Summary

One source of Dell E6400/E6500 laptop performance problems is the output wattage of the power brick. This page contains details and a repeatable experiment to confirm the results. At time of writing, I cannot find references to this cause in the blogs, or on Dell's website. Further, when I received the system, there was no obvious documentation advising of a performance reduction by using a low powered brick (e.g., the travel charger).

UPDATE - users are reporting the same problem with the Dell XPS and Studio XPS series, and the same solution.

Performance Problem Symptoms

This problem may be indicated by the following performance problems:

System Identification

My system is a Dell E6500:

Experiment and Results

Obtain two power bricks, “LOW” and “HIGH”. The one called “LOW” must be rated at 3.34A output -- the “travel charger” that comes with the unit, model PA-12, has this output. The one called “HIGH” must be rated at 4A or higher, for instance, the PA-E3, 4.62A output, though any “large” brick from previous Dell generations will work.

Obtain a Fedora or Ubuntu live CD distribution that contains the app “glxgears.” I'm using Fedora 12 (64 bit) Installable Live CD, and these instructions are specific to that. Here is a link to 64 bit Fedora 12 iso. Other graphical-boot Linux live distributions might have this. (If you have Fedora or Ubuntu installed, you don't need the live boot CD.)

  1. Connect the system to the HIGH power brick.
  2. Boot Linux into graphical mode.
  3. With HIGH plugged in, start the application “glxgears” in a terminal window: go to Applications, System Tools, Terminal, then enter glxgears. Let it run for about 30 seconds, and observe the sequence of performance numbers.
  4. With glxgears running, unplug HIGH and plug in LOW.
  5. Let glxgears run for about 30 more seconds.
    ** you should observe a lower frames per second rate. **
  6. Unplug LOW, and plug in HIGH. Let run another 30 seconds.
    ** you should observe a higher frames per second rate. **
  7. ^C to kill glxgears, then enter “openssl speed dsa2048” in the terminal window. Try one run with LOW, wait for it to quit, then try one run with HIGH. Observe a performance difference.

I observed a difference of 1172 FPS on HIGH, versus 439 FPS on LOW for glxgears. I observed 967 sigs/sec on HIGH, versus 194.1 on LOW for dsa2048. HIGH is 267% faster than LOW on glxgears, and 498% faster than LOW on dsa2048.

See the Data Section for detailed results.

Conclusions

If you are having steady, continuous performance problems with your E6400 or E6500, are not seeing any BIOS errors complaining about “non Dell adaptor,” and have tried upgrading video drivers without success, try changing power bricks.


This section contains additional information.

Background

I obtained a Dell E6500 in April 2009 with a docking station. After noticing significant video performance problems and finding links to sites suggesting that the Intel GM45 video card is to blame, I upgraded through Ubuntu 8.10, 9.04, 9.10, before switching to Fedora 11 and later 12. Many users have cited the Intel GM45 video card as the problem, which is why the upgrades, but in my case, it's not. Links on Intel performance: Intel Graphics Performance Guide, Troubleshooting Intel Graphics, one bug report, second bug report, downgrading X video drivers.

Other users have cited “throttlegate,” indicating that the performance goes down related to a heat issue. This may be an independent problem that I have not yet observed, reported on Dell Tech Support Website, and covered by a wonderful detailed thermal analysis (pdf), allegedly suppressed by Dell. Other reports of the thermal problem, that I have not noticed but still might affect me, include Techspot, Engadget, Ubergizmo, Tweaktown, Technews, ZDNet, Hardware IT News, and someone's personal blog.

Problem Discovery

I observed that performance depended on whether I was docked, or undocked and working at home. At work, I tried undocking (hastily, without AC adaptor), and found that performance was stellar. Finding that odd, I tried connecting each peripheral that I take with me, from my backpack, one by one into the unit – mouse (no effect), USB stick (nothing), then power cord – and saw that the travel charger power cord that I take with me caused the performance to drop. I then tried the power cord from the dock, and performance was superb. I borrowed a bunch of different Dell power bricks, various vintages, of different amperages / wattages (some the same power but different vintage), and obtained the experimental results.

It is frustrating that these performance problems went on for literally months when all I had to do was change power bricks, not Linux distributions.

Data

3.34 Amp Raw Data

$ openssl speed dsa2048 ; glxgears & sleep 26 ; kill $!
Doing 2048 bit sign dsa's for 10s: 1940 2048 bit DSA signs in 9.99s
Doing 2048 bit verify dsa's for 10s: 1647 2048 bit DSA verify in 9.99s
OpenSSL 1.0.0-fips-beta4 10 Nov 2009
built on: Wed Nov 18 09:06:54 EST 2009
options:bn(64,64) md2(int) rc4(1x,char) des(idx,cisc,16,int) aes(partial) blowfish(idx)
compiler: gcc -fPIC -DOPENSSL_PIC -DZLIB -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -DKRB5_MIT -m64 -DL_ENDIAN -DTERMIO -Wall -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector --param=ssp-buffer-size=4 -m64 -mtune=generic -Wa,--noexecstack -DMD32_REG_T=int -DOPENSSL_IA32_SSE2 -DOPENSSL_BN_ASM_MONT -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DMD5_ASM -DAES_ASM -DWHIRLPOOL_ASM
sign verify sign/s verify/s
dsa 2048 bits 0.005149s 0.006066s 194.2 164.9
2202 frames in 5.0 seconds = 440.289 FPS
2127 frames in 5.0 seconds = 425.384 FPS
2309 frames in 5.0 seconds = 461.643 FPS
2054 frames in 5.0 seconds = 410.644 FPS
2287 frames in 5.0 seconds = 457.164 FPS

4.62 Amp, 6.7 Amp Raw Data

The difference between the 4.62A versus 6.7A runs was not statistically significant, but both runs are statistically significant compared to the 3.34A run.

$ openssl speed dsa2048 ; glxgears & sleep 26 ; kill $!
Doing 2048 bit sign dsa's for 10s: 9677 2048 bit DSA signs in 9.99s
Doing 2048 bit verify dsa's for 10s: 8377 2048 bit DSA verify in 10.00s
OpenSSL 1.0.0-fips-beta4 10 Nov 2009
built on: Wed Nov 18 09:06:54 EST 2009
options:bn(64,64) md2(int) rc4(1x,char) des(idx,cisc,16,int) aes(partial) blowfish(idx)
compiler: gcc -fPIC -DOPENSSL_PIC -DZLIB -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -DKRB5_MIT -m64 -DL_ENDIAN -DTERMIO -Wall -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector --param=ssp-buffer-size=4 -m64 -mtune=generic -Wa,--noexecstack -DMD32_REG_T=int -DOPENSSL_IA32_SSE2 -DOPENSSL_BN_ASM_MONT -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DMD5_ASM -DAES_ASM -DWHIRLPOOL_ASM
sign verify sign/s verify/s
dsa 2048 bits 0.001032s 0.001194s 968.7 837.7
5907 frames in 5.0 seconds = 1181.264 FPS
5843 frames in 5.0 seconds = 1167.671 FPS
5805 frames in 5.0 seconds = 1160.925 FPS
5819 frames in 5.0 seconds = 1163.717 FPS
5817 frames in 5.0 seconds = 1163.341 FPS

About

Russ Fink is a Ph.D. Student at UMBC, and an open source aficionado. A co-worker witnessed the experiment.