Differences in x201 540M benchmarks

I purchased a x201 in Australia with a 540M and 4GB of memory and all the options To sum up I am getting benchmark results that are slower than those reported by this site (and Notebookcheck). I reformatted the laptop and installed Win7 64bit with the current drivers listed on the Lenovo site to do the testing.

I did have the battery installed during all the testing, and I have the 65W adaptor. Using a Kill-a-watt style meter I measured the draw as 46W peak on any test I could throw at it.

The only difference I can think of is I spec’ed the laptop with a single 4G mem stick. Maybe the 2nd stick accounts for the differences (can those who have 2 sticks comment on this for me)

I am getting some of my comparisons from Notebookcheck: Intel Core i5 540M Notebook Processor
_____________Observed_______Expected
SuperPi 1M_______15s___________15s
SuperPi 2M_______36s___________36s
SuperPi 32M_____887s__________851s
wPrime 32M_______21s___________18s
wPrime 1024M____654s__________576s

So you can see that the single threaded SuperPi doesn’t present enough of a difference to worry about. However, the multi-threaded wPrime does show statistical differences.

So I whipped out my trusty IntelBurnTest v2.5 and Tmonitor v1.03. IntelBurnTest is run with Threads set to Auto and Stress Level set to Maximum.

What I observed is CPU throttling. I did not get throttling when running wPrime. It seems I am running into the thermal limits of the CPU on this laptop. When running wPrime I averaged around 64C. This was low enough for the Intel Turbo to kick on every now and then and boost the multiplier to x21 from the max of x20. However, running the IntelBurnTest gives me an average CPU temp of 72C and the Turbo kicks in only at the very start. After about 10 seconds I start to see throttling. I can see throttling is occurring because the multiplier is rapidly changing on Tmonitor – it is not constant as you would expect to see when the CPU is 100% loaded. All my other CPU monitors are showing an average of 50% load across the cores/threads. After a few loops I am averaging 17.2 GFlops.

I tried to use the TPFanControl software to try and boost the CPU fan up to max, but it did not result in a major difference of temperatures.

I also downloaded ThrottleStop v2.0 (Build 227) to try and block the CPU from throttling – with no success. I can lower the multiplier lower than the maximum and the Core i5 will comply, however it will still throttle on IntelBurnTest with a multiplier of anything over x9.

I have a Core i7 750 in my desktop computer, so I am familiar with how IntelBurnTest will operate on a desktop CPU. I can assure you this is not normal on the desktop. It does not throttle under any circumstance.
So, now that I have presented some values I was hoping that others can replicate my results on their laptops to see if they observe the same results. As I mentioned at the start, I am running with one memory stick. That may be accounting for the differences in wPrime but some opinion from others would be appreciated

In terms of IntelBurnTest, I appreciate that it does not represent a test of real world conditions. It is specifically designed to be a CPU/memory stability test. I have no other software that is able to load the CPU as much as this. I just want to see if others are also getting throttling as well.

Apologies for the super long post, just thought I would be thorough.

 

while it is not exactly the same, but running 2 ram module setup does boost the performance of laptop.

http://forum.notebookreview.com/lenovo-ibm/490138-your-memory-score-t510.html

 

I see the point made in the thread you mentioned. Good to know about the graphics card performance increasing. It will be an upgrade I tackle in the future for sure.

Something just occurred to me based on teh linked thread. What if what I think of as heat throttling in IntelBurnTest is actually a lack of memory bandwidth?

I presume since I am running one stick of memory that I have half the bandwidth available. IntelBurnTest is a CPU and Memory test. What if the single stick is starving the CPU of data from memory in this particular test? It may be throttling itself automatically for lack of data and not heat related issues.

Maybe if someone with 2 sticks of RAM can run the program and verify if they see the same behaviour with IntelBurnTest and Tmonitor?

Thanks again.

 

I have a X201 540M with 2x2GB, Windows 7 32Bit, Lenvovo install with other crap added as well. Right after a reboot it does wPrime 1024 in 667 secs. I then swapped the RAM with a single 4GB, and after a reboot got 662 secs.
I would say that was within the margins of error.

Not the precise test you wanted, but it was a direct comparison.

G.

 

Thanks for doing that Gunge.

It seems that for wPrime that memory bandwidth accounts for very little of the performance. Your result was also in agreement with mine.

It still irks me a little that there is a difference between the benchmark results that Notebookcheck reported on and what we are seeing.

It is a good laptop, but I would still like to get the maximum performance out of the CPU.

 

And that's why Notebookcheck should be taken with a grain of salt

 

I wonder about things like ambient temps. I know back in my brother's biochemistry days, there were issues with things being done in India in unairconditioned labs giving unreproduceable results in controlled labs because nobody noted the conditions.
They also don't say which version of wprime in the test.

 

It isn't right after a reboot, but I just tried wprime out of curiosity and I got ~21.4second for the 32 test and 679 for the 1024 test (though I might have skewed the 1024 test results by pulling up my auto-hiding task bar a few times to see what the cpu temp was).

The interesting thing is that in use Turboboost will take the CPU to a higher speed than Wprime does, so maybe it's just another instance of benchmarks being an inaccurate indicator of performance.