This post is more than 5 years old
1 Rookie
•
5 Posts
0
371085
August 29th, 2013 12:00
Alienware 18 CPU Throttling
I've had my AW 18 for a month now and it became clear shortly after I started using it that performance was not where it should be. Making a long story short, the entire throttling issue is neatly chronicled in this forum post.
Bottom line, all the throttling and heat/fan issues occurring can be resolved with an updated BIOS so please Dell and Alienware, help us out.
No Events found!



shadowhand
4 Posts
0
December 11th, 2013 14:00
In my case, the system goes from 45-50 degrees C while idling to 85 degrees C under load (running 3dMark, hottest during the last test, of course.) with Processor Current Limit set at 55.000A
And the same temperature range give or take a degree C while running XTU's built in CPU benchmark.
Changing Processor Current Limit to 95.000A gives me temperature ranges between 48-50 degrees C while idling and up to 90 degrees while running the XTU benchmark.
The fans seem to throttle up enough to lower the temperature to about 70 degrees whenever it has spiked during the hardest part of the benchmark though.
The average temperature during the XTU benchmark is 67 degrees at 55.000A and 69 degrees at 95.000A.
Hope those numbers are of some use to you :)
gqman69
25 Posts
0
December 11th, 2013 19:00
Chris,
Check out this Review from NOTEBOOKCHECK
http://www.notebookcheck.net/Review-Alienware-18-Notebook.102566.0.html
Here's the interesting part:
Turbo Boost
Unfortunately, our review unit cannot use its full Turbo Boost potential. While the CPU worked with 3.1-4.0 GHz in single-core benchmarks, it only ran with 3.2 GHz in multi-core benchmarks. This result is sobering and not even close to the alleged potential of the Core i7-4930MX. It could reach up to 3.7 GHz in theory; 3.2 GHz are comparable to a Core i7-4700MQ. If you spend that much money for a processor, you can expect the corresponding performance or the CPU upgrade makes no sense. We will have to wait and see if Alienware fixes this issue; some benchmark results show a negative impact of the limited Turbo Boost.
CPU Performance
The Core i7-4930MX gets a good result in the Single CPU Test of Cinebench R11.5. 1.85 points for the top version are around 9% faster than the Core i7-4900MQ (1.70 points @ Schenker XMG P703) and 13% faster than the Core i7-4800MQ (1.64 points @ One K73-3N). The entry-level Core i7-4700MQ even falls behind by almost 20% (1.51 points @ DevilTech Fragbook DTX). The situation changes in the Multi CPU Test, which is a bigger challenge for the notebook. Because of the comparatively low clock the Core i7-4930MX only manages 6.97 points. Despite its lower nominal clocks, the Core i7-4700MQ is actually on the same performance level (6.91 points @ Toshiba Qosmio X70). We should see around 7.5-8.0 points from the 57-Watt CPU under ideal conditions. We recommend you get a Core i7-4800MQ at most since the expensive models do not offer a good price performance ratio anyway.
Alienware 18
Schenker XMG P703
Schenker XMG P723
Toshiba Qosmio X70
Alienware 18
Schenker XMG P703
Schenker XMG P723
Toshiba Qosmio X70
drekkerd
1 Rookie
•
5 Posts
0
December 12th, 2013 09:00
Dell-Chris M, recreating the problem in my case is very straightforward:
1 - buy an Alienware Haswell-based system, in my case the AW 18 with 4900MQ
2 - do anything with it that puts a load on the CPU (gaming, video encoding)
3 - use whatever tool you want to see the CPU multipliers throttle way back, usually to around 3.1 Ghz but sometimes as low as 2.8 (actually UNDERclocking!)
