Any updates on the XPS 16 heating/ throttling issue?

I am pretty sure putting the brightness on minimum while gaming reduces the likelihood of CPU throttling, but it's a lot harder to prove. I'll see what I can do.

Prime+Furmark can be used to uncover a lot of throttling facts even though Dell won't recognize these programs as "real-world" usage of the laptop. The fact that changing the brightness changes the multiplier within 2 seconds is very insightful as to what's going on. Watching Prime+Furmark is like watching someone commit a murder - you know who's guilty of what, but it's not evidence.

The next time you or someone else is playing a game and experiencing CPU throttling, try lowering the brightness to the minimum, and see if you experience a speedy recovery. This worked for me the one time I tried it.

 

If the problem is indeed a power issue, how can there be two laptops with similar configurations, but one throttles and the other doesn't? I know there are many variables in this equation, so perhaps it is just the smallest thing or something that we haven't thought of yet that is making the difference. But as I've stated a number of times, I have yet to experience throttling while gaming for hours and hours, while others experience it after only minutes of gaming. It's not like I'm going easy on my system or anything. Usually full display brightness, using the internal laptop speakers, on wifi, external mouse plugged in... etc etc. It seems to me that if some laptops throttle and others don't (or at least not appreciably during gaming), then this should be a solvable problem. I would be fun to get a bunch of us together in one room with our laptops and start to test this out. I haven't had a good lan party/geek-out in a LONG time.

 

That would be awesome, but I really can't see this happening.

If only Dell had test engineers and quality control that cared and/or were honest.

 

A public SXPS meeting would be kinda cool to compare bencmarks and show/prove that throttling is a case by case basis and not an overall SXPS issue.

As for taxing your system, nobody is claiming you arent hard on your system. But also realize that even with hard gaming and multitasking put together does not come close to what prime95 + furmark do to a system. I work my system pretty hard, but I do not come close to full tilt on my GPU and CPU constantly.

 

Ok finally had a chance to run the new AC adapter (from 1645 w/5730) vs. old AC adapter (from 1645 w/4670) on my old 1645 machine to see if it would make a difference and also ran some test with cooling...

5 TS logs - 1645, w/4670, 9-cell, 130w AC, A09 BIOS, WLED, 320GB 5.4k HD (Ambient air temp ~75F)
1) ThrottleStopLog-old1645-oldAC.txt
- old AC adapter computer open full
2) ThrottleStopLog-old1645-newAC.txt
- new AC adapter computer open full
3) ThrottleStopLog-old1645-newAC-lid closed.txt
- new AC adapter screen almost closed all the way
4) ThrottleStopLog-old1645-oldAC-max_cool.txt
- old AC adapter screen almost closed all the way, ghetto cooler ON
-->Best Result 5) ThrottleStopLog-old1645-newAC-max_cool.txt
- new AC adapter screen almost closed all the way, ghetto cooler ON

CLIFFNOTES:
Short of it the AC adapter isn't a cure all, seems to be a small improvement with the new AC adapter...BUT seemingly the COOLER you can get the laptop the better it runs.

Oh and still NO THROTTLING on my new 1645 w/5730...knock on wood...
(Prime95+FurMark and LFD(1) run fine)

Unclewebb Throttlestop question
...how to you get it to log GPU temp?
I have it set GPU=2(and TJMax=100), but for some reason it dosn't want to log it any more...it used to work fine but something happened?

 

I agree 100%. The 130 watt ac adapter alone should not cure throttling in an RGBLED 4670 XPS. The wattage is still well above the 130watt maximum output. In my humble opinion the only way to make such an XPS not throttle would be to use a 160watt ac adapter. The full cure, imho again, to the power throttling issue is a combination of things: WLED, 5730 GPU, 130watt ac adapter, and a09 bios. All those things combined lower the power useage and heat output by alot.

 

Ya, it will very likely never happen, but it's a dream that I can dream happily. Not since highschool have I been to any epically awesome lanparties, though a few of us would get together occasionally during University.

