3d games over w510

[Adda]

I'm not really doing a lot of testing, sometimes playing with some 3D gaming with the fan set to max, and the rest of the time I use the w510 with ubuntu, however, not known high temperatures, indeed ..

if you want me to do a test with my tell me as well, for comparison, I also have the cpu i7 720

bye

Well since you have the same CPU as I do, your W510 should behave the same as mine.
So I'd keep an eye on it, you never know what game might make it run HOT.
Before I did the tweaks I talked about before, I never managed to make my W510 run completely stable.

Running Ubuntu and such there are no worries, the only time where there is a chance of overheating, is when the GPU and CPU is under heavy load at the same time.
But not every game will overheat it, some might, others not at all.

The tests are just a way of seeing if the system is capable of overheating.

[aldimeola81]

seems to me that the w510 is not very warm , however today I disassembled the fan , applied the paste and reassembled , the temperature has remained the same , then I disassembled again , and reassembled pushing hard on the heat sink down before tightening the screws and moving it to the side so as to spread the thermal paste , now operating temperature of the GPU at the same game with the same settings dropped from 61C to 58C but the important thing is that the fan stops during normal use on windows or ubuntu , before she ran constantly.

For example, here on ubuntu as I write, the temperature of the GPU is 37 C and the cpu 41C

Test yourself , I think that will improve heat dissipation and stability ' in 3D games .

the important thing is that the gpu / cpu are fully in contact with the heat sink , then you have to push very hard on the heatsink during assembly, so as to drain the excess .

so far I've done this work on two computers and the results were excellent , bye.

[Adda]

I have repasted my W510 many many times, using AS5, I'm the one who discovered that some W510 units have poorly fitted coolers.
As I have said many times, some games make the W510 run hot, others do not, the one you are running does not.
The easiest way to test if your W510 can handle any load, is to run furmark and prime95 at the same time.
When I play Skyrim, my W510 barely exceeds 60C, does that mean all games run as cool as that? no.

It seems you don't want to, that's fine, but you don't know if your W510 can handle any load unless you try and make it run as hot as your can.

http://forum.thinkpads.com/viewtopic.php?f=48&t=90854

[aldimeola81]

I have repasted my W510 many many times, using AS5, I'm the one who discovered that some W510 units have poorly fitted coolers.
As I have said many times, some games make the W510 run hot, others do not, the one you are running does not.
The easiest way to test if your W510 can handle any load, is to run furmark and prime95 at the same time.
When I play Skyrim, my W510 barely exceeds 60C, does that mean all games run as cool as that? no.

It seems you don't want to, that's fine, but you don't know if your W510 can handle any load unless you try and make it run as hot as your can.

http://forum.thinkpads.com/viewtopic.php?f=48&t=90854

excuse if we do not understand very well, it will be 'for my bad English.
I appreciate your advice,
I certainly do not want to teach anything to anyone, it was just to understand.
one thing I do not understand is why the w510 fan turns on at 38C, and does not stop before 34C seems absurd to me, turn to such a low temperature.


I will post the test results as soon as possible....

[Adda]

You don't have to test anything, if you don't find it to be important.
But if you don't, I recommend that you keep a temperature monitor going while you play games, so that if you run in to a game that runs hot, you'll know, or if a game suffers from slowdowns, the temp monitor can tell you if overheating caused it.

The default fan control in the W510 is terrible, as you say, the fan turns on at very low temps, and doesn't really turn off again once it has started.
But at high temps, it's not aggressive enough, and allows the system to run very hot indeed.
It's definitely best to use TPFanControl or IBM_ECW to fix that.

I have been doing some more testing, I have "CPU Power Management" enabled so that the CPU can go as high as 2.8GHz, and played some Borderlands 2.
With the GPU undervolted, the highest temp remains around 70C, with the GPU at default, temps go as high as 76C.
This is with the fan at max speed (64).

I decided to undervolt and overclock the GPU, and have "CPU Power Management" enabled.
I tested this config with furmark and prime95, and it just about remains below 90C, with heavy throttling, not good, but acceptable since the games I play at this time, don't make it run that hot.
It's probably just a matter of time though.

