*OFFICIAL* Alienware M18x Benchmark Thread

Can I ask you guys, which software you're using to OC your GPU's or which would you recommand?

I am using Trixx at the moment and it works fine except that it doesn't save my fan control settings when I reboot my computer. It always reset it to automatic.

I checked the option to automatically start Trixx at startup.

 

Heres for comparison purposes. I hav not been able to even get close to the GPU or CPU limits due to PSU issues. Thought there may be some interest in these for comparison.

M17x R2 - i7 940XM (ES) / 6970 X-Fire

3DMark Vantage: P 22724
GPU: 24438
CPU: 18823




3DMark11: P 7234

 

Simply awesome, DR650SE! How did you manage to score 7700 on 3dmark11 for the GPU's? did you tweak the tessellation in CCC to off?

 

Yea, I have actually scored 7900 in 3DMark11 GPU, but during the combined test I was asking too much from my PSU, and my PSU throttled back from 280w, down to 200w, so it dragged my entire score down. And people say CPU OCing doesn't help in 3DMark11 It dragged my score down to 6700 or something like that, maybe even lower.

I didn't even save a screenshot of the 7900 GPU score because the overall was so low.

 

@DR650SE,

Amazing! Now, grab a 330W PSU from Dell and swap/mod the connectors

 

Lol, thanks, I need to find a way to make that work. I'm not even pushing the components. This is the best so far.

BOYAKASHA!
3DMark Vantage: P 23405
GPU: 24381
CPU: 20896

 

Lol, are those temps...chest freezer??

 

I case up some dice and put it on heatsinks. Didn't want to got directly on them. Especially given how wonky the cards have been acting. They're getting RMAed

 

question: My first run stock settings on 3dmark 11 is ALWAYS about 1000 lower than my second run. Is this normal? I got from 5600 to 6400-6800. Also, when benchmarking do you all have tesselation and "surface optimization" off? that seems to affect scores drastically.

 

I have my M18x set to Level 3, Bclk @ 103, and it is amazing watching it beat down my desktop in CPU (Single and multi) and tie GPU (6970M = Desktop GTX280). Vantage GPU, they both score ~10.8-11k at stock.

I was like, "Hey, my Desktop GTX280 isn't THAT old!?!" till I dug up my old newegg invoice and realized I purchased it Summer 2008 lol. Time to order a 580 for the desktop lol.

@ L3 / Bclk103, All 8 threads run at 3.61ghz rock solid. In Cinebench R10, it put the hurt to my OC'd 920 d0 @ 3.6ghz both single and multi threaded.

Ridiculous performance and very quiet under load IMHO.....

 

Yeah, I'm pretty shocked at how well these sandy bridge mobile CPUs perform. Its a beautiful thing.

 

I'm impressed with the CPU's as well. These things are beasts. The 6970s in CF are pretty amazing too. I did say it before, but to play Crysis/Crysis 2 on max settings on a freaking laptop is insane. I truly cannot believe how powerful mobile computing as gotten. I easily get 3 years worth of good gaming from each AW I buy. That is a lot of value over 3 years.

 

ha, awesome dude! I had a GTX280 overclocked to that fiery place and back in my old rig. Loved it! To have that power in a laptop.. times 2. Oh yes!

 

so where is the 580m sli 6990m xfire benchmarks?.........

 

My thoughts exactly! Everybody seemed so chatty until their M18x's arrived. Probably a good sign though, I guess...

 

AMD Radeon HD 6970M video card benchmark result - Intel Core i7-2820QM Processor,Alienware M18xR1 score: P23823 3DMarks



Not much of an OC

 

keep on pushing) 850/1100 i would aim for! try to download hwinfo64 to check your temps, it's free!

 

Highest I got to was 71.

I haven't repasted so I'm not going to go above this.

It's been 100% stable in all tests and Fall out 3 Vegas.

Other scores
AMD Radeon HD 6970M video card benchmark result - Intel Core i7-2820QM Processor,Alienware M18xR1 score: P6984 3DMarks

 

3DMark06: 21335 / Vantage: 21907 (GPU 23270 / CPU 18634)
Clocks: 815/1125

Spoiler

 

Here's the result from my stock 580 Sli rig that turned up on Monday - having to use the cruddy dell drivers still though......

Vantage 580SLI

 

Not bad at all, pal! Could you run 3dmark 11 and check the temps for both cards, please. After, you might wanna try the newly released Nvidia Verde 280.19 Beta driver from here

 

Thanks for that - just tried the installer - but I still have the problem of "no compatible hardware" ....

