Tested MY Spiral finally.... - Page 2

bigben2k · 08-13-2002, 09:42 AM

So... are you going to come up with a new/revised Radiate?

pHaestus · 08-13-2002, 09:54 AM

Nope. I have a feeling that the whole concept of "heat watt calculators" are missing the boat. Ever had Prime95 crash due to aggressive memory timing? I have. The power that your CPU is capable of producing under 100% software load is influenced by other components. So if you set a CPU to say a low multiplier and a really high FSB, you may change the heat load under programs like Prime95 by changing memory bandwidth. Interesting questions, but difficult to test/prove. Here are some numbers from my diode reader while idling via WPCredit CPUHLT:

(0.125C res and 4Hz)

See that there's still a regular oscillation in CPU temp? No such thing as 100% idle. Not so sure there is 100% load either. Such is the challenge of having monitoring equipment that is decent (but still perhaps not accurate).

webmedic · 08-13-2002, 10:00 AM

Ph what are using to monitor your temps? I've been looking for a digital thermometer. I've been looking at the fluke 2190a. Any other recomendations?

bigben2k · 08-13-2002, 10:13 AM

Quote:

Originally posted by pHaestus
Nope. I have a feeling that the whole concept of "heat watt calculators" are missing the boat. Ever had Prime95 crash due to aggressive memory timing? I have. The power that your CPU is capable of producing under 100% software load is influenced by other components. So if you set a CPU to say a low multiplier and a really high FSB, you may change the heat load under programs like Prime95 by changing memory bandwidth. Interesting questions, but difficult to test/prove. Here are some numbers from my diode reader while idling via WPCredit CPUHLT:

See that there's still a regular oscillation in CPU temp? No such thing as 100% idle. Not so sure there is 100% load either. Such is the challenge of having monitoring equipment that is decent (but still perhaps not accurate).

Hum... Did you consider that it may be a full