Well I got my attenuators in and they don't fix every PWM noise problem. I don't have that many extra fans around to test.
The NMB fans I'm using don't like PWM, any lower noise from less duty is canceled out by an increase in pwm-noise. Perhaps these fans are not balanced perfectly or something... I'm not sure they're totally comparable to other fans. (they cost me like $3 or $4)
I currently hooked up a fanmate-like controller set to ~6V after the tbalancer ALL the pwm-noise go away... the fans work between 18% adn 100% (of 6V) with no extra noise at all, <18% they fans just stop spinning.
I got some attenuators in, I tried swapping the voltage controllers with attenuators... lots of noise. Plus they seemed not to spin up as easily.. which I guess is to be expected..... PWM they get a nice 6V for an instant to start moving... attenuators give them a constant 3V or something and never can get spinning. This might be fine for some fans that arn't trying to go so slow but 95% of the time my fans are between 25-35% of 6V... I don't think the attenuator can do that.
And then for smething really wierd... I put bakc the voltage controllers... and I added the attenuator and i got noise back again at all settings.
Reallly this is just one fan, which might be a little unique, but for sure the attenuator won't fix every fan...
Review of mCubed T-Balancer in English
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
This entire thread got me very curious, so I just tried my 120mm fans with with the T-Bal. The fans I tried:Becks wrote:If anyone has the AL 120mm fans, evercool or any other brand they're sold udner.. is there any noise around 20-30% duty?
- Papst 4412. I tested this while still installed in my running secondary rig (not very scientific, I know). There was almost no resonance to be heard or felt. Sped up perfectly at 20%.
- Titan TFD-12025M12C (aluminium body). PWM noise and vibration clearly present, although somewhat dampened by the weight of the fan. Sped up at ~20% but very (PWM) noisy at that speed. Became bearable at 40% but already too much airflow for my purposes and still vibrating.
- Adda AD1212LB-A73GL (ball bearing). These are the standard fans that came with my PC-V1000. No noticable PWM noise or resonance (held fan in hand). Spins up at 15%, noise level similar to Papst up to ~40% (which is more that I use).
For all 3 the attenuator did not help much if anything. The Titan (Evercool) did not spin up until about 45%, the others at around 30%.
Since this made me even more curious I decided on another "test" (hardly worth the title).Becks wrote:Well I got my attenuators in and they don't fix every PWM noise problem. I don't have that many extra fans around to test.
The 80mm fans
I tried a Papst 8412NGMLE, a Titan TFD-8025M12C (plastic, not aluminium) and some brandless thing. All have more PWM noise than the bigger fans. Without the attenuator the Titan has least PWM noise, but for these fans the attenuator works. Noise and resonance are not eliminated but are lessened. Minimum spin-up speeds go up to about ~35% (from about 20%) but that's worth it.
I also tried replacing it with a Fanmate-II. I found this to have less influence (but still some) on the PWM noises, but this differed between the fans. The attenuator does seem to mess up the rpm detection of the T-Bal though, haven't really looked into this.
conclusion?
Does this ramble need it? Probably. I'll make it as short as possible.
- Attenuators work for 80mm fans and smaller, but effect varies from brand to brand;
- Attenuators don't add much for 120mm fans but do raise minimal rpm considerably;
- The Titan/Evercool alu's are better suited for regular undervolting than PWM;
- (these) Papst and Adda 120mm's are comparable at lower rpm's but the Papst wins at higher speeds noisewise; Adda seems to have even less PWM noise;
- I'm still in love with the T-Balancer.
Don't buy extra fans based on my experience; my remark had to do with the 8412's I've owned. I only have 2 4412's and am happy with them but they may or may not suffer from the same inconsistent quality; I just don't know. Besides, they're expensive enough.DrCR wrote:So right now, for the heatercores, I'd like to pair them with the Papst 4412. Think I’ll have to buy extra ones to make sure I get quiet ones?
IMO and based on my "tests", the Titan/Evercool is not an option for case fans with the T-Bal. Since I only use my current case fans at 20-30%, the lower airflow of the Nexus fans would be ideal I think.
Will do! I'll probably order them tomorrow and simply make time for these important comparisons!DrCR wrote:Yeah, I would love to hear how you like the Nexus fans when you get them in! PM me or reply to this thread when you post about them would you?
Contacted mCubed today to see whether they couldn't ship a T-Balancer out to MikeC et al. for testing... but they already beat me to it.
I just hope one of MikeC's newly appointed apprentices will get around to it...mCubed Sales and Support wrote:we are trying since 2 months to get a review by mike chin, who is the editor of silentpcreview. They have a test sample over there. But he has a lot of products and ours is still not the next!
Got my Nexus fans in today (both 120mm and 80mm) and I am very impressed! So impressed that I did another small fan comparision using those scientific tools of yore: hands & ears.
