Overclocking Performance - 3DMark
Everyone is basically familiar with the 3DMark benchmarks, and while they aren't games per se, they provide a relatively consistent way of comparing system performance. They are also useful in that they run some CPU tests that tend to put a heavy load on the processor for a relatively long period of time. On the later versions of 3DMark, the CPU tests are also multithreaded, so they can utilize 100% of a dual core processor. After the QuickTime H.264 results, we wanted to see if other lengthier CPU stress tests would show similar behavior.
The GPU oriented tests all demonstrate the same thing: 3DMark is a great way of seeing just how fast your graphics card is. None of the graphics tests show anything more than a very slight increase in performance from the increased processor speed. If these results correlate to actual gaming performance, we shouldn't expect to see much difference in performance at the various CPU overclock settings.
On the other hand, the CPU tests show exactly the sort of behavior we expected based on the QuickTime results. In all three CPU tests, Bin-5 scores lower than Bin-4, and in 3DMark05 and 3DMark06 Bin-3 is the best setting for maximum CPU performance. If our testing with Folding@Home holds true for all situations, users shouldn't experience CPU throttling at any of the overclock settings up through Bin-3, and it appears that some of the more CPU intensive applications may actually result in enough thermal throttling that Bin-4 and Bin-5 end up running slower than Bin-3. However, this appears to only hold true when both CPU cores are placed under 100% load for an extended period of time.
Everyone is basically familiar with the 3DMark benchmarks, and while they aren't games per se, they provide a relatively consistent way of comparing system performance. They are also useful in that they run some CPU tests that tend to put a heavy load on the processor for a relatively long period of time. On the later versions of 3DMark, the CPU tests are also multithreaded, so they can utilize 100% of a dual core processor. After the QuickTime H.264 results, we wanted to see if other lengthier CPU stress tests would show similar behavior.
The GPU oriented tests all demonstrate the same thing: 3DMark is a great way of seeing just how fast your graphics card is. None of the graphics tests show anything more than a very slight increase in performance from the increased processor speed. If these results correlate to actual gaming performance, we shouldn't expect to see much difference in performance at the various CPU overclock settings.
On the other hand, the CPU tests show exactly the sort of behavior we expected based on the QuickTime results. In all three CPU tests, Bin-5 scores lower than Bin-4, and in 3DMark05 and 3DMark06 Bin-3 is the best setting for maximum CPU performance. If our testing with Folding@Home holds true for all situations, users shouldn't experience CPU throttling at any of the overclock settings up through Bin-3, and it appears that some of the more CPU intensive applications may actually result in enough thermal throttling that Bin-4 and Bin-5 end up running slower than Bin-3. However, this appears to only hold true when both CPU cores are placed under 100% load for an extended period of time.
Facebook
Twitter
RSS
20 Comments
View All Comments
Tommyguns - Friday, May 11, 2007 - link
$4000 for a laptop to play games on?!Not trying to bash the artical, but whats the point of testing products if the actual market niche for such products is sooooo freaking small!?
$2000, on a laptop is a lot of money. adding in the fact it needs to play games, do that many people really spend past 3k on one? let alone 4k+?
If 10-20total fps in DX9 games is worth $2000, then someone let me know...
Dell inspiron E1705 with a C2D T7200, 7900GS vid card, gig of ram, 9cell + 6cell battery, DVDRW, vista home pre., 17 inch UltraSharp Wide Screen UXGA Display, wireless N card,and a few other odds and ends can be had for $1875. Toss in one of the coupons available online, like the $500 off and its just shy of $1400 for a decently effective gaming laptop. is it the best buy? idont know, but the best well rounded one. also, the XPS 17inch laptop on dells site, with the same parts, retails for around $2500....wtf? yes, no coupon...but still.... I like dell, yet dislike them equaly as much.
Thus, review a 7900gs dell offering so us poor people can see what an additional $1200+, really can do...Thanks... I would buy tommorow if i knew what all these machines did in terms of gpu power
JarredWalton - Friday, May 11, 2007 - link
Part of the point of the article was to show that the added CPU speed did little to nothing for gaming, so if you're interested in a gaming laptop I'd recommend getting something like the XPS M1710 but sticking with a T7200 and just maxing out the GPU. You can do that for under $3000. Also remember that the configuration as tested has a $500 Blu-ray drive thrown in that many won't care about. So, cut down the CPU ($700) and the BRD ($500) and it's a lot more reasonable.ziddey - Wednesday, May 9, 2007 - link
Hi, this is definitely an interesting article, but I was a little disappointed in how it was presented and the testing done on it. I hope to not offend, but if you still have the ability to retest further, here's some ideas.So we're talking about throttling. If that's the case, how hot is it getting? You can track coretemps of meroms. The popular rmclock can also show any form of throttling in log form as well, and also shows it in graphical format as well. It also features ability to manipulate some finer features of it, although of course, at your own risk.
Further, I'm assuming this is only possible on upwards unlocked cpu's. In that event, could you just use something like crystal cpuid to change the multiplier?
Most significantly, how about this: If temperatures are still in line (and based on the track record of conroes, I see no reason they shouldn't, as long as vcore is as low as merom's are), can you set to maximum overclock in bios, and then use rmclock to force maximum multiplier while on AC
Deusfaux - Wednesday, May 9, 2007 - link
either you're homeless or you have an excessive amount of money and a butler or i dont know whatbut for all the reasons that make a laptop advantageous over a desktop, you strip them all away by attempting to make it one of the gaming variety.
ridiculous cost
battery life
size
weight
hardly upgradeable
I feel bad for the continual string of people who are convinced they can get all the benefits of a powerful desktop in a portable package (and not end up being inferior in every aspect)
Get a laptop that has good portability, get a powerful desktop for gaming. dont try to do both in one.
araczynski - Tuesday, May 8, 2007 - link
sounds like its a nice laptop for 1280x800 gaming. curious what the screen looks like at that resolution, in comparison to the native resolution 'sharpness'. Does it scale very well or look as crappy as you'd expect the desktop to look at that resolution?suryad - Thursday, May 10, 2007 - link
More like 1920 x 1200 gaming all except for Stalker which I dont care for much anyway. I am getting this bad boy tomorrow or next week Monday. Can't wait. Worst part is the wait to receive it in your hands...hope i can last the weekend. Also the 3dmark scores area bit low...ntoebooforums people are reporting slightly higher scores...but anywho...super speed here I come!kalrith - Tuesday, May 8, 2007 - link
The spec page on page 2 shows *q* every place that quotation marks should be.Gary Key - Tuesday, May 8, 2007 - link
Fixed, sorry about that, we are on a new engine now, bugs still seem to pop up. ;)JarredWalton - Tuesday, May 8, 2007 - link
Heh... Gary's being nice to me. I screwed up my own table by reusing parts of an old article and forgetting to properly "convert" the generated HTML file into a clean format. That's what sleep deprivation will get you.redfirebird15 - Tuesday, May 8, 2007 - link
Is it really possible to get thousands of FPS on Supreme Commander? I've never actually played it, but that looks like a typo. If that is correct, what is the difference between getting 500 FPS and 1000 FPS? I thought it was and RTS anyway.It must be a typo.