G51JX-A1/X1/X3 GTS360M performance significantly lower than G60JX GTS360M - Investigation Thread

Result in Last Post - The Difference in Performance was because the G60JX must of been running 3DMark06 at panel maximum 1366x768 rather than reported 1280x1024 therefore there was a 20% difference in Performance.

Thanks all who replied here, pm'ed or replied in other threads.

Below is the original data/thread as it was before the answer was found:

-20% Stock Performance @ Same Temps (G51JX vs. G60JX)
Minimum -20%+ Maximum Overclock Performance @ Same Temps (G51JX vs. G60JX)

I noticed in the G60JX Best Buy thread a few days back that the G60JX users were reporting that they were getting GPU performance 20% higher at stock speeds than G51JX users had reported. The G60JX systems were also able to overclock the GTS360M GPU up to levels of 650/1900/1500 and get 11,900 3DMark06 points - performance also 20% higher than G51JX users are reporting (and note that # include a slower CPU than the flagship G51JX-A1 model has) and with an overclock much higher than the maximum 615/2000/1600 that the G51JX with the same GPU appears to be capable of. All G60JX and G51JX systems use the same GTS360M GPU.

So far we can confirm we are both using the same drivers (namely modded 196.86). PSU and BIOS, VBIOS differences are still up in the air - as is the potential that the cards operate on different voltages.

I'm investigating it now to see if there might be some difference in VBIOS, voltage, drivers, or something else that might be causing it. The G60Jx is nearly identical to the G51JX-X3 systems, except with only 2 RAM Slots (4GB preloaded) and a 16" 1366x screen instead of our nice Full HD 15.4". The G51JX-A1 has an i7 CPU different as well and 4 RAM Slots. These systems should definitely get the same performance with the same video card. Note that the G51JX and G60JX both use the G60JX motherboard (though there could be sub-model differences).


Theories of Cause:
- i7-720QM CPU of G51JX-A1 requires more power than the i5-430M CPU of the G60JX - False, since the G51JX-X3 has the same processor and the same lowered performance (could it be it shares some kind of limit throughout the G51JX series? - Maybe)
- GTS360M ASUS MXM modules are inferior to those in G51JX models
- Differences in preloaded software
- Differences in driver (False, tested)
- Differences in operating voltages
- Difference in BIOS
- Difference in Motherboard
- 16" 1366x768 would allow for the 20% better performance and difference in overall overclock because of the fewer pixels to be driven (all G51JX systems have 15.6" FHD screens)

Here is a comparison of the stats btw:

viper3885 (G60JX):


My Stock G51JX-A1:


viper3885 (G60JX) Overclocked to 650/1900/1500 w/same temps as before:


My Overclocked G51JX-A1:


The performance and OC differences are real and have been reproduced on multiple G51JX systems. The G60JX reported performance matches up more closely with what the GTS360M was advertised as being capable of (eg. 11,000 3DMark06 points).

Note also that the temperatures of both are approximately the same, it's just the performance values that differ.

G51JX-X3 users with the i5-430m processor have reported similar CPU values while also showing the CPU performs the same and with the same temps. The i7-720QM of the G51JX-A1 operates at 15% better as expected. It is just that the GPU values are lower in both cases.

I am collecting all the findings and facts I have so far into this post and the following one, and will update the info as more becomes known. It would be great to do a 3DMark Vantage test on the G60JX and compare with the G51JX results.

Any reports, ideas, anything are welcome.

Thanks,
Peter

 

Reserved just incase.

As reference for people, to get 196.86 drivers working on the laptop use the inf below - however this does not appear to help the performance.

G51JX-X3 3DMark06 Results:

Another G60JX Benchmark Result from another user:

Link to one of the original posts/detailed system comparison:

Original post in G60JX Best Buy thread: http://forum.notebookreview.com/showthread.php?t=465562&page=5#49

 

I can't confirm this being the cause, it's an interesting theory that had not occurred to me. The best way to check this in my mind was to compare the previous G60J which has the same low res screen and the G51J which had the same GPU but the smaller Full HD screen.

In the results I've found, the difference in performance between the two models is not visible and does not apply the same way as in this case. I will continue to research this though, the results I found so far:

Stock Clocks:

http://www.notebookcheck.net/Review-Asus-G60J-Gaming-Notebook.22357.0.html
G60J: 3DMark 06 -10199 points (All Stock) - 1280x1024
G60J: 3DMark Vantage - P4900 points

http://forum.notebookreview.com/showthread.php?t=427948
G51J: 3DMark06 - 10168 points (Stock GPU (stock drivers), CPU on turbo mode: ) - 1280x1024
G51J: 3DMark Vantage - P5228 points

Other results:
After system cleanup - http://forum.notebookreview.com/showthread.php?t=457808 G51J: 3DMark 06 - 11,125 points

This theory could still be valid if there is some difference in how things work in this new model, but historically it does not seem to match the facts imho.

Peter

P.S. To clarify all the results in this thread were specifically run at 1280x1024 including the NC result on the G60JX (done by forcing 1280x1024).