I am very happy to report that both sizes Nexus fans work absolutely smoothly without audible PWM resonance. Of course the Nexus fans have lower airflow than my other fans (the Nexus has rpm/cfm of 1000/37 whereas the Adda has 1800/72 and the Papst 1600/55) so I compensated fan speeds for the comparision. I compared the beforementioned 120mm fans with the Nexus fan; I ran the Nexus at 100% to test noise compared to the others at 50% and ran all at 25% to compare resonance. The Nexus is more quiet at comparable airflow and does not appear to resonate at all.
Same goes for the 80mm basically: it does need higher "voltage" (I know, PWM means 12V all the way), around 35% to get started but after that the Nexus is again unbeatable versus Papst, Titan and an Enermax I dug up (PL80B12HH). The Papst 80mm's tend to resonate a lot more and need an attenuator. Even with an attenuator the Papsts are not resonance-free, whereas the Nexus does not need an attenuator at all.
Btw, this time I did take the 120mm Papst out of the server to test it: it also had some tendency to resonate due to PWM, but nowhere near as much as the Titan or the 80mm Papst. It ran smoothly (but with more noise than the Nexus).
My very simple conclusion: if you do not need higher airflow than what Nexus fans offer, they are among the best choice you could make and combine wonderfully with PWM.
Something else: I looked into rpm monitoring more thoroughly. The attenuator plays havoc with the T-Bal's ability to monitor fan speed and blockage at low speeds. Not a big deal as long you don't run the fan at near-stopping speed. Also found out that only one of the two fan connectors on the attenuator passes rpm info along.
Now if you'll excuse me: I have some fans to build in
edit: fixed some spelling & unclear sentences
I am very happy to report that both sizes Nexus fans work absolutely smoothly without audible PWM resonance. Of course the Nexus fans have lower airflow than my other fans (the Nexus has rpm/cfm of 1000/37 whereas the Adda has 1800/72 and the Papst 1600/55) so I compensated fan speeds for the comparision. I compared the beforementioned 120mm fans with the Nexus fan; I ran the Nexus at 100% to test noise compared to the others at 50% and ran all at 25% to compare resonance. The Nexus is more quiet at comparable airflow and does not appear to resonate at all.
Same goes for the 80mm basically: it does need higher "voltage" (I know, PWM means 12V all the way), around 35% to get started but after that the Nexus is again unbeatable versus Papst, Titan and an Enermax I dug up (PL80B12HH). The Papst 80mm's tend to resonate a lot more and need an attenuator. Even with an attenuator the Papsts are not resonance-free, whereas the Nexus does not need an attenuator at all.
Btw, this time I did take the 120mm Papst out of the server to test it: it also had some tendency to resonate due to PWM, but nowhere near as much as the Titan or the 80mm Papst. It ran smoothly (but with more noise than the Nexus).
My very simple conclusion: if you do not need higher airflow than what Nexus fans offer, they are among the best choice you could make and combine wonderfully with PWM.
Something else: I looked into rpm monitoring more thoroughly. The attenuator plays havoc with the T-Bal's ability to monitor fan speed and blockage at low speeds. Not a big deal as long you don't run the fan at near-stopping speed. Also found out that only one of the two fan connectors on the attenuator passes rpm info along.
Now if you'll excuse me: I have some fans to build in
edit: fixed some spelling & unclear sentences
Last edited by teejay on Tue Oct 05, 2004 2:03 pm, edited 1 time in total.
I got my T-Balancer the other day, and installed it today.
At the end of the week I'll get some new 80mm fans so I'll have more to say then.
Currently I tested it with the small stock fan and heatshink, using a 1.13ghz AMD Athlon CPU.
At 26 degrees ambient, I found out I could use the stock fan at 15% (which is ~700 rpm, out of a maximum of 4500 rpm), and still get no higher temps than 65 degrees, under full load (CPUburn for 1 hour). Apart from some clicking noise, the fan seems to be silent at such low rpms. I guess with a high quality heatsink and 80mm fan I can get practically silent operation, but the whole point is that I want to upgrade the CPU. Later this week I'll experiment with faster CPUs.
On the down side, the temperature sensors seems to be off, really, really off. I think I understand why. First, the facts:
When I let the sensors measure the ambient temp, they all read ~28 degrees, which is only 2 degrees higher than the real temp in my room (26 degrees). This is all within specs and no complaints here. The problems start when you actually try to measure computer parts. I let my PSU run without any cooling devices for a while (don't try that at home), after several minutes it became so hot, I couldn't even put my finger 1 second on it, without screaming from pain. Anyway, I installed the sensor on the place I put my finger, and to my surprise the T-Balancer reported a "cool" 51.5 degrees (and remember, the ambient temp reading was precise with the T-balancer sensor, and I DID mount it the right way on the PSU).