[aldimeola81]

You don't have to test anything, if you don't find it to be important.
But if you don't, I recommend that you keep a temperature monitor going while you play games, so that if you run in to a game that runs hot, you'll know, or if a game suffers from slowdowns, the temp monitor can tell you if overheating caused it.

The default fan control in the W510 is terrible, as you say, the fan turns on at very low temps, and doesn't really turn off again once it has started.
But at high temps, it's not aggressive enough, and allows the system to run very hot indeed.
It's definitely best to use TPFanControl or IBM_ECW to fix that.

I have been doing some more testing, I have "CPU Power Management" enabled so that the CPU can go as high as 2.8GHz, and played some Borderlands 2.
With the GPU undervolted, the highest temp remains around 70C, with the GPU at default, temps go as high as 76C.
This is with the fan at max speed (64).

I decided to undervolt and overclock the GPU, and have "CPU Power Management" enabled.
I tested this config with furmark and prime95, and it just about remains below 90C, with heavy throttling, not good, but acceptable since the games I play at this time, don't make it run that hot.
It's probably just a matter of time though.

I ran the test as you said, more 'times in a row, with Furmark and prime 98 set at maximum heat production:

I checked the temperatures with HWMonitor, here is the result.

in the bios the cpu management is on automatic mode for ac and windows power manager is set to maximum performance

73C for the GPU and 78C, the gpu.

the fan set at 64 with tpf

I think the ambient temperature is about 18-20C

What do you think?

here the screenshot
http://i943.photobucket.com/albums/ad28 ... a71a01.png

[Adda]

I need a bit more info, how did you test it?

There are several tests to do.

All of them involve furmark (I use v1.8.2) configured to run a "stability test" in "extreme burning mode", I run it at 640x480.
Prime95 should be configured for "in-place large FFTs" for maximum heat.
Use CPUz to monitor CPU core clocks.

1. Run furmark at described above, and four threads of prime95, keep going until temps have stabilized.
The four CPU cores should run at 1.73GHz during this test.

2. Run furmark (again same config) and two threads of prime95, use the task manager "set affinity" feature to select witch cores to test.
Try all combinations, fx. furmark and one thread of prime95 on core 0, another thread of prime 95 on core 1, and so on, till you have tired all combinations.
The two loaded cores should run at up to 2.4GHz in this test.

3. Furmark and one thread of prime95, on the same core, test all cores one by one (this is by far the most important test), again use the task manager "set affinity" feature to select witch cores to test.
Furmark and prime95 MUST run on the same core and that core should turbo boost to just short of 2.8GHz.

It looks to me like you tested with furmark in it's default configuration, and four (or more) threads of Prime95.

I'm getting curious though, what is the production date of your W510? It seems Lenovo put different video memory in W510 at some point (one of the BIOS updates mention support for new video memory).
You can see the production date under the bottom, on the sticker with a bar-code on it, right after S/N, mine is a 10/04, so just about four years old, and definitely an early model.

[Adda]

Here is a pic of a test run I just finished, I did all tests I described above.
My W510 has "CPU Power Management" enabled, "core parking" is enabled in the OS (default setting) the GPU is undervolted to 0.85v and overclocked to 594/1308/790MHz.

http://cdn.head-fi.org/a/aa/aa64fc55_stabilitytest.png

The CPU starts throttling quite a bit once it reaches 80C.
I found that core 0 can't park it self, so it is the only core that is able to run 2.8GHz, though usually it stays in the 2.6-2.7GHz range during testing, once it reaches 80C, it starts throttling like crazy.

If I test core 1,2,3 the CPU behaves as if it's running two threads, limiting the max clock to 2.4GHz, though usually it stays in the 1.73-2.3GHz range during testing.

