Throttle stop help 920xm :S

Hi! I'v been playing with TS for a day or two now and finally think i got it working.I have clock,chipset mod ticked set multi ticked and set to turbo and i have EIST ticked set to 26 across the board and my TDP/TDC = 90/80.
MY problem is performance in games is actually worse if i turn TS on no matter if its over clocked or monitoring or if i turn it off and close program still bad performance.If i reboot and don't turn TS on i get the performance i would expect.
Does anyone have any ideas why this could be?
Forgive me if i haven't gave the correct info TS is a little bit confusing to me :S

 

great now run wprime 155 @ 8 threads to test it out. if it show 26x throughout the whole test then you are good.

 

It does show 26 throughout the test but still im seeing a big drop in gaming performance 15-10 fps when TS is on or has been on.
I can see a improvement on the TS bench No overclock=22.659s on the 32m test
with overclock its 13.114 seconds...just the games don't work :S..Oh and thanks for the super fast reply again xD

 

hmm something seems off with those numbers. with ts enabled @ 26x you should be under the 10 sec mark. wished i had backed up my 940 wprime benches. hopefully someone else can chime in.

 

It might be worth mentioning wprime cant detect my hardware so i have to skip this part by pressing shift and i installed TS when i had my i5 540m installed could it be worth re installing it now? And if i turn TS to monitoring mode and run the test i still get 26 in the fid column

 

Did a quick run of the TS benchmark with my 920m. Got 19.589 seconds at stock settings (24-23-17-17) (55W/47A). At 26x flat (92W/80A), it took 12.99. It doesn't matter whether TS is on or off insofar as the multiplier and TDP settings are concerned; once set in TS, it stays set until you change it or reboot. You're running a little slow, but not dramatically so on the bench. I suspect the game slowdown is either heat throttling or power starvation caused by adding the power draw of the 5850m, particularly if o/c'ed, to the power needs of the o/c'ed 920xm. Log your CPU performance with T-Stop (monitoring mode) and your 5850's performance with GPU-Z while gaming. That should tell the story. A 920xm at 26x running under load generates serious heat in relatively short order. For gaming, you're usually better off with the CPU at or near stock, particularly if the GPU is overclocked.

 

I agree, you're just drawing too much power from the system and it's probably getting too hot as well.

Throttlestop has four different profiles, set one up for max CPU and another with the TDP dropped down a bit for gaming. There isn't a game out there that isn't going to bottleneck at the 5850m first when you're talking about 3GHz+ on an i7. I run 65TDP for boot and gaming, 80TDP for when I'm working on the desktop.

 

thanks rev. always thought it was lower for 26x.

man these new cpu are impressive, just ran wprime and got 8.419 sec for the 32m run.

 

Remember that's the TDP +25% short term boost though. Run the 1024m and see what you get, I'm curious.

 

That is impressive. 920xm has to run at about 22.5x to achieve that score. Sure am looking forward to your experiment with the 6970m and the M15x. That combo with a 920/940 could create a beast if it can handle the power needed to run it all.

 

this is the 1024m run. pretty good imo.

 

Nice indeed! I'll post a run as soon as I'm done teaching this class.

 

I think so too. The nicest thing is the temps. My 920xm can blow you out on the 32M speed test, but can't keep up on the long test due to heat throttling. I need new paste, but the 920/940's always bounce off their thermal ceilings when put under near full loads with all cores/threads heavily o/c'ed. The saving grace is that it rarely happens in the real world. What's really impressive to me is that your 2720 (thought you were going to order a 2820?) is showing almost 20,000 CPU in Vantage. I recall that I was trying to break 20K/CPU in Vantage when I blew up my first motherboard. I was at 28x or 27x when it cashed out. So that's about what it would take to equal your score. But the 920/940 at those settings will finish wPrime 32M in under 7 seconds. It's not yet clear to me how this reconciles. What's your theory?

 

80TDP/70TDC 23/24/27/27 That's my usual desktop setting.