I did another test, I put the sensor on the CPU's heatsink, while the CPU's internal sensor reported 61 degrees, the T-balancer reported 36 degrees (and while I agree the heatsink will never be as hot as the actuall CPU, still the heatsink was definitely not 36 degrees celsius, it was at least 50 degrees, because it was pretty hot to the touch). I really didn't care about it, as long as it stays linear. But it didn't. When I lowered the fan's rpm, and the CPU's internal sensor read 71 degrees, the T-Balancer reading raised only to 38 degrees. So, for about every 5 degrees celsius, the T-Balancer's sensor felt only a raise of 1 degree. And I'm not sure this linearity is constant either.
Now, I think the reason why this is happening is because the T-Balancer's sensor head is engulfed inside a small plastic case. Apparantly It must be like this, because this sensor is digital, and needs to be powered on with electricity, and without the plastic case on it, it might short your pc components when it is activated. The plastic case does not conduct electricity, so it solves this problem. Only problem is, plastic does not conduct heat either, so what you get is a temp sensor, which is insulated inside a plasitc case, kind of beats the purpose, don't you think ?.... This explains why it can give precise ambient measurements, but becomes totally off (by 15-30 degrees celsius) while trying to meausre a hot panel.
Overall the T-balancer's ability to reduce rpms down to a full halt is very impressive, the total control it gives you over the fans rpm is very important, and can not be found at any other product that I know of today. but the digital sensor is so off, it really takes a lot from the product.
At the end of the week I'll get some new 80mm fans so I'll have more to say then.
Currently I tested it with the small stock fan and heatshink, using a 1.13ghz AMD Athlon CPU.
At 26 degrees ambient, I found out I could use the stock fan at 15% (which is ~700 rpm, out of a maximum of 4500 rpm), and still get no higher temps than 65 degrees, under full load (CPUburn for 1 hour). Apart from some clicking noise, the fan seems to be silent at such low rpms. I guess with a high quality heatsink and 80mm fan I can get practically silent operation, but the whole point is that I want to upgrade the CPU. Later this week I'll experiment with faster CPUs.
On the down side, the temperature sensors seems to be off, really, really off. I think I understand why. First, the facts:
When I let the sensors measure the ambient temp, they all read ~28 degrees, which is only 2 degrees higher than the real temp in my room (26 degrees). This is all within specs and no complaints here. The problems start when you actually try to measure computer parts. I let my PSU run without any cooling devices for a while (don't try that at home), after several minutes it became so hot, I couldn't even put my finger 1 second on it, without screaming from pain. Anyway, I installed the sensor on the place I put my finger, and to my surprise the T-Balancer reported a "cool" 51.5 degrees (and remember, the ambient temp reading was precise with the T-balancer sensor, and I DID mount it the right way on the PSU).
I did another test, I put the sensor on the CPU's heatsink, while the CPU's internal sensor reported 61 degrees, the T-balancer reported 36 degrees (and while I agree the heatsink will never be as hot as the actuall CPU, still the heatsink was definitely not 36 degrees celsius, it was at least 50 degrees, because it was pretty hot to the touch). I really didn't care about it, as long as it stays linear. But it didn't. When I lowered the fan's rpm, and the CPU's internal sensor read 71 degrees, the T-Balancer reading raised only to 38 degrees. So, for about every 5 degrees celsius, the T-Balancer's sensor felt only a raise of 1 degree. And I'm not sure this linearity is constant either.
Now, I think the reason why this is happening is because the T-Balancer's sensor head is engulfed inside a small plastic case. Apparantly It must be like this, because this sensor is digital, and needs to be powered on with electricity, and without the plastic case on it, it might short your pc components when it is activated. The plastic case does not conduct electricity, so it solves this problem. Only problem is, plastic does not conduct heat either, so what you get is a temp sensor, which is insulated inside a plasitc case, kind of beats the purpose, don't you think ?.... This explains why it can give precise ambient measurements, but becomes totally off (by 15-30 degrees celsius) while trying to meausre a hot panel.
Overall the T-balancer's ability to reduce rpms down to a full halt is very impressive, the total control it gives you over the fans rpm is very important, and can not be found at any other product that I know of today. but the digital sensor is so off, it really takes a lot from the product.
Yea the temp probes have issues, the probe itself doesn't really come in contact with what its measuring.. they're kinda big so you can't fit tehm between fins on a heatsink for example... I've got 8 of them and use 2 on the side of my radiators (little redundancy) and one on my harddrive... the rest are really doing nothing hehe. Mostly because with issues I have no use for them, but also becasue I have nothing else I want to measure.
If you've ever played with other temp probes, the flat/thin ones, those (imo at least) are many times more useful. Taping one to the bottom of a heatsink near the die.. that'll be many times more accureate than the big temp probe things stuck on the side of the heatsink.
ANYWAYS...
"We are developing at the moment a water cooling kit add-on incl. water
temp probes, which are directly in the water via a t-part and incl. a
flowmeter and some more analog thin temp probes!"