**Scrap that - just seen your other post about how to get around it - trying it now.

Cheers

 

Ignore please

 

Let's see if bar-code manages to get them work and provides us the installation procedure)

 

Yeh i've got it working now using bar-codes modded inf thanks

 

guys here is my stock 3Dmark 11 with 11.8 drivers AMD Radeon HD 6970M video card benchmark result - Intel Core i7-2720QM Processor,Alienware M18xR1 score: P6969 3DMarks

 

Wow, that's amazing! How's that possible? 7300 points for the GPU? it outperforms 580M by 20%? there should be smth wrong in here! Did you tweak CCC for the best performance and no tessellation?

 

AMD Radeon HD 6970M video card benchmark result - Intel Core i7-2630QM Processor,Alienware M18xR1 score: P20987 3DMarks

21K Vantage with stock everything RAM, CPU, HDD, 6970s etc.

OC'd to 800/1050.
Broke 21K ... yay...

http://3dmark.com/3dmv/3351273?show_ads=true&page=/3dmv/3351273?key=LnRJxFEjwTLka8bNHrvsnvr96NY06y

 

yes and here is one overclocked to 825-1000 AMD Radeon HD 6970M video card benchmark result - Intel Core i7-2720QM Processor,Alienware M18xR1 score: P7615 3DMarks

 

Hey, Speedy... Could you post 3dmark11 run with everything stock and no CCC tweaks at all, just raw performance and the same but with 825/1000 OC (no tweaks)?

 

here you go AMD Radeon HD 6970M video card benchmark result - Intel Core i7-2720QM Processor,Alienware M18xR1 score: P6362 3DMarks

 

Here is my (everything stock) OC'd 6970 to 800/1050 3dMark 11: 6900. I know its no biggie compared to some with the 2920xm.... but this is a 2630 ... with 4mb ram and standard HDD ...


AMD Radeon HD 6970M video card benchmark result - Intel Core i7-2630QM Processor,Alienware M18xR1 score: P6882 3DMarks

 

I went through all the posts I could find to try and wrangle up some numbers. Here is what I got:

moved to new thread: http://forum.notebookreview.com/alienware/598852-alienware-6990-vs-580-thread.html#post7756312

 

great job putting on the benchmarks together in one place

 

Things are looking good for the cheap people like me who went with the 6990's! Haha.

I want to see how far someone can OC their 6990's and remain stable, though.

 

It will be interesting to see how stable the 6990m is after 825/1150. In reality, some users such as myself and scook9 got shafted and picked up lemons that can't overclock well. So it will be intriguing to see if the 6990m is capped by heat or by clocks. Hypothetically, there might be users that get out benchmarked by 6970m xfire that hit in the 7900-8000 3dmark11 range.

By the way, is there any way to mess with the registry in windows to allow your cards to be pushed higher? My 3dmark11 score is terrible compared to others and I'm capped not by heat but by stability.

 

Räy - It will be interesting to see how the 6690 CFX performs. I have discovered the same things about the 6970 CFX. In fact, the OC ability of the 6970 CFX seems to fluctuate for me. Sometimes I can push 825/1150 without issues and other times I need to back down to 815/1125 for stability. This does not seem related to heat and I have encountered no problems with temps. They run nice and cool. I think it has some to do with drivers and maybe some degree of instability affected by the switchable graphics. Obviously, the registry has to have keys and database entries to provide handling instructions for both of the graphics subsystems, but only one of them is visible to the OS on the M18x. (On the R3, the hardware for both subsystems is visible to the OS simultaneously because on-the-fly switching is possible from what I understand.)

Anyone that has not installed the updated Intel HD 3000 driver should consider doing so, as the M18x works better with that installed than it does with Dell's version of the driver, even when running on discreet graphics and not the IGP. I anticipate that as the Intel HD 3000 and AMD 6xxxM drivers mature and become more suited to the switchable graphics platforms, we will see some improvements in stability. There are some limitations in voltage adjustments on the 6900 series, so the drivers probably will not increase the OC'ing overhead. For me that's OK, because the HD 6970M CFX handles all the games I am throwing at it on high settings just fine and the benchmarks scores are certainly nothing to sneer at.

The only two registry tweaks I have found are enabling (DWORD value 1) all instances of EnableCrossFireAutoLink and disabling (DWORD value 0) all instances of EnableUlps found in the registry. (Of course, most of us know that the latter is required to OC without problems.)

 

Here's some info I posted last night on another thread that may interest anyone that has not seen it...

 

weird Fox I was hitting 21k today with the 11.7 drivers ...best scores I have hit.

Also, randomly seeming, today I could OC farther then ever before, I could usually only run 775/1050 and stay stable. Today I ran 800/1050 and then 825-1050 ... nothing has changed, room temps the same etc ... just let me do it today ....

