenovo provides similar numbers in this post.

Why is this important to anybody? I'll say why it's important to me and it might help somebody.

I do some computationally intensive tasks for my work. I was thinking about building a desktop computer with a i5-2500K as the CPU. The 2500K has can easily sustain 3.3Ghz+ on all four cores with the stock cooler/fan.

When I saw the W520 with the 2720QM, I thought this might be a good chance to fully migrate to a laptop. In theory, the 2720QM can turbo to 3Ghz will all cores active. The difference between this and the 2500K is .3Ghz per core, which isn't all that much in the big scheme of things.

However, the tests show the W520 w/2820QM/2720QM settles at ~2.4Ghz when all cores are fully loaded. This means the 2500K actually has a difference of ~1Ghz per core. That's a ~40% difference, which is quite a lot.

I've read that the 2500K can also be easily overclocked to 4Ghz Turbo with the stock cooler, so the performance difference grows even more with that.

Please note that Prime95 with 8 threads is an extreme scenario and probably doesn't matter to most people. With 4 threads the 2820QM was running at 3Ghz, which is actually quite comparable to the 2500K.

Once you past 4 threads (more than 2 cores fully loaded) the situation starts to favor the desktop processors.

 

Couple thoughts/questions here, kto:

* in your referenced posts, the tester mentioned turning down the brightness on the display panel to reclaim a slight amount of juice. Would pressing Fn+F3, which turns off the display, be better? Or, perhaps, closing the lid enough to turn off the display during the heavy part of a job run?

* as far as loading down cpu cores affecting the chip speed, are the reductions only applicable if loading cores at 100% max? I don't do that, but would be running maybe at 80-90% total utilization. Almost maxed out, but some breathing room such that the cores would not be running at 100% all of the time. So, would turbo work, or would the laptop throttle back to much slower frequencies??

 

I don't think it's a power issue as the 8 thread run uses 95.0 Watts, which is well under what the W520 power brick can supply. From a cooling perspective closing the lid might make make the laptop run hotter and this may further restrict turbo.

Keep in mind that the CPU is not being "throttled" in this situation. The 2720QM has a base frequency of 2.20Ghz and the 2820QM has a base frequency of 2.30Ghz. Throttling only occurs if the CPU drops under this base frequency due to high temperatures.

What is really happening is that the CPU cannot reach full *theoretical* turbo frequencies due to thermal restrictions. Many other laptops also cannot reach maximum turbo frequencies - this is understandable since laptops have limits to how much heat they can dissipate.

80%-90% sustained CPU use is approximately 7 threads in Prime95. According to the thread I linked to the reported frequency for the 2820QM is 2.6Ghz. This is a .3Ghz turbo over the base frequency.

Keep in mind that Prime95 is an extreme test scenario. A more realistic test would be needed to see what happens when the CPU actually has some time (even if it's just a second) of lower CPU usage.

 

^ Aha. Thx, kto for the clarification. I'm new to this turbo thing

Appears then, that a fairly heavy load -could- receive turbo boost, particularly if one uses a cooler, which would be what I'd likely do.

 

According to that link all loads will receive some degree of turbo boost, but that you should have ~70% or less CPU usage to really see a practical gain from Turbo boost if you are doing something that takes a while (e.g. video encoding).

If your computing jobs take only a few minutes (e.g. 10 minutes) I wouldn't worry at all about any of this.

I suspect you are correct that a cooler will help to raise what turbo frequencies can be reached.

I'm sorry I cannot be of more help and provide more concrete numbers. I don't own a W520. Otherwise I would have run some more comprehensive tests to give you a more clear answer.

 

I think it's useful to note that the maximum design TDP of the 2720QM is 45W. The extent to which it is possible to exceed this may depend on how well the cooling system performs.

You should assume that the peak turbo frequency will be reached when only one core is in use. The maximum frequency progressively drops as more cores become busy (and use power / generate heat). I think the the quad core / 8 thread approach starts to provide dividends is with multi-threaded programs when the cumulative output of more cores running slowly is more than the output of fewer cores running quickly.

HWiNFO32's Sensor tab makes a fair attempt at reading the instantaneous CPU power consumption. It may be revealing to take a look at that under different core loading conditions (the program also shows the speed of each core). You can do this using wPrime and selecting the number of threads to be run.

John