So yay on that... something to look forward to
If you've ever played with other temp probes, the flat/thin ones, those (imo at least) are many times more useful. Taping one to the bottom of a heatsink near the die.. that'll be many times more accureate than the big temp probe things stuck on the side of the heatsink.
ANYWAYS...
"We are developing at the moment a water cooling kit add-on incl. water
temp probes, which are directly in the water via a t-part and incl. a
flowmeter and some more analog thin temp probes!"
So yay on that... something to look forward to
Sad the Papst 4412 don't do to well on PWM. Even if I don't go for PWM and just use good old undervolted DC, I would hate to have to buy new fans if I ever go an all PWM route.
Glad to hear the Nexus fans do so fine on PWM though! Nexus fans are my 80mm and 92mm of choice. Thanks again teejay.
Or perhaps you could cut off their probes and splice on your own nonplastic coated probes? (And use something like thermal tap to affix the probe)
DrCR
__________
Yeah, 12V airflow is one of the reasons I'll probably go for the Globe offering for my heatercores. True, I doubt I'll actually need more than ~35CFM (about Nexus 120mm fan's max at 12V), but I would like to have it if needed. Guess you could say it's all in my head lol. I'll probably use the Globe for my 120mm case fans needs as well, instead of a Nexus, if only due to asthetics (orange won't work in my format).teejay wrote:My very simple conclusion: if you do not need higher airflow than what Nexus fans offer, they are among the best choice you could make and combine wonderfully with PWM.
Glad to hear the Nexus fans do so fine on PWM though! Nexus fans are my 80mm and 92mm of choice. Thanks again teejay.
Could you just remove the plastic or would sanding be required? I use thermal pads to attach digital probes -- heat conductive and an electrical insulator (not electrically conductive).jones_r wrote:...Now, I think the reason why this is happening is because the T-Balancer's sensor head is engulfed inside a small plastic case. Apparantly It must be like this, because this sensor is digital, and needs to be powered on with electricity, and without the plastic case on it, it might short your pc components when it is activated. The plastic case does not conduct electricity, so it solves this problem. Only problem is, plastic does not conduct heat either, so what you get is a temp sensor, which is insulated inside a plasitc case...
Or perhaps you could cut off their probes and splice on your own nonplastic coated probes? (And use something like thermal tap to affix the probe)
DrCR
__________
http://www.mcubed-tech.com/eng/images/es_big.jpg
If you don't have one thats what the probes look like.. on the other side is a chip and some other little things (resistors and bleh, and a mini-led). ANyways you can see the plastic.. its like vacuum formed stuff... its real thin and kinda stiff, you could peal it off easy. Its only purpose is to not let anything short out, its not holding anything together. If you used thermal tape there wouldn't be any problems. (as far as I can tell)
With our without removing the plastic themral tape should make a difference however... if you look at the side (the one not shown in the pic) its not even flat... thermal tape should get rid of the airgaps... and if you remove the plastic the thermal tape might form a little better around the temp probe thing as well.
On mine I also stuck it to my radiator and then put electircal tape over it to seal of off from any airflow near it... imo there was some airgap issues and I was afraid if any air flowed though those gaps the temps would be even more off.
If you don't have one thats what the probes look like.. on the other side is a chip and some other little things (resistors and bleh, and a mini-led). ANyways you can see the plastic.. its like vacuum formed stuff... its real thin and kinda stiff, you could peal it off easy. Its only purpose is to not let anything short out, its not holding anything together. If you used thermal tape there wouldn't be any problems. (as far as I can tell)
With our without removing the plastic themral tape should make a difference however... if you look at the side (the one not shown in the pic) its not even flat... thermal tape should get rid of the airgaps... and if you remove the plastic the thermal tape might form a little better around the temp probe thing as well.
On mine I also stuck it to my radiator and then put electircal tape over it to seal of off from any airflow near it... imo there was some airgap issues and I was afraid if any air flowed though those gaps the temps would be even more off.
I made some more tests today.
I removed the plastic case from one of the sensors, to see if it is the culprit. Well, it is not. The sensor without the plastic case measured +/- 2 degrees celsius in comparisson to the one with the case. Furthermore, the sensor without the plastic case is kind of useless. When I mounted it on the PSU, the moment it touched the PSU, the T-balancer went totally crazy, all the fans started to shift between slow and fast rpm's. Conclusion: this sensor should NOT run without electrical barrier between it and the pc component it measures.
Well, I basically came to the conclusion that these M-Cubed sensors just suck in telling you the real temps. Don't expect them to report more than a 3-4 degrees celsius increase for an actual 10-20 degrees raise. They are sensitive as the hearing of an 85 y/o WW-II veteran. Also, because they can't be mounted like the thin analog sensors, they might be 30-40 degrees off from the real temperature.