 

Yeah, not sure why it fluctuates. I'm glad you're getting better scores and it looks like you're getting essentially the same type of benchmark scores from 11.7 that I am getting from 11.6. As I pointed out earlier, I think the results will be more consistent as the drivers mature. I may have been able to get 11.7 to produce better scores with some tweaking, but I am more interested in Aero behaving properly than getting slightly higher benchmarks, and 11.7 and Aero definitely did not play nice on my M18x.

I just ran my second-best 3DMark11 score this morning. I was able to hit P6659 compared to my P6691 back on 07.16.11. I haven't tried to put my clocks back where they were yet, so I am still running lower clocks not than I was then.

Spoiler

Past Results:

Spoiler

Scores with AMD 11.7b RC/beta...


As you can see, the 3DMark11 score suffered most with the AMD 11.7 final release...

Here is a link to my previous (best) P6691 score: LINK
.

 

Not to be a stickler, but if you exclude the OC'd dual 6990m score from that set, isn't the average 22244.75? In that case, the dual 580m's score 1826 higher, which is 8.2% more. Of course, until we have more samples in the 580m set it's tough to be sure what we're really looking at.

 

That was a lot of work, syntax53. I agree with Speedy, good job. Thanks for putting that together for us. +1 rep.

In addition to what Pearson Park is mentioning, did you confirm all of the scores are with the same CPU? To have a meaningful comparison of 6970M CFX vs 6990M CFX vs 580M SLi, we need to be sure the scores are only from systems with a specific baseline CPU. There is going to be too much variance in performance with a mixture of 2630QM, 2720QM, 2920XM, etc. to get a clear understanding of what (if any) meaningful difference in performance there is between the NVIDIA-based and AMD-based systems.

 

Fox, have you found changing xf auto link does anything specific?? I can see the logic behind it being that xf is never disabled, right?

 

Hmm... I've got the same 2720 CPU and have everything set for performance and no tessellation. Wonder why the difference in 3DMark 11 scores? A spread of 1000 is pretty huge compared to my top score of 6691. I'd love to know what is affecting two systems with the same CPU and GPU so differently.

 

Yes, exactly stevie. In my earlier testing I did a driver-only installation. No CCC package was installed. I stripped out everything AMD, went into Device Manager and manually installed the drivers by browsing to the INF files. With the DWORD values changed to 1 (enabled), as you point out, CFX was working great without the Catalyst package. I did have (and still do have) RadeonPro installed and confirmed CFX was working without CCC, as the slider-bar overlay visibly confirmed this during benchmark runs.

I had no issues with Aero under that configuration, but it sort of sucked not having the ability to change any settings with CCC. So, rather than run without CCC, I simply disabled it from starting with Windows. I have noticed that once I start CCC manually, Aero get a little bit flakier until the next shutdown/restart. When I have not activated CCC manually for 6 days Aero has been holding strong.

 

Fox, have you got all the sliders in CCC set over to the far left with everything disabled AND all the boxes unchecked for options......thats a sure fire way to pick up more points on mark 11......

 

Assuming my settings stayed as I had them, yes they should be. I am intentionally staying out of CCC to make Aero behave, but that's a very good point. Thanks for mentioning it.

It could be that unless it is running at start up some of those options are enabled by default and CCC needs to run for them to be disabled.

I will double check the settings, temporarily ignore the Aero bug and re-enable CCC to run at Windows startup. I will run some benchmarks that way tonight and see what shakes out.

 

Thanks for the rep. I figure as long as we get as many benches as possible the CPU difference will average out as the GPU has the largest effect on the score. No, it's not perfect, but at this early stage I think it suffices. The reason I didn't exclude the OC scores was because the notebookcheck numbers don't say which are and aren't, and I definitely assume that some are. So again, I'm relying on averages and just trying to get as many numbers as I can from people.

 

I can't think what can be different other than the OS mine is a fresh install w7 pro sp1 with all the updates applied and drivers from DELL + ccc 11.8 beta

 

Don't get me wrong, I can see value in that, for sure. And, I agree that the averages are accurate as averages among a variety of systems. From the information you have compiled, we can make statements to the effect of "on the average" or "by and large" based on the data captured. There is definitely value in that. It's more than we could say before you gathered the data, but no conclusive statements are possible. (This is sort of like the "unscientific polls" where a newspaper or TV station will call people to ask questions. The answers are driven by the opinions of whom they called upon and the poll results are not conclusive in relationship to the nationwide public opinion of the masses.)

My only point is that we need to have some head-to-head showdowns between two or more identically spec'd systems (identical other than GPU), running at factory CPU, memory, and bus clock speeds to accurately declare whether or not a 580M SLi setup performs better than a 6970M or 6990M CFX setup, or vice versa. Until we have that, a statement of fact about who the winner of benchmarks is--AMD or NVIDIA--will not be possible.