CPUz didn't show the right clocks, Core Temp shows that all cores can turbo boost an equal amount, but rarely above 2.5GHz, I can't seem to make it go above 2.6GHz.
Once a core reaches 85C, it throttles down to around 2GHz, causing the temp to instantly dip to 80C and then clocks back up to 2.3-2.5GHz, and once the temp reaches 85C, well you get the idea.
Core 2 and 3 run hotter though, probably due to the cooler construction.
This is all single thread testing

This is different from earlier testing I have done (long time ago), at some point I could make ALL cores run at 2.8GHz, maybe something has been fixed, maybe my OS has a non-standard config, I'm not sure, someone will have to confirm this.
I was not thorough enough in my testing.

I think Lenovo made changes to the power management driver (was updated recently), to make sure the W540 doesn't crash and burn, and these changes affect other models as well.
I'm starting to doubt this is the case.

More testing is needed.

Edit: I should add the the GPU didn't throttle at all during testing.

Edit2: forgot to mention the the fan ran at max speed (64) throughout testing, and that the CPU protection feature in IBM_ECW was disabled.

Edit3: I should also add that the CPU did throttle to 933MHz a few times, but stayed above 1733MHz at most times.

[Adda]

Just finished a more extended run, and it really is hard to find a way to read the CPU clocks accurately, the truth is probably a mix of what CPUz and Core Temp shows.

The killer test is to run four threads of Prime95 until temps have stabilized, in my case, the GPU reached 80C and the CPU 84C.
Then I use the task manager "set affinity" feature to put furmark on core 3, and then all four threads of Prime95 on to core 3.
I then cycle through all the other cores in the same fashion.
Doing that causes CPU temps to soar, 93C max for core 3, 92C for core 2, 89C for core 0 and 1.
Fan speed was at max throughout the test (64).

http://cdn.head-fi.org/8/86/86bf0a23_exttest.png

The GPU runs quite a bit hotter at the default 0.93v, compared to 0.85v even if the GPU is overclocked beyond default clocks, at 0.85v.

Edit: my W510 was not powered down between this extended test, and the previous one.
The previous test was started not long after cold boot, this test was started after the system had warmed up.

[aldimeola81]

hello, I have done the test you asked, setting the affinity ', activate the turbo mode but the temperature does not rise much, the results you see them here.
http://s943.photobucket.com/user/aldime ... 4040333318


using a new version of Furmark at the native resolution, enabling all the cores and hyper trading, the maximum temperature is what you see here.

http://s943.photobucket.com/user/aldime ... 8452384184



more 'so the temperature does not rise, i've tried everything but no way.

build date is 10/10

sorry for the poor photo quality, that's photobucket recompress fault.
bye

[Adda]

The first test was too short, it takes at least 20 minutes to heat the system, but temps can still rise by a degree or two beyond 30 minutes of testing.

The second test looks similar to my 20 minute test with four threads and furmark.

I think your system will run at similar temps to mine if tested the same.

The new video memory for W510's came with BIOS v1.38, but I can't seem to find the release date.

You might want to disable Hyper Threading, it increases DPC Latency (a lot) and doesn't do games any good.

[aldimeola81]

What test do you want to see for more time? First or second? Bye

[Adda]

I always do all the test in each run, I use the "set affinity" thing to shuffle threads around, I don't try to prove that it's stable, I do everything I can to prove that it's not stable.

The killer test is to run four threads of Prime95 for at least 20 minutes (or until temps have stabilized), then open the task manager, and use the "affinity" feature to make furmark run on core 3, then put the four threads of Prime95 on to core 3 as well.

Alternatively, you can just run furmark and four threads of Prime95 for at least 30 minutes, that test is easy to replicate.
But if you do this, why not finish it off by putting Prime95 and furmark on core 3 for a grand finale.

The release date of BIOS v1.38 is 2011/06/13, so W510 mobo's made before that date must all use the same video memory.

[aldimeola81]

I have done hours of testing, if I select a single core, the temperature drops dramatically.