And yes, if only the test engineers had a passion for their work... I'd love to have a job like that...

 

m715,

So I thought mine was perfect too, until I paid attention to the GPU.... you are positive it isn't being affected?

What was your max temp for the GPU??

 

All right, I have an exam tomorrow, so not a long post:

I know of at least one 5730/WLED user who throttles, so it ain't over yet.

P95+FurMark = I reckon most laptops would have issue running this test. That's like shooting a nuke at Kevlar to prove that it isn't strong enough. Throttling/weakness of Kevlar are never going to disappear. All any computer can do is minimize throttling so it does not appear during regular usage. That's it. But if you push hard enough, ANY computer will throttle. P95+FurMark is not anywhere near any kind of realistic test. Period.

I can accept the idea the RGB is too power-hungry. No wonder they took it off.

So, what is the solution here besides a totally redesigned laptop?

 

Is this the right thread to post TS logs or some other thread.

Thanks

 

GPU max temp without cooling has been 78C, that's both while gaming and running FurMark + Prime95 together for over 1 hour each...

Not saying its cured but the new system seems to work.

 

Yeah, same here.... that is the new throttle point... are you SURE your GPU didn't thottle??????

Only reason is like I said I had the same feeling that it was fixed.. brotha it aint!!!!

 

Dell used to build and ship laptops that didn't overheat or throttle. There used to be enough watts available so a user didn't have to constantly worry what application was going to send them over the edge. Here's a Prime95+Furmark test of a D830 with an Nvidia GPU. No throttling and the temperatures were great.

http://forum.notebookreview.com/6382811-post552.html

The D830 shipped with a 90 watt adapter. I have no idea whatever possessed Dell to try shipping the same sized adapter with the original XPS 1645. The 130 watt adapter is an improvement but it's still marginal. If you're lucky and it doesn't throttle, there's still next to nothing left over to charge the battery while gaming.

m715: I'm not sure why the GPU=2 trick isn't working for you. ThrottleStop asks the ATI driver for temperature info so maybe the driver you are using no longer supports this feature. All I have to test with is an ATI 5770 desktop GPU and GPU temperature reporting still works on that with the 10.6 desktop driver.

 

Hmm...I'll watch out, next time i'm gaming I'll run GPU-Z and watch the temps and seeds as well...

Thank you...It was a ATI driver issue, reinstalled 10.6 and GPU temps are back Thanks!

 

Just tried playing Gears of War for about 6 mins. Max my temps got up to were 82 (CPU) and 82.5 (GPU), then throttling kicked in. I can't believe how how this thing gets, I cant even touch the bottom of the unit ANYWHERE without burning myself and the palm-rest is just flat out uncomfortable for gaming. Calling Dell right now to see what can be done.

 

What do you think of this idea for linking power consumption with (cpu) throttling:

Run a game that throttles, on minimum brightness (I know it's a pain, I just did it). Observe the multiplier when throttling as compared to on maximum brightness.

For me, I get 7x or 9x multipliers when throttling in game on max brightness, but 11x when on minimum. I can back this up with logs (although the logs don't say what brightness I'm using)

Anyone on A09 who experiences CPU throttling in games- what do you think of this idea for IdeaStorm?

 

gpig, I get throttling even when using an external monitor with the onboard laptop screen SHUT OFF...so I just really don't know what to think, this cpu must need a hell of a lot of juice when shutting the monitor off still doesnt give it enough power.

 

Is this with the A09 BIOS (and 5730)? If so, then your laptop is indeed in horrible shape.

 

That was with A09...I have since upgraded to A10 but can't tell the difference. Still get throttling after the same amount of time in every game I've tried to play. How hot this damn thing gets after having it on for a while reguardless of what I'm doing really bothers me too, even though according to HWmonitor my temps are within acceptable range with a 6-cell battery. Throttling CAN'T be temperature related because my temps stayed below 75 C when gaming on an external monitor with the vent not being blocked. I'll try again tomorrow to talk to someone at D3ll, todays 1/2 hour was a waste of time.