The only way to use these sensors is by a qualitive mesure and not a quantitive one. For the CPU for example, if you don't want temps to ever go above 65 degrees, see what the T-balancer reads for this temp (for me it is 38 degrees), and just design the T-Balancer's fan curve so that you'll get a constant 38 degrees for the CPU. You should be very carefull not to pass 38 degrees even by 1 degree ceslius, cause on a bad T-balancer's day, it might translate into an actual 10 degrees celsius raise.
Does anyone know if you can give the T-Balancer a certain temp you want to keep constant, and it will adjust the fans in real time in order to accomplish that ? (without creating a pre-made curve).
I removed the plastic case from one of the sensors, to see if it is the culprit. Well, it is not. The sensor without the plastic case measured +/- 2 degrees celsius in comparisson to the one with the case. Furthermore, the sensor without the plastic case is kind of useless. When I mounted it on the PSU, the moment it touched the PSU, the T-balancer went totally crazy, all the fans started to shift between slow and fast rpm's. Conclusion: this sensor should NOT run without electrical barrier between it and the pc component it measures.
Well, I basically came to the conclusion that these M-Cubed sensors just suck in telling you the real temps. Don't expect them to report more than a 3-4 degrees celsius increase for an actual 10-20 degrees raise. They are sensitive as the hearing of an 85 y/o WW-II veteran. Also, because they can't be mounted like the thin analog sensors, they might be 30-40 degrees off from the real temperature.
The only way to use these sensors is by a qualitive mesure and not a quantitive one. For the CPU for example, if you don't want temps to ever go above 65 degrees, see what the T-balancer reads for this temp (for me it is 38 degrees), and just design the T-Balancer's fan curve so that you'll get a constant 38 degrees for the CPU. You should be very carefull not to pass 38 degrees even by 1 degree ceslius, cause on a bad T-balancer's day, it might translate into an actual 10 degrees celsius raise.
Does anyone know if you can give the T-Balancer a certain temp you want to keep constant, and it will adjust the fans in real time in order to accomplish that ? (without creating a pre-made curve).
I've asked them that same question. They are planning on it but don't have that yet. I suppose one could write a custom software solution overtaking the manual control but that kind of defeats the idea of an autonomous fan controller.
As to your sensor issues: this is not a happy story!
Have you contacted mCubed on this matter? They often react within a day.
My personal experience is somewhat different: after calibration my T-Bal cpu sensor (mounted on the side of the heatsink close to the cpu) reads the same as the socket diode, from startup (~30C by the time Windows boots) up 'til 55 with the cpu fan temporarily off (mbm reported chip temp to be 64C at that point, time to switch the fan back on). There is some delay in the measurement between chip temp and heatsink temp but that is to be expected: it takes a while to cool down after all. Same with my VGA Silencer: it starts at 30 and goes up to 60 while 3D-ing (I have no temp measurement for the gpu but these temps are similar to my previous card which did have one). I am starting to doubt the validity of my own observations however...
Would really like to know what mCubed has to say in this matter... if only MikeC would get around to testing his T-Balancer sample
As to your sensor issues: this is not a happy story!
Have you contacted mCubed on this matter? They often react within a day. My personal experience is somewhat different: after calibration my T-Bal cpu sensor (mounted on the side of the heatsink close to the cpu) reads the same as the socket diode, from startup (~30C by the time Windows boots) up 'til 55 with the cpu fan temporarily off (mbm reported chip temp to be 64C at that point, time to switch the fan back on). There is some delay in the measurement between chip temp and heatsink temp but that is to be expected: it takes a while to cool down after all. Same with my VGA Silencer: it starts at 30 and goes up to 60 while 3D-ing (I have no temp measurement for the gpu but these temps are similar to my previous card which did have one). I am starting to doubt the validity of my own observations however...
Would really like to know what mCubed has to say in this matter... if only MikeC would get around to testing his T-Balancer sample
Ok, here is the question I emailed mCubed:
Hello,
I've just reveived the T-Balancer, and tried it today.
I love the full control it gives me over RPMs. It really works.
The thing which annoys me about the T-Balancer is the temprature sensors. How could you pick such lousy sensors for the T-Balancer ?. First, they are HUGE, and so you can't mount it in tight places in order to get good readings. Furthermore, the round white mounting material that you provided for the sensors, slowly loses its grip when the temps go up high, and so the sensor sometimes fall from where it was mounted, and even if it doesn't fall, it lose the good grip it had before, and so the readings become even less accurate.
Talking about the sensor's accuracy, this thing has NO accuracy at all. Calibration doesn't help much, because the sensor problem is that it is not sensitive to heat. In order for the T-Balancer's sensor to go up 1 degree celsius, the real temperature sometimes need to go up 5-6 degree celsius. This is crazy!!!.