In this test I run on all cores:
Version 1.13 furemark 1920x1080 (last 25 min)
Furemark 1.8 640x480 for more than one hour
Prime 95 64bit (maximum heat production)
Original frequencies and voltages.
Try it yourself and post the result.
http://s943.photobucket.com/user/aldime ... sort=3&o=0

[Adda]

If I test a single core, temps drop as well, but if I go from testing four cores and go directly to a single core (no pausing the tests, in between) then the core I test will get very very hot before slowly cooling down.

I don't think two instances of furmark simultaneously will produce higher temps then one instance.
You have made your W510 run hotter then this, with a much shorter test.

I have already run all the tests I need to run, and I have done it many times over the last four years, I have tried to explain how you can replicate the tests I run, but without luck.

I'll try one last time.

1. make sure the CPU is running at max clock, use the windows power manager for this, or IBM_ECW, just make sure the CPU is running at max clock all the time, no matter the load.

2. run furmark v1.8.2, stability test in extreme burning mode 640x480.

3. run four threads of Prime95 "maximum heat".

4. wait for temps to stabilize (at least 20 minutes), then put furmark on to core 3, and then all four threads of Prime95 on to core 3 as well, don't stop the tests or anything, just use the "set affinity" feature in the task manager.

5. make sure the CPU and GPU isn't throttling.

[Adda]

Here is a straight 20 minute, four max heat threads of Prime95 + extreme burning mode Furmark v1.13.

http://cdn.head-fi.org/b/b0/b0fdc9eb_fm1.13.png

[aldimeola81]

Only 20 minute? Try with 1 hour if you can and add furemark 1.8 as me for do a compare
I've forgotten to say :I've modified w510 cooling system

[dr_st]

What exactly do all you guys hope to achieve by repeatedly subjecting your laptops to a torture test like Furmark, which is not even remotely representative of a typical workflow under load?

[Adda]

I have owned a W510 for four years, and it has never been completely stable until it could survive the tests I throw at it.
It might have been stable for weeks or months, but then it would suddenly suffer from slowdowns or even crashes, caused by high temperatures.

However, after I configured my W510 in a way so that I could handle these tests, it has been completely stable, even through hot summer days.

These tests stress the system more then MOST things, but it is not a completely unrealistic scenario, if it was, I would not get the results I have.
Starcraft 2 made my W510 run just as hot as furmark + four threads of prime95, due to a bug in the menu system, that caused extremely high framerates.

And this is a mobile workstation, it should handle every load you throw at it without throttling or crashing, just like the T500, T43p, R50p, A30 and every other performance oriented ThinkPad that came before it (that I have tired).

If there where no problems with the cooling system in the W510, I wouldn't care about these tests, but that is not the case, and it is not the case for W520/530/540 either.

If the cooler cannot handle the maximum heat output of the components it is cooling, then the cooler isn't good enough.

Edit: also, remember that we are running the fan at speed 64, that won't happen if the fan is controlled by BIOS, in fact, the BIOS fan controller seems to think that there is no reason to spin the fan up to speed 7 until 80C.

[dr_st]

Fair enough.

[aldimeola81]

I tried to disable the hypertrading the bios and I did some settings that do not remember, now the temperature has dropped quite a bit, so here is a test 1 hour with 3 programs.

games run even more 'fluid now

bye

http://i943.photobucket.com/albums/ad28 ... g~original

[Adda]

I think you have tested so much now that we can say that your W510 doesn't crash or overheat easily, so the cooler must be working as intended and running the fan at max speed, works well enough.

There are two options in the BIOS I know of that affect temperatures significantly, "CPU Power Management" witch enables/disables core parking, "Core Multiprocessing" witch enables/disables three out of the four CPU cores.
Disabling HT may reduce temps a bit too in certain situations.

[aldimeola81]

I think you have tested so much now that we can say that your W510 doesn't crash or overheat easily, so the cooler must be working as intended and running the fan at max speed, works well enough.

There are two options in the BIOS I know of that affect temperatures significantly, "CPU Power Management" witch enables/disables core parking, "Core Multiprocessing" witch enables/disables three out of the four CPU cores.
Disabling HT may reduce temps a bit too in certain situations.

say that you do not overheat it is too little, my w510 is a rock about the temperatures.