 

Uhh... are the benchmarks run at the same resolution?

 

Yes, all the benchmarks I posted were run at the default 1280x1024 resolution in 3DMark06. There is also a 1366x768 result as well, but because of the resolution difference I didn't include it - thanks for your reply . I'm assuming as well that none of the other 3DMark06 settings were changed from default - and the preference is for getting data from G51Jx/G60Jx results from stock systems with as little differences as possible from the initial config (my tests were preformed the same day I got the system with only minimal bloatware removal).

Peter

P.S. I didn't mention above since the difference in the OC is also up for comparison but the Stock Clocks on the two systems are the same 550/1800/1323.

 

Actually, it's true if the G60J/JX has a max 1366x768 resolution, Notebook Check reports the 3DMark06 test was run at 1280x1024, but how can it be? I will run my own test at 1366x768 now to compare.

Thanks,
Peter

 

Update: I think that all the tests on the G60J and JX despite people reporting them as being run at 1366x768 or a similar resolution rather than the 1280x1024 default that all computers with higher than that resolution default to.

Therefore if you add it up:
1280x1024 = 1310720
1366x768 = -1049088
= 261,632 = 20% fewer pixels. Therefore 20% better performance.

Comparison of 3DMark06 1366x768 results:
G51JX: 10058, SM2.0 4265, SM3.0 4305, CPU: 2988
G60JX: 9911, SM2.0 4318, SM3.0 4429, CPU 2591

Therefore within normal variance.

There is the fact that he Viper can overclock further than me, but it may be just because he is running the test at the lower resolution. So this case is settled, QED .

Peter

 

You know, I don't think there is a linear relationship between 3dmarks and resolution. At lower resolutions, the CPU makes more of an impact, etc.

Anyway, this is expected, the GTX 260M was underclocked but I think the GTS 360M is at designated speeds because of the reduced TDP. At stock speeds the GTX 260M should be a bit better than the GTS 360M just looking at the specifications, though.

 

I just came to say, like in the other thread, that performance does not scale linearly with the ammount of pixels present.

I gave a lecture about it before but was a long time ago haha.

The topic still stands, the G60JX was getting higher numbers because of the resolution, though it was not 20% performance difference at all.

 

what is this thread about? The 3dmark of g60jx was obviously run at 1280x768.

/thread

 

And you have your result in the other thread - and the reasoning behind it. It was not a theory I came up with, but it is one that fits the facts. Nothing else but the resolution is different. It may not scale linearly but in this case it did correspond within 5% by actually calculating the #s (which themselves have some variance of course) so it works . I'm not saying it's exact 1:1 - sorry if it sounded like that - but if the only thing changing is the # of pixels in this test it explains the difference in the actual results .

Update:
~30% difference in pixels and the difference in pure GPU performance at stock was approximately 13%. In the overclock tests, the difference was closer to 20%, but it could get much higher depending on the quality of his chip and factoring in that I maxed my GTS360M out at 1280x1024 while he just set a random overclock and tried it out. I was originally assuming that he was running at 1366x since I lacked any data but 3DMark06 was running 1280x720 instead so the pixel difference was even bigger. But to me I'm not trying to argue a theory, just trying to explain the facts so I'm good with it either way - Significantly Fewer Pixels = Significantly Better Performance - Problem Solved

Actually they were run at 1280x720 I've been told, but it's a similar effect. In 3DMark01, 03, 05 and 06 there used to be a note saying "This test was NOT run with Standard Benchmark settings" - however it seems that is no longer visible. You can run the tests at the default 3DMark06 res value of 1280x1024 by using an external monitor, which for the sake of comparison NC and other review centers do. I had assumed since no one said they were getting that note that they had run the test at the Standard Res - but since that is not the case and I didn't know it, I created this thread.

If you read the top post though you'll see I already said it's solved so, yes we can definitely close it .

Peter

 

Oh I am not saying that performance does not increase or anything like that. It is based on facts: Performance does not increase in a linear manner to resolution. You have your facts indeed, as you have shown, 30% pixel difference was 13% performance difference.

Basically, a fixed % number of pixels =/= fixed % performance decrease.

But yeah, less pixels = more performance, and the same holds true the other way around

Now I wish I had a GTS360 just because it runs much cooler But in secret I long for an HD5870m

 

Any findings in the previous few days?

 

My g51jx-a1 stock 3dmark was: 9525. I'm also a bit disappointed by that as I've expected it to beat 10k mark...

 

Hey mahdi,

That is a normal result - in this case users with G60JX models were doing tests at 1280x720 and giving the results without stating the resolution wasn't default and it's been adding to confusion in this thread and in the G60JX thread itself. If you run the test at that resolution you will also find you get a score of 10000+ right away.

You can check the G51JX thread or my review here to see how to overclock the GPU to improve the score. We get slightly higher scores than a stock G51J running at the same temperatures and using less power - and actually with 3DMark Vantage we get even better scores (even with PhysX disabled). No it isn't as powerful as the GTX260M, but it likely will last longer, can be overclocked further, runs cooler and therefore quieter and takes less power.

Peter