In conclusion, the sensor you picked for the T-Balancer is WAY too big, impossible to mount without later movements with the sticks you provided, and not accurate after calibration, to the point that using it becomes dangarous to the computer.
One last thing, why haven't you provided a feature which allows you to set a constant temperature for each computer component, and let the T-Balancer automatically keep this temperature constant, using real time fan corrections ?, at least this way it wasn't so dangerous to use the temp sensors...
And here is their answer:
Dear Jones,
Yes, you are right, in some cases it is better to have smaller sensors which you can place nearer to the heatsource! We are working on it! We will offer in about 1 month an extension kit with thin foil analog sensors. They will react more actively.
To get along with the digital ones I have some advices:
- take superglue to fix them
- take the cable binder to fix them to the heatsink
The effect with 1:5° may happen over a short while, because the sensors will react more slowly if they are fare away from the heatsource.
The integrated sensors in the motherboard are not always true: we had also effects, where our digital sensors showed a higher temperature than the internal ones!
The digital sensors are very accurate over the whole temperature range (+/-1°), with calibration (+/-0,1°). They really show the actual temperature. It is only a question how to place them.
Your mentioned last feature of controlling all fans automatically to a target temperature (without curve): we will implement it in one of our next firmwares. Then you can also update yours and it works with your TBAN.
We hope, that we could help you for so long and we will keep you informed about the thin analog sensors!
Best regards,
Maik!
Maik Berendt
Sales and Support
Hello,
I've just reveived the T-Balancer, and tried it today.
I love the full control it gives me over RPMs. It really works.
The thing which annoys me about the T-Balancer is the temprature sensors. How could you pick such lousy sensors for the T-Balancer ?. First, they are HUGE, and so you can't mount it in tight places in order to get good readings. Furthermore, the round white mounting material that you provided for the sensors, slowly loses its grip when the temps go up high, and so the sensor sometimes fall from where it was mounted, and even if it doesn't fall, it lose the good grip it had before, and so the readings become even less accurate.
Talking about the sensor's accuracy, this thing has NO accuracy at all. Calibration doesn't help much, because the sensor problem is that it is not sensitive to heat. In order for the T-Balancer's sensor to go up 1 degree celsius, the real temperature sometimes need to go up 5-6 degree celsius. This is crazy!!!.
In conclusion, the sensor you picked for the T-Balancer is WAY too big, impossible to mount without later movements with the sticks you provided, and not accurate after calibration, to the point that using it becomes dangarous to the computer.
One last thing, why haven't you provided a feature which allows you to set a constant temperature for each computer component, and let the T-Balancer automatically keep this temperature constant, using real time fan corrections ?, at least this way it wasn't so dangerous to use the temp sensors...
And here is their answer:
Dear Jones,
Yes, you are right, in some cases it is better to have smaller sensors which you can place nearer to the heatsource! We are working on it! We will offer in about 1 month an extension kit with thin foil analog sensors. They will react more actively.
To get along with the digital ones I have some advices:
- take superglue to fix them
- take the cable binder to fix them to the heatsink
The effect with 1:5° may happen over a short while, because the sensors will react more slowly if they are fare away from the heatsource.
The integrated sensors in the motherboard are not always true: we had also effects, where our digital sensors showed a higher temperature than the internal ones!
The digital sensors are very accurate over the whole temperature range (+/-1°), with calibration (+/-0,1°). They really show the actual temperature. It is only a question how to place them.
Your mentioned last feature of controlling all fans automatically to a target temperature (without curve): we will implement it in one of our next firmwares. Then you can also update yours and it works with your TBAN.
We hope, that we could help you for so long and we will keep you informed about the thin analog sensors!
Best regards,
Maik!
Maik Berendt
Sales and Support
I can only agree
Well I stuffed my CPU sensor next to were my Hyper 6 heatsink meets the CPU and so far MBM has measured temps from 43 degrees C in idle to 63 after using CPU burn for 20 minutes. T Balancer was calibrated to 43 degrees C in idle and still records 43 when MBM says 63. I dont know how to make use of this, it's dangerous as you say.
Stajo, it didn't move even 1 degree up ?, you are talking about a 20 degree increase here, for such an increase I usually get about a 3 degree increase with the T-balancer's sensor.
What happened to you is exactly why I decided to calibrate the T-Balancer's sensor to the maximum CPU temp, and not to the idle (i.e I calibrated it after running CPUburn for several minutes). The downside for this is that the T-balancer's sensor always report more than 60 degrees, even when the real temp is way below, but at least it's safe...
There is something really wrong with the T-balancer sensor. Maybe they have bad quality control. I asked them about it and I'm waiting for a reply.
What happened to you is exactly why I decided to calibrate the T-Balancer's sensor to the maximum CPU temp, and not to the idle (i.e I calibrated it after running CPUburn for several minutes). The downside for this is that the T-balancer's sensor always report more than 60 degrees, even when the real temp is way below, but at least it's safe...