 

Hey I got my new laptop, here are REALLY quick logs.
On Adapter
On battery

Looks like no throttling on adapter. Don't know about when I'm on battery. I set powerplay to performance and min. processor state to 100% on battery

*I didn't update anything yet, its on A09 and video driver version 8692.1.0*

 

Hmm..."onboard laptop screen SHUT OFF"...that might just give me enough power

I'm running a test on my old 1645 w/4670 ( NOT PRACTICAL)
- screen off
- "custom cooling"
- A10 BIOS
- AC adapter from my 1645 w/5730

So far NO THROTTLING of CPU or GPU...will post results when the hour test is up...
(w/FurMark + Prime95)

I feel your pain...I just finished going through all kinds of problems with my 1645 w/4670 and finally for me the 1645 w/5730 seems to be good so far...

 

I am weary about requesting a replacement with the same specs...thinking of going with an i5 instead, for everything I do (and most people I would imagine), there shouldn't be a difference between any of those and the i7 IMO.

 

Agreed a higher spec i5 with the 5730 would be great, and it likely wouldn't throttle!

 

Ok, 1+ hour stress test of FurMark + Prime95 on a 1645 w/4670 and NO THROTTLING

Practical? NO
DO NOT try this at home...

- 130w AC from my 1645 w/5730
- remove back plate from laptop
- 1-20" box fan
- "wetcardboard" idea screen OFF (used external)
- laptop screen almost closed
- A10 BIOS
- 78F Ambient temperature

CPU max temps 65/61/65/64C
GPU max temp 69C

With the 20" box fan the temp concern was taken off the table, this clearly highlights the power issue

No screen on no throttling
Screen on max brightness instant throttling
Screen set to 1/4 brightness no throttling within 5 minutes

 

What was GPU core and mem reading during all of this?

 

675 / 796.5 (GPU-Z) - A10 BIOS

If the GPU was allowed to reach 84C then it would throttle the GPU to 300 / 500 (in the GPU-Z log you can watch it bounce off this 84C limit), and interestingly this would allow the CPU to go back to 12.00

Also throttling is somewhat temperature based, either 77/78C? on the CPU or on the GPU? (not sure which one) will start mild throttling (see 23:00:03 in the log), where on the test above the temps were kept under this and no throttling occured.

Do higher temperatures require higher wattage draws?

Any tests for a 1645 w/4670 people want run before I send it back?

 

I would worry about EMI sitting it that close to a house fan like that.

 

How did you get it to run at those numbers? 675 is higher than it is even rated... its rated 650...

 

Thanks, I agree...that's why you see the 1645 that is being sent back...and not my new 1645 I'll put a do not try this at home disclaimer in...

I think the 5730 is 650 and the 4670 is 675
(at least that's what the two 1645's I have sitting here run)
*edit* Lists the core speed @ 675MHz for the 4670
http://www.notebookcheck.net/AMD-ATI-Mobility-Radeon-HD-4670.13881.0.html

 

Oh I see, that was for the 4670.. that makes sense now....

 

Did another quick test after I updated to A10 and 10.6.
TS LOG HERE

I unplugged at about 24:00 (12:00) and my multipliers instantly dropped. Anyone know what I can do to prevent this? I already set battery to performance and minimum processor state on battery to 100%.

 

How about plug the thing in?

Stress testing with Furmark and / or Prime95 takes a lot of watts and that much load can damage a battery so a little bit of throttling in this case isn't such a bad idea.

 

Oh! Thanks for your knowledge! X) Now I feel dumb

 

See what I mean? If it throttles in FurMark+P95, that isn't that huge of a deal. The D830 is just a single example, anyways, though the exception is noted.

Thanks for the test m715. I'll forward that to Dell guys. Someone quoted 116W was the maximum power draw for the SXPS?

Preferably, I need someone with a meaty adapter and TS to see what wattage adapter is needed.

The temperature throttling will be harder to cure, I'm sure, but power just needs a different adapter.