 

Very nice, JohnnyFlash. Did it hold 23x throughout the 1024M run without throttling? +Rep

 

@ Rev Did a TS bench 32m 24-13-17-17 55w/47a score was 19.870.
26-26-26-26 92w/80a score was 13.304.
I'v run the logs like you suggested and it seemed like the CPU and GPU would throttle back at the same time I expected to see the CPU staying high and the GPU going low.GPU was not overclocked
Taking the advice you guys have given before i just run the CPU at stock for gaming,i mean its great at stock :O I'm trying 60w/50a 24/24/18/18 and gonna run the GPU at 800/1050...does this sound safe to you guys? or an tweaks you might suggest...Thanks again for the help

 

wow nice run guys. still would love to see the 2920 in action. was thinking of getting one, but way to expensive at the moment.

 

No throttling, although I do have the alarm set to 90C, so when it hits that temp the TDP goes back down to 65 for a second.

 

Looks good, all normal.

Not sure what CPU settings you were using here. Having both CPU and GPU throttle back simultaneously suggests a power deficit, with the available power being allocated between the two on some ratable basis. With thermal throttling you would expect one or the other to drop back, with the other largely unaffected except as a response to the reduced overall throughput. A power deficit is logical at 26x or thereabouts, particularly in a benchmark context, but would not be expected with near stock settings.

Sounds fine. You might even try something like 27-26-18-18 with a 70W/60A power limit given that many games are still dual threaded. A single core at 27x or two cores/threads operating at 26x require much less power than 24x across 4 cores/8 threads, and the TDP acts as a power governor anyway that should keep things in hand. Don't want to scare you away from enjoying the full capabilities of the 920xm, but you have to be aware of the balancing act sometimes involved. It is also possible that your PSU is not providing full power. These 150W power units are capable of supporting reasonably heavy overclocking, but hardcore guys like Inap use the M17x-R2's 240W PSU to eliminate the power issue altogether. I don't know how to check the power output of your PSU, maybe some of our electrical wizards can help.

 

@ Rev i thought the scores looked ok too,nice to get it confirmed and thank you for that explanation on why its throttling and I've tried the multi you suggested on a few games and i cant see a difference from stock or OC'd, I'm putting this down to the CPU not being so important in most games.
I'd like to try a bench mark so there are no variables in my tests can you please suggest one to use something that will stress the CPU and GPU and give a indication to real world performance i.e gaming.I don't mind if i have to pay for it.
Thank you again for helping me.

 

You guys should adjust your heat sink screws while wPrime is running so you can find the optimal screw settings for max heat disipitation. I did so and here is what I was able to came up with

i7 940XM Using SetFSB and Throttle Stop.

wPrime 32M = 6.457 sec


wPrime 1024M = 226.001 sec

 

These are some nice scores... what were your TDP/TDC values and turbo ratio limits for these runs if I may ask?

And what were the temps during the 1024M test? No thermal throttling at all?

 

Thanks, the 32M is definately not everyday use, but the 1024 was zero thermal throttling. The TDP was probably around 100W. The key to getting no thermal thottling is to properly adjust the screws on the heat sink. On the M17x R2 when the heat sink has all screws completely screwed in, it lightly warps the heat sink which makes it a night and day differance. It causes the heat sink to not sit perfectly flush (not optimizing heat transfer). If you run wPrime at stock settings or a higher TDP, you can manualy adjust the screws as 1024M runs. Then you can adjust each screw as the benchmark runs and you will find that when properly adjusted, it should prevent thermal throttling. Right now I run a 940XM or 920XM (depends on what I leave in the system) and both are set for about 90W TDP for everyday use. Works well for me and temps never go above 85* Even with wPrime1024 running, at 100% load, I can run 100W and not get any thermal throttling, unless I've increased voltage. That will give you thermal throttling because the voltage is a huge heat multiplier.

 

That's an interesting hypothesis. If I have a day to kill I may test it out.

 

Its what I always do before closing my laptop. It makes a night and day diff. Trust me, you'll be glad you did. May have to keep the keyboard plugged in to turn the laptop on. Run it as a naked laptop. Hope to hear your results. If its not soon pm me cuz id like to know how it worked out in the M15x.

 

It'll be awhile, probably during exam week next month. I'll have a week of just chilling in my office.

 

@DR650SE: Thanks for the all the info, I guess I need to try this when I find some time, hopefully next week, but I'm pretty busy atm.

 

From 5800-6500 stock 920xm

To 8200. Pics included.