There is something really wrong with the T-balancer sensor. Maybe they have bad quality control. I asked them about it and I'm waiting for a reply.
Jones, I followed your example and I reposisioned the sensor even closer to the CPU, under the heatpipes. This time MBM measured 56 degr max after 20 mins CPU Burn (due to higher fanspeed) so I calibrated T-Ban to that. Then I thougt if I choose a progressive curve the fan would go down to slower rpm but the T-Ban measures 52,5 degr at idle (MBM 42). 3,5 degrees differense for T-Ban compared to 14 degrees for CPU Diode. The fan went down to 80% from 90% at full load and that's still to high and noicy for idle mode. I can only hope for thoose analog smaller sensors mCubed announced to come soone, cause those sensors are useless.
The sensors are kinda bad but they're not horrible... on my radiators i have 2 tbalancer sensors and 2 thin temp probes in different spots...
all 4 sensors are within 1C of eachother or so... the thin ones react alot faster.. well they'll show a rise and then a few seconds later the tbalancer will start rising... they never really get that far off from eachother... the sensors are accurate just hard to get them to read what you want.
Whatever your stickin it on... when you put your finger on there do you feel much of a temp change?
all 4 sensors are within 1C of eachother or so... the thin ones react alot faster.. well they'll show a rise and then a few seconds later the tbalancer will start rising... they never really get that far off from eachother... the sensors are accurate just hard to get them to read what you want.
Whatever your stickin it on... when you put your finger on there do you feel much of a temp change?
This is exactly what happened for me too.3,5 degrees differense for T-Ban compared to 14 degrees for CPU Diode.
Stajo, the T-balancer can still do miracles with this 3.5 degrees change. In order to get a normal quiet curve, set the maximum temp of the T-balancer to 60 degrees, and then go to the curve's table, and delete all the temperature values there (except 0 and 60, which you won't be able to delete), and now set new temp values of 58, 56, 54, 52, 50. And now set an appropriate fan speed for each temp value.
This way even with such a small temp range (52.5-56) you'll still get a quiet operation.
Becks, how exactly did you mount the T-balancer sensors to the radiator ?, have you used the round white mounting sticks which were supplied with the T-Blalancer ?.all 4 sensors are within 1C of eachother or so... the thin ones react alot faster.. well they'll show a rise and then a few seconds later the tbalancer will start rising... they never really get that far off from eachother... the sensors are accurate just hard to get them to read what you want.
BTW you don'/t have to pick those generic curves.. click the advanced button
The temp sensors also don't have to be perfect... if you got it so when the cpu was loaded the temp sensor always read 56C and when it was idle it was always 52C thats enough info.
My goal was to keep the water at 32C (deltaT for cpu/water is at most maybe 12C.. giving a max load temp 44C.. sometimes water will get to 34C or so... max load temp 46C yay)
So for the curve I did
_TEMP 00 30 31 32 33 34 035
SPEED 20 22 25 35 50 80 100
If you look at this curve... its pretty flat and starts increasing fast at 31C and after 32C it increases very fast... that's pretty much what will keep something at a certian temp. The temp is pretty much always 32 or 33C .. load or idle.. (at night when room temp drops alot it'll go down to 28-30 but the fans are barely spinning so i don't really care)
If your trying to make a curve for a heatsink.. I'd find out what is the max fan speed needed... running cpuburn do manual and find out how fast it needs for 50C load temp.. if 80% keeps it good then write that down... then read what tblancer is reading and set that temp @ 80%.. 1C more set to 100%... 1C less set to 70% 2C less set to 50% I dunno you get the idea..
If you play with the curves you can get some better results... then again if you really can't get any good temp reads then that wont help. The thin probes will/should work way better on heatsinks, you should be able to get them very close to the die.. but I'm pretty sure you should get some results from the current temp probes.
--edit--
doh i was writing this when you posted... same idea heh
on each end of the radiator is a 'tank' the things the barbs are screwed into.. i just put the probe on those... they're full of water.. works pretty well. I didn't use the white things (kinda looked junky to me) I just held them against the rad and then put tape over them (black electircal tape).
The thin temp probes I suck down in the fins so they've got heat on all sides of them.. which I would think would be more accurate.. I could see how air flwoing around the outside could potentially affect the tbalancer's temp probes.. tho really I don't see much of a difference.
The temp sensors also don't have to be perfect... if you got it so when the cpu was loaded the temp sensor always read 56C and when it was idle it was always 52C thats enough info.