~Ibrahim~

P.S. m715, do you have the RGB?

P.P.S. Shiz, I just realized the tests you have are P95 + FurMark, which Dell won't accept. I guess you've dismantled your rig already? If not, can you test with some games, if you have any?

 

I believe that was liquid's average draw under max power load, but his has the 5730/WLED config. The 4670 and RGBLED probably add 10 watts, each.

 

Everything is still sitting on my desk so...

I have just the WLED 1080p, only game I have installed is Left 4 Dead (1)
Same conditions (on the box fan w/max cooling, screen off and closed), all graphic settings maxed it will run at a 13.00 multiplier, does this mean it is throttling midly?
I guess the usb mouse and keyboard draw just enough power to limit it to 13.00? Temps are not a problem (max cpu 54C gpu 59C - ambient temp 72F)

I even tried disabling anything I could think of in hopes it would cut the power to it (CD, WiFi, BT, IR, IEEE, webcam, sound 10/100...) and pulled one of the 2 RAM chips...but same result locks in at 13.00

Oh I a different note, 1st problem with my replacement 1645 w/5730
- When trying to wake from hibernation today (I seem to remember other people talking about hibernation problems)...it won't period, tried ac+batt, batt only, ac only (would give the tinest flash of orange on the hinge and that was it), finally pulled batt held power button 30 sec, plugged back it and it finally started up...BUT all my touch media buttons are dark and dead now. Hopefully this is only a software issue...time to go update all the drivers.

 

The default multiplier for a Core i7-720QM is 12. Anything above that means you are getting some turbo boost. The amount of turbo boost you get depends on how many cores are active and if you are within the Intel power requirements for turbo to be working. Anything below 12 is a sign of throttling but anything above 12 is fine. Your log file shows a lot of 13.00 numbers which is equivalent to one bin of turbo boost.

The multiplier locked at 13.00 looks suspicious but it's hard to say if it is a limitation of your CPU or if it is throttling related.

 

13x seems disappointing because 18x (2.4 GHz) is supposed to be "dual core" turbo mode. If the game is L4D, try disabling Multicore Rendering in-game, and see if it does anything.

 

posted elsewhere

 

Yep I think your right...here is an old log with BIOS A09 & same game, multipliers range from 13.xx to 14.xx, until it overheated and throttled to 7.00

Game is L4D1, Multicore disabled yield higher multipliers
Maybe i'll roll back to A09 and see what it does.

Tommorow UPS will be coming for the old 1645...

 

ran a game (need for speed shift all max settings)....on A09 (I've been switching between 7,8,9, and 10 to see what works best)

Code:

06/30/10 16:06:08  13.94   20.1  100.0  100.0   31
06/30/10 16:06:09  14.09   20.7  100.0  100.0   31
06/30/10 16:06:10  12.73   22.2  100.0  100.0   30
06/30/10 16:06:11  13.21   22.4  100.0  100.0   30
06/30/10 16:06:12  14.00   19.8  100.0  100.0   31
06/30/10 16:06:13  13.31   23.2  100.0  100.0   31
06/30/10 16:06:14  12.45   20.9  100.0  100.0   30
06/30/10 16:06:15  13.41   20.3  100.0  100.0   33
06/30/10 16:06:16  14.19   17.5  100.0  100.0   32
06/30/10 16:06:17  14.44   17.7  100.0  100.0   32
06/30/10 16:06:18  13.75   19.3  100.0  100.0   32
06/30/10 16:06:19  14.49   15.6  100.0  100.0   32
06/30/10 16:06:20  14.41   16.2  100.0  100.0   32
06/30/10 16:06:21  14.20   18.1  100.0  100.0   32
06/30/10 16:06:22  14.09   17.6  100.0  100.0   32
06/30/10 16:06:23  14.69   16.3  100.0  100.0   32
06/30/10 16:06:24  13.84   18.7  100.0  100.0   33
06/30/10 16:06:25  13.81   20.2  100.0  100.0   33
06/30/10 16:06:26  14.42   21.2  100.0  100.0   31
06/30/10 16:06:27  13.37   21.6  100.0  100.0   31
06/30/10 16:06:28  13.62   20.8  100.0  100.0   31
06/30/10 16:06:29  13.84   21.7  100.0  100.0   31
06/30/10 16:06:30  15.17   20.5  100.0  100.0   31
06/30/10 16:06:31  14.31   19.9  100.0  100.0   31
06/30/10 16:06:32  13.51   20.7  100.0  100.0   32
06/30/10 16:06:33  13.21   20.6  100.0  100.0   32
06/30/10 16:06:34  14.56   20.2  100.0  100.0   31
06/30/10 16:06:35  14.81   22.9  100.0  100.0   31
06/30/10 16:06:36  15.16   21.7  100.0  100.0   31
06/30/10 16:06:37  14.46   20.5  100.0  100.0   32
06/30/10 16:06:38  14.12   17.8  100.0  100.0   32
06/30/10 16:06:39  13.47   20.5  100.0  100.0   32
06/30/10 16:06:40  14.59   20.1  100.0  100.0   31
06/30/10 16:06:41  14.03   20.5  100.0  100.0   31
06/30/10 16:06:42  14.46   24.1  100.0  100.0   30
06/30/10 16:06:43  14.76   23.9  100.0  100.0   30
06/30/10 16:06:44  14.73   22.9  100.0  100.0   30
06/30/10 16:06:45  15.12   21.8  100.0  100.0   30
06/30/10 16:06:46  13.56   21.9  100.0  100.0   31
06/30/10 16:06:47  14.44   20.8  100.0  100.0   32
06/30/10 16:06:48  12.90   21.6  100.0  100.0   32
06/30/10 16:06:49  12.06   21.7  100.0  100.0   32
06/30/10 16:06:50  12.40   22.0  100.0  100.0   31
06/30/10 16:06:51  11.74   22.7  100.0  100.0   32
06/30/10 16:06:52  12.50   20.5  100.0  100.0   32
06/30/10 16:06:53  12.51   20.9  100.0  100.0   32
06/30/10 16:06:54  12.50   21.9  100.0  100.0   32
06/30/10 16:06:55  12.26   22.2  100.0  100.0   31
06/30/10 16:06:56  13.30   25.1  100.0  100.0   29
06/30/10 16:06:57  13.42   32.3  100.0  100.0   29
06/30/10 16:06:58  13.03   23.9  100.0  100.0   32
06/30/10 16:06:59  13.80   22.5  100.0  100.0   30
06/30/10 16:07:00  14.60   20.7  100.0  100.0   31
06/30/10 16:07:01  14.79   24.2  100.0  100.0   30
06/30/10 16:07:02  14.49   26.3  100.0  100.0   31
06/30/10 16:07:03  14.13   23.3  100.0  100.0   30
06/30/10 16:07:04  15.01   21.5  100.0  100.0   30