Using Intel XTU and/or Throttlestop, Windows users have a functional workaround to at least get the Intel-rated turbo speeds out of our processors but forget about tweaking beyond that since the fan tables are messed up and the CPU quickly goes into thermal shutdown much past the rated turbo speed. What we need is all these options in the actual BIOS so the system performs as expected out of the box.
gqman69
25 Posts
0
December 12th, 2013 14:00
Just got my Alienware 18 4900MQ and I am getting by the minute. On load it never goes over 3.1Ghz. I will return it 10 days if there is no bios update.
murpdogg
4 Posts
0
December 12th, 2013 14:00
gqman69
25 Posts
0
December 12th, 2013 15:00
Well I tried everything, can't do nothing for Linux.
Sure XTU works ok when in windows but I use Linux so bye bye Alienware.
Nice machine.
DELL-Chris M
Community Manager
•
56.9K Posts
1
December 12th, 2013 16:00
I have escalated all of your comments as well as the thread on NRF. Our engineers are discussing what we can do. This will take time.
drekkerd
1 Rookie
•
5 Posts
1
December 12th, 2013 16:00
Thanks Dell-Chris M, the vast majority really love their machines and just want a proper solution to this throttling issue. I'll be happy to provide more details if needed.
gqman69
25 Posts
0
December 12th, 2013 19:00
Ok I did some tests and this might be interesting.
If I switch to the onboard Intel graphics then everything looks good!
The "Power Current Limit" seems to be 95A which is good.
If I switch back to the nvidia setup then the "Power Current Limit" becomes 55A and the CPU never goes over 3.1Ghz when ALL cores are 100%.
DELL-Chris M
Community Manager
•
56.9K Posts
0
December 13th, 2013 02:00
Thanks all! I have gathered your recent notes and sent them to the team.
gqman69
25 Posts
0
December 13th, 2013 08:00
Ok, so people who uses Linux like me.
I found a temporary solution until BIOS is fixed.
The Power Current Limit register is 0x601.
Solution
Use msr-tools and kernel module msr.
On the 4900MQ:
rdmsr 0x601
0x101414000001b8
Only the first (16-bits) is interesting.
So, 0x01b8. Now The first 3-bits are for something else (I won't go in details here).
So, 0x01b8 shifted 3-bits = 0x37 = 55 (55A)
To set your own current (i.e. 95A):
wrmsr 0x601 0x101414000002F8
This will set the power current limit to 95A.
drekkerd
1 Rookie
•
5 Posts
0
December 13th, 2013 09:00
Epic Linux workaround, bravo!
gqman69
25 Posts
0
December 14th, 2013 02:00
I use my alienware for work and my work requires lots of processing.
This machine is incredible for what it is (with the fix of course).
However, the fans algorithm need to be tweaked. Temperatures are erratic when pushed, sometimes they kick in too late.
Temps are usually ok, but they sometime peak mid-90s.
BJFox
2 Intern
•
763 Posts
1
December 14th, 2013 12:00
I realize you are using this technique as an alternative since XTU and ThrottleStop do not work for Linux. However, I am very curious about one thing. Are the changes written to the MSR registers permanently, or do they have to be set again at every reboot or cold boot? If the changes are committed and written to NVRAM--effectively replacing what is currently stored in NVRAM--then perhaps using those tools with a Linux Live USB or DVD bootable distro would get the job done for Windows users as well.
If Linux hacks (or UEFI shell hacks) can eventually lead to a fix that we can use to break, disable, circumvent or remove Secure Boot and Secure Flash and firmware signature enforcement, that would be even more awesome. Then we could flash our own unlocked BIOS like we have in the past. If we had that luxury right now we would still need Alienware to assist us with fixing the fan tables, but not having the correct default CPU power settings would be largely irrelevant.
murpdogg
4 Posts
0
December 24th, 2013 08:00
Dell,
Any ideas on when a new bios will be released for the Alienware 18?
I understand it is the holidays, but my hopes of getting a new bios before Christmas is probably not going to happen.
I have purchased numerous laptops / desktops of Alienware and Dell, but this will be my last if Dell decides to ignore this issue.