My goal was to keep the water at 32C (deltaT for cpu/water is at most maybe 12C.. giving a max load temp 44C.. sometimes water will get to 34C or so... max load temp 46C yay)
So for the curve I did
_TEMP 00 30 31 32 33 34 035
SPEED 20 22 25 35 50 80 100
If you look at this curve... its pretty flat and starts increasing fast at 31C and after 32C it increases very fast... that's pretty much what will keep something at a certian temp. The temp is pretty much always 32 or 33C .. load or idle.. (at night when room temp drops alot it'll go down to 28-30 but the fans are barely spinning so i don't really care)
If your trying to make a curve for a heatsink.. I'd find out what is the max fan speed needed... running cpuburn do manual and find out how fast it needs for 50C load temp.. if 80% keeps it good then write that down... then read what tblancer is reading and set that temp @ 80%.. 1C more set to 100%... 1C less set to 70% 2C less set to 50% I dunno you get the idea..
If you play with the curves you can get some better results... then again if you really can't get any good temp reads then that wont help. The thin probes will/should work way better on heatsinks, you should be able to get them very close to the die.. but I'm pretty sure you should get some results from the current temp probes.
--edit--
doh i was writing this when you posted... same idea heh
on each end of the radiator is a 'tank' the things the barbs are screwed into.. i just put the probe on those... they're full of water.. works pretty well. I didn't use the white things (kinda looked junky to me) I just held them against the rad and then put tape over them (black electircal tape).
The thin temp probes I suck down in the fins so they've got heat on all sides of them.. which I would think would be more accurate.. I could see how air flwoing around the outside could potentially affect the tbalancer's temp probes.. tho really I don't see much of a difference.
This is starting to sound (to me) more and more like a quality issue with the sensors, since my (8) sensors seem to work fine. CPU heatsink temp (T-Balancer) and mobo sensor are always in sync over a range of ~20C with a delay of max 1min, if I suspend my machine and wake it up an hour later I can see the hdd sensors go from 21C to 32ish in matter of minutes, my VGA Silencer sensor reports in ranges from ~30 at startup to >60 under load, if I hold my temp-outside-the-case sensor it goes immediately from 22 to ~34, etc.
What I was wondering about: if you set the response curve to change fan speed significantly within a few degrees, don't you get a hysteresis effect? How have you guys set your hysteresis range? 2 degrees?
Are all these problems on models with the latest firmware? (stupid question, I know...)
What I was wondering about: if you set the response curve to change fan speed significantly within a few degrees, don't you get a hysteresis effect? How have you guys set your hysteresis range? 2 degrees?
Are all these problems on models with the latest firmware? (stupid question, I know...)
Heat conductive glueing
I must admit that I under my experimenting with the sensors has only "stuffed" the sensors as close to the heatsource as possible. My guess is, as some former speakers suggests, that the trick is to get the sensor as tight to the heat source as possible (ie no air between). I know no one will compare me to Einstein in that conclusion but I'd really want to know about best practices here. Superglue, is that heat conductive enough or is there better compounds? Is heatpipes the best place for measuring?
In my t-Balancer the temperatures change as expected, more or less the same amount peak to peak than the MBM readings but with the differences due to the sensor being mounted at the side of the CPU heatsink (delay in readings, not completely lineal compared to the diode temp, etc.).
I think it will make a big difference if the temperature sensor is not properly attached to the surface. I use the sticky dots that came with the t-Balancer and they work fine with my HDD and CPU but tend to get loose on my Video card (that Zalman HS doesn't have a flat surface to place the sensor).
I think it will make a big difference if the temperature sensor is not properly attached to the surface. I use the sticky dots that came with the t-Balancer and they work fine with my HDD and CPU but tend to get loose on my Video card (that Zalman HS doesn't have a flat surface to place the sensor).
Thank you rperezlo! At least now I know I am not the only one whos sensors appear to work as expected... I also used the sticky pads and attach them on a flat surface, with the chip side facing towards the pad. My only gripe is that the pads are almost impossible to remove without leaving residue behind.
I myself would never use superglue since removing would become almost impossible... and I tend to swap components rather often
I myself would never use superglue since removing would become almost impossible... and I tend to swap components rather often
About those sticky pads, I did a small experiment. I used an old reading lamp, which gets too hot for the touch after several minutes of operation.
What I did is I put a sticky pad on it (without a sensor), then turned the lamp on. After 15 minutes, when the temp was stable at a very high temp, I touched the lamp at a place near the pad (but not at the pad), and couldn't hold my finger there for more than 0.5 seconds. When I touched the pad instead (and I pushed my finger to the pad with great force, for good thermal conductivity), I could keep my finger on the pad for almost 10 full seconds.
So, how much of a degree difference do you think we are talking here ?.
What I did is I put a sticky pad on it (without a sensor), then turned the lamp on. After 15 minutes, when the temp was stable at a very high temp, I touched the lamp at a place near the pad (but not at the pad), and couldn't hold my finger there for more than 0.5 seconds. When I touched the pad instead (and I pushed my finger to the pad with great force, for good thermal conductivity), I could keep my finger on the pad for almost 10 full seconds.
So, how much of a degree difference do you think we are talking here ?.