06/30/10 16:07:05  15.04   23.0  100.0  100.0   29
06/30/10 16:07:06  14.64   22.5  100.0  100.0   31
06/30/10 16:07:07  14.73   23.0  100.0  100.0   29
06/30/10 16:07:08  14.87   23.8  100.0  100.0   30
06/30/10 16:07:09  14.60   24.3  100.0  100.0   30
06/30/10 16:07:10  14.05   23.6  100.0  100.0   29
06/30/10 16:07:11  14.87   22.4  100.0  100.0   30
06/30/10 16:07:12  14.16   21.4  100.0  100.0   30
06/30/10 16:07:13  14.13   23.0  100.0  100.0   29
06/30/10 16:07:14  14.96   21.5  100.0  100.0   31
06/30/10 16:07:15  14.75   21.6  100.0  100.0   30
06/30/10 16:07:16  14.72   22.2  100.0  100.0   29
06/30/10 16:07:17  15.11   20.5  100.0  100.0   29
06/30/10 16:07:18  13.90   21.3  100.0  100.0   30
06/30/10 16:07:19  14.48   21.7  100.0  100.0   30
06/30/10 16:07:20  14.68   22.0  100.0  100.0   29
06/30/10 16:07:21  14.62   19.5  100.0  100.0   30
06/30/10 16:07:22  13.60   22.3  100.0  100.0   29
06/30/10 16:07:23  15.07   21.1  100.0  100.0   29
06/30/10 16:07:24  14.16   21.8  100.0  100.0   30
06/30/10 16:07:25  14.85   20.8  100.0  100.0   30
06/30/10 16:07:26  14.84   21.5  100.0  100.0   30
06/30/10 16:07:27  14.75   21.5  100.0  100.0   30
06/30/10 16:07:28  14.92   21.6  100.0  100.0   31
06/30/10 16:07:29  14.49   21.8  100.0  100.0   29
06/30/10 16:07:30  14.96   23.7  100.0  100.0   29
06/30/10 16:07:31  14.81   24.6  100.0  100.0   29
06/30/10 16:07:32  14.73   24.0  100.0  100.0   29
06/30/10 16:07:33  14.70   22.4  100.0  100.0   30
06/30/10 16:07:34  14.89   22.0  100.0  100.0   29
06/30/10 16:07:35  13.85   22.9  100.0  100.0   30
06/30/10 16:07:36  13.30   21.8  100.0  100.0   29
06/30/10 16:07:37  13.78   20.4  100.0  100.0   30
06/30/10 16:07:38  12.39   21.0  100.0  100.0   30
06/30/10 16:07:39  13.33   18.9  100.0  100.0   30
06/30/10 16:07:40  14.57   28.2  100.0  100.0   27
06/30/10 16:07:41  13.99   30.4  100.0  100.0   24
06/30/10 16:07:42  13.57   31.1  100.0  100.0   29
06/30/10 16:07:43  14.78   20.7  100.0  100.0   29
06/30/10 16:07:44  13.63   20.4  100.0  100.0   30
06/30/10 16:07:45  13.80   21.5  100.0  100.0   30
06/30/10 16:07:46  15.05   19.7  100.0  100.0   30
06/30/10 16:07:47  14.37   20.9  100.0  100.0   29
06/30/10 16:07:48  14.16   21.1  100.0  100.0   30
06/30/10 16:07:49  13.40   21.1  100.0  100.0   30
06/30/10 16:07:50  13.20   22.6  100.0  100.0   30
06/30/10 16:07:51  14.84   20.4  100.0  100.0   29
06/30/10 16:07:52  14.17   20.9  100.0  100.0   29
06/30/10 16:07:53  13.50   21.3  100.0  100.0   30
06/30/10 16:07:54  13.44   21.1  100.0  100.0   30
06/30/10 16:07:55  13.19   21.7  100.0  100.0   30
06/30/10 16:07:56  13.51   21.9  100.0  100.0   30
06/30/10 16:07:57  14.05   19.2  100.0  100.0   30
06/30/10 16:07:58  14.27   19.3  100.0  100.0   30
06/30/10 16:07:59  14.06   20.5  100.0  100.0   31
06/30/10 16:08:00  14.52   17.2  100.0  100.0   30

