M17 OC Benchmarking - The Next Generation!

you could push it a few more gigz here. we have -30C.

 

i'm definitly trying to see how far i can push my q9000 outside tomorrow

 

Drake, are you just upping the CPU voltage until it's stable? You've got to be using FSB by now. You're out of multipliers!

 

might want to submit that to CPU z for the recored


also find some 4850's at any cost

 

Yes multi and FSB both changed.

There's a cpuz record?

 

SuperPi 2M. My M17x did 33 seconds.

 

to verify the speed

HWbot ... CPUz validator ..... etc

 

run 06 and lets look at those cpu scores. Also you should be able to break 14k on vantage cpu now. Unless you have stuttering.

 

That is better than 920XM! Hooah!
ON par with i5 750 Lynnfield. Scratching the Bloomfields!
Go for 3Dmark! It should be even more interesting.

 

DAM SON!

Lets see some benches!

 

I'm inside now. Got too cold! sorry.

 

called a coat .....

 

I was wearing multiple layers. I went ahead and lowered the FSB down a bit and did a run since you guys wanted to see it.

 

who's m17 did you pick up ?

nice score BTW you really need 4850's you should of got the ones off Ebay a while back

 

you should of overclocked the gpu's. I know you want to do it.

 

Just for a reference : stock desktop i7 950 score is 5300-5400; i7 920 is around 4700-4800.
I'd say - you rock!

 

I bought it from Alienware off of eBay.

Maybe my next outdoor benchmarking session.

lol, thanks!

 

Congratulations Mandrake! What speed would you say you're willing to run on an everyday basis?

 

I'm only doing this for benchmarking. I'll reset my bios when I'm done. I'm currently running 3.8Ghz and I'm not having any heat issues but I haven't tried gaming with it.

 

nice job. Which one do you like better the whitebook or the allpowerful?

 

Whitebook if I had 4850's

 

I like them both.

 

Allpowerful = Mandrake

For most of us, it's a task to try to get there, but for Mandrake, it's a walk in the park !!

 

Thanks. I wish it was a walk in the park. Clipping copper a couple days ago, sitting out in the cold. Testing different multipliers, fsb, memory (speed and timings) and cpu voltages. It's been time consuming.

 

I was about to say it sure does not sound like a walk in the park to me lol... Awesome job though ..

 

I have a strong suspicion that QX9300 in M17X could be OC'ed to the same level if it wasn't so brutally undervolted.

 

I've been investigating this. The chipset is limiting but I have a few things I want to try when I swap procs again. Right now I'm not running an extreme proc in my M17x.

 

there is better cooling, but the limitation of the chipset...

The intel chipsets have ALWAYS overclocked better. 4GHz is pretty good though for a Core 2 Quad....ANY Core 2 Quad (desktop or laptop)

 

I'm surprised no one here has attempted this before since the system is over a year old.

 

No one got the XTU to work, that is what allowed you to go so far

I would have been ALL over this in July if we had found a working XTU

 

What's been the problem with XTU?

 

I am not sure, people knew what it was but were unable to get it to work on the M17. I think no one had tried to use desktop versions. I was much less matured as a laptop bencher at this time unfortunately

 

me to.... me to....

 

Nvidia chipsets suck for overclocking

 

They just lack the tools for laptops.

 

well we have nst that brought it up alittle. there's nothing we can do about that. and i have looked for nvidia programs where we can change the volts and there isn't any. nst is the only program but it is locked out.
my 7600 gt desktop has more features then the m17x. unless there was somewhere to edit the bios for the m17x, m17x can only which how the older model out paces the newer one.

 

Except the desktop chipsets by nvidia... I had a 650i up to 2000mhz on the front side bus, pulling one of the first conroe chips from 2.1 to 3.5ghz. Did that for about half a year when the rig SELF DESTRUCTED!

 

put your rig in your sig

 

Intel chipsets could usually go as high as 2400 MHz......(the x series, not the P)

 

No such luck with the T9900 (for obvious reasons) but at least I can change memory timings and the FSB. Much better (stability) than setfsb.

 

Roughly 30 minutes into OCCT with no issues. This can run 3.5Ghz all day.

 

Awesome stuff, HOWEVER I'd recommend using either Intel Burn Test OR OCCT's Linpack CPU test for stress testing, rather than OCCTs normal test, it's much hotter and more intensive under 64 bit than most other testing mechanisms (hell its used by Intel).

It also tends to tell you whether an overclock is stable much more quickly. 5-10 runs on maximum stress with IBT is pretty much a surefire on whether a CPU is stable or not. (and usually only takes half an hour for 10 runs full stop!)
I've had CPUs 24 hours prime stable, fail or reboot with IBT in 5-10 minutes!


Dont get me wrong or take me as belittling your results, Im not that's an awesome clock, but IBT/Linpack would save you time (and is MAN's test for your CPU haha)!!
IBT is also a pretty damned small programme so easy to run, install and use (smaller than OCCT, and just needs unzipped)

 

lol, thanks. I also used the stress test that comes with XTU. Don't know if that's the same thing. Regardless if it can complete a benchmark then I call it stable.

 

Haha, if you can complete Linpack 64/IBT 64 bit max stress 5 run then your CPU is pretty much solid. Dare you try and see what your temps hit

I've seen it push temps 5-15 degrees higher than prime(dont bother with 32bit version, tis weaker)

If it ain't Linpack64 stable, it's not stable

Only takes 10-15 mins so what's your worry

 

You can always find a better way to stress your gear.
At some point you have to stop

 

Thing is Linpack is hard AND fast (oo-er!)-er than most of the other stress tests. I'd say that's win win. The fact I can stress test a CPU in less than 30 mins compared to 8/12/24 hours of prime/orthos/occt etc is a godsend.

Linpack is used by Intel to test thier own chips which I think says something for it.

 

Convinced me. I'll give it a shot.

 

Intel Burn Test is the easiest to use version (and on version 2.3 last time I checked), just download, unzip, open it, set to maximum stress (it defaults to 5 runs), press go, and then see if it finishes successfully.

If you machine restarts/bluescreens or similar, then your system isn't stable somewhere along the line.

Watch temps too...it gets hot, pretty quickly!

 

that is not entirely true...

i use to think so till, i was set straight on it all.
64 bit os with 6 gigs of ram...your work size needs to be 25000 @ 20 passes
64 bit os with 4 gigs of ram...your work size needs to be about 18000 to 20000 or max ram allowable
and it has to run for at least an hour. @ 20 passes
5 or 10 minutes does not cut it. trust me, i tried...lol

 

Thats why you need to use maximum stress (which selects all useable RAM), and yes there's a chance of finding errors after say 2 hours, but even so what I said still stands, its much more demanding and shows errors much quicker than the likes of prime

There's always a point where more testing is good, but you get what I meant

A reasonably unstable machine usually wont last more than 10-15 minutes with linpack, and knife-edge machines won't last 5 in general, it'll have shutdown/restart or failed pretty quickly, whereas it might have passed prime/orthos for a few hours (or a 3dmark06/vantage run), linpack shows the errors up that much faster.

Testing for 30 minutes at max stress gives you at least a good idea it's reasonably stable, and will kick out a lot of machines that may pass 8 hours of prime, and will generally be absolutely fine for day to day use, this is what I was getting at.

Even if you test for 2 hours (or overnight as I've done before), there's a tiny chance it may error further on, BUT it picks things up quickly in comparison to the others, and I've never had a machine overnight linpacked become unstable, I cant say the same for prime/orthos, which means I really couldnt go back from using it as my stress-tester of choice. Time is money and all that