I made ​​a small mods just i've buyed it: I do not know if it helped or not, I think so.

I got in touch the heatsink on the CPU with the base of the keyboard using the thermal pads that I had taken from the t43.

I removed a piece of tape under the keyboard to expose the metal and I replaced it with a thermal adhesive panasonic.
for those who say that these tests are useless:

I sell and shelter thinkpad for years, and all models with 3d card nvidia or ati intel after a year die for lack of gpu resulting from the heat. (t41, t42, t43, t60, x60, T61P ...)

so if I think about it a bit 'before .. it is not a mistake, what do you think?

[dr_st]

I sell and shelter thinkpad for years, and all models with 3d card nvidia or ati intel after a year die for lack of gpu resulting from the heat. (t41, t42, t43, t60, x60, T61P ...)

so if I think about it a bit 'before .. it is not a mistake, what do you think?

Not sure I understand the claim you are making here.

Is it that all Thinkpads die due to GPU overheating?
Is it that it happens after only a year?
Is it that it's intentional?

[aldimeola81]

I sell and shelter thinkpad for years, and all models with 3d card nvidia or ati intel after a year die for lack of gpu resulting from the heat. (t41, t42, t43, t60, x60, T61P ...)

so if I think about it a bit 'before .. it is not a mistake, what do you think?

Not sure I understand the claim you are making here.

Is it that all Thinkpads die due to GPU overheating?
Is it that it happens after only a year?
Is it that it's intentional?

no, not after a year, but after a few years, maybe 3 / 6 (depending on usage and models)

for what I saw thinkpad die for two things:

gpu with cracked welds under the bga video card or cpu (series x) 80%

power issues on the motherboard, all of a sudden will not turn more 'and it all seems dead, (maybe depend from some mosfet power). 20%

I speak of the series from t40 onwards, preceded the series I never had.

This is the approximate average that I've noticed.

For example, now I have to hand a t60 with ati graphics card, the same defect T61P, who had the [censored] fx-570m graphics card ..

heating gpu the pc is turned on, otherwise nothing happens(black screen and error code)

same thing with t41-t42-t43, (no boot)and so on.

Only thinkpad with intel integrated graphics seem to be bombproof.

I do not know if it is intentional on the part of Lenovo and other (other manufacturers have identical defects)

certainly depends on the lack of lead in solder under the bga, and overheating of the GPU, it would be nice to see if before the 40 series, when it was still using lead in solder defect already 'existed or not.

so before your thinkpad abandon you, check the temperature of your laptop.

[dr_st]

Thank you for the detailed explanation.

I think you must work in very hot environments and/or subject your Thinkpads to unusually high stress.

Most Thinkpads I had with ATI GPUs lasted longer and haven't even started showing signs of GPU-related failures. This includes a T42 from 2005, a T60 from 2007, an X32 from 2006 - these are all units I currently possess. I've had a few other such units, which are no longer in my possession, but were 3+ years old when sold, and again had no issues.

The only laptops that failed/are failing on my hands are an A31p with GPU issues (which was assembled for me from hardware, most of which is probably 10+ years old), and an X32, whose board failed with the power-related symptoms you mentioned.

The collective experience also suggests that, while such failures do exist, and are more common in some models than in others, they are by no means a certainty or anywhere close to a certainty within the span of several years.

[Adda]

I think the problem with dying GPU's is a complicated one.

They heat up quick and cool down quick, especially moderns ones with aggressive power management.
Lead free solder did cause problems early on, but modern GPU run hotter, and cooler then older ones (due to advanced power management), this stresses the solder more.
I'm quite convinced that if we used the best old school leaded solder and compared it to modern lead free solder, there would be no competition, lead free would win as it is designed to work with modern chips with aggressive power management.

Of cause, there have been certain GPU's with problems that may be at least in part, caused by lead free solder, like the early Radeon GPU's up to and including Radeon 9600 (and similar GPU's).
And then the infamous nVidia FX570 and similar GPU's those seem to be the most plagued of all.

But laptop vendors are to blame as well, who at IBM thought that a peak GPU temperature above 90C was acceptable?
At least it didn't throttle, and many T43p (including my collection of 3) still work fine, they have southbridge problems in stead though.
But to be fair, the high GPU temps in T4xp's was caused by a design flaw in the heatshield, ATi is to blame, but IBM should have noticed and demanded it be fixed BEFORE they released laptops with a faulty GPU's in it.

The T60's would probably last longer with a bigger cooler, but at least they don't throttle.

the T/W500 would almost certainly last longer with a better cooler, they run pretty hot, and GPU failures seem relatively common in those models.
I think switchable graphics might play a part in that though, you want the GPU to stay at a relatively stable temperature, and turning the GPU on and off quite often, will probably cause more heat cycles.

The W510 runs extremely hot, but dead GPU's seem rare so far, but we can't really know for sure till more time has passed, when the model turns 7 years of age, we'll know more, same goes for W520.

As for the W530/540, who cares.

No what we need is that the IT industry starts taking cooling seriously, and use 85C as the max peak temp, no matter the load as long as the laptop is used under the conditions (air moisture, ambient temp and elevation) mentioned in the service manual.
The peak temp for the W510 is ~93C.

[aldimeola81]

Thank you for the detailed explanation.

I think you must work in very hot environments and/or subject your Thinkpads to unusually high stress.

Most Thinkpads I had with ATI GPUs lasted longer and haven't even started showing signs of GPU-related failures. This includes a T42 from 2005, a T60 from 2007, an X32 from 2006 - these are all units I currently possess. I've had a few other such units, which are no longer in my possession, but were 3+ years old when sold, and again had no issues.

The only laptops that failed/are failing on my hands are an A31p with GPU issues (which was assembled for me from hardware, most of which is probably 10+ years old), and an X32, whose board failed with the power-related symptoms you mentioned.

The collective experience also suggests that, while such failures do exist, and are more common in some models than in others, they are by no means a certainty or anywhere close to a certainty within the span of several years.

clearly I do not speak only of my thinkpad I've owned, but also the 50/60 that I had in assistance over the years, sooner or later, the defect occurs in thinkpad with gpu 3d.

the problem is that those with onboard video graphics have a really slow, even the menu 'windows and animations go in slow motion

Also my girlfriend has a t43 with ati card that works fine, heats as a bakery but 'no problem .. I have a t40 still working with ati card, but we're talking about computers that do not work hardly ever.
the same can 'say those who are the thinkpad collectibles.

another cause I think they are heat sinks, sometimes the conductors of heat punctured, exits the conductive liquid and the operating temperature becomes very high.


however as I said before the fault of the bga is generalized, computer, playstation, xbox, smartphones, macbook .. maybe it's a way designed to make them die after a while 'time.

[aldimeola81]

@adda:

Why do not you care about the W530-540? you think are scarce products?

according to a friend of mine who deals with reballing, the temperature is not the real problem, the real problem is the lack of lead into balls., lead makes the ball more 'elastic welds more' accurate, easy mounting tab without creating tensions within the installation is carried out at temperatures more 'low and do not hassle components etc..

if this were true models before 2005 so they should not suffer from these defects .. in fact my toshiba m30 2004 with NVIDIA GeForce FX GO 5200 has toured for years 24H a day, played with overclocking, the fan makes a noise hell, the keyboard works very badly but still works in the hands of a friend who keeps him always on.

I did a reflow on the t60 , bringing the temperature of the GPU at 260C for 10 seconds, now running again , who knows ' as last' ?

on ubuntu I installed a program that monitors temperature and speed ' fan :

the temperature takes a cycle every 2 minutes going from 40C - 90C- 40C etc. ..
clear that sooner or later, the expansion of the components does break some welding

the system should adapt to the load and stabilize the temperature , do not work this way.

[dr_st]

the problem is that those with onboard video graphics have a really slow, even the menu 'windows and animations go in slow motion

I must say I never experienced this issue. Not with drivers properly installed at least.