 

woops shoulda put it in a tesxt file..anyways its just a small snippit of it because it doesnt deviate from this for the entire 1 hour that ive been playing it

 

A couple of code html tags are also useful if you are pasting a chunk of a log file. Just use this [code ] without the space at the end and [/code ] without the space at the end and paste your log file in the middle. It will take up less room that way and be easier to read. That log file looks good.

 

Here's an example. You could have 4 threads fully loaded at 100% while the other 4 threads are at 0%. If each full load thread was on a different core then all 4 cores would be in the active state and the maximum multiplier would be limited to 13. These 4 threads could also be running on the first two cores, one main thread and one hyper thread for each of the first two cores could be fully loaded. The second two cores could be idle in the C3/C6 sleep state. Now the maximum multiplier could be as high as 18 because that's the maximum for when 2 cores are awake and 2 are asleep. In both cases the overall load in the task manager is going to be 50% but depending on how these threads are distributed within the CPU can make a big difference to overall performance and what multiplier you get. Some of this is controlled by the CPU and some of this can depend on how a piece of software is coded. A program can take advantage of how these CPUs operate but a well threaded program might backfire and keep more cores awake and reduce the multiplier and overall CPU speed.

On a Core i7-720, if more than 2 cores are active, the maximum multiplier is limited to 13. To fully load one core, you only need to fully load one of its two threads. So two fully loaded threads running on two cores can bring you up to the point where the maximum multiplier will start to drop to 13. That's only an overall CPU load of 25% but depending on how the load is arranged within the CPU, going beyond this can start to significantly limit the amount of turbo boost available and will lower your maximum multiplier.

It might be interesting if you could use Set Affinity... in the Task Manager to lock a game to 4 threads that correspond to 2 of your 4 cores. That might help significantly boost your multiplier if it allows the other two cores to spend more time asleep. It might let an i7-720 use the 18.00 multiplier a little more often while gaming.

For testing, run 4 threads of Prime95 - Small FFTs and then move these threads around using Set Affinity. You should get a much better average multiplier if you arrange them so only 2 cores are active compared to if you arrange these threads so all 4 cores are active.

 

I just tried this - small FFTs:

Set it to only to use CPU0 and CPU1 and the Multi is between 18-19s. When I increase that to CPU0, CPU1, CPU2 and CPU3, they hover between 15-16.

When I include CPU4 and CPU5 it's just under 13 and of course with everything it's at 12.

 

I just tried this too. Same exact results as Error.

I also ran furmark at the same time and the numbers didnt change.

Could set affinity be used to increase performance in games or other gpu intense applications?

 

FSX allows you to chose set the affinity in the cfg..there are litterally 1000000 topics about this regarding quad cores and i7's. But none have the such insight about the boost mode of the i7. Which threads would need to be set to only use 2 cores?

 

The threads are not always ordered the same on different computers. Even on the same computer, it's possible that the bios can change the ordering of the threads each time you boot up so it's a good idea to do some testing.

The Task Manager is designed to simply show you 8 threads but doesn't put any effort into displaying those threads in the correct physical order on your CPU. That information is contained in the APIC ID number for each thread. If you look in the RealTemp Settings window you should see your APIC ID number.

http://www.fileden.com/files/2008/3/3/1794507/RealTempBeta.zip

The typical number is 01234567. If you have booted up and your APIC ID is that number then the first two threads in the task manager belong to core 0, the next 2 belong to core 1, the next 2 belong to core 2 and threads 6 and 7 belong to core 3. If this is the case then use Set Affinity... to lock your game to either the first 4 or the last 4 threads.

The other common APIC ID is 02461357. If this is your APIC ID and you lock your game to the first 4 threads in the Task Manager then it will be running your game on all 4 cores which is not what you want to do if you are trying to maximize your multiplier.

For this situation, you would use task manager to lock it on the first 2 threads, then skip two threads and then check off two more threads and leave the last two unchecked. If the APIC ID sequence is nothing like these two examples then it gets a little more complicated.

It will be interesting to see some testing of this to see if managing your CPU a little better can give you a higher average multiplier and better performance. If this helps avoid or delay throttling, that would be a bonus.

 

All right, thanks m715.

So, I just want to get this right: the system does not throttle because of power concerns in any type of use.The only exception to this is P95+FurMark.

I realize m715 could only test L4D1, but he did max it out...

 

I played L4D1 for about 30mins maxed out (720p) and didn't see any throttling. Frame rate did seem like it was a solid 60 for some parts but got as low as 25 FPS in other parts. Don't know if that was just because of a wide open area or gpu throttling? Anyways I didn't get the usual throttling like I experience with EVERY OTHER GAME after 5mins. My temps while playing L4D1 were just under 80 though compared to playing Crysis, Mass Effect 2 and Bad Company 2 were my gpu hits 84.5 C and my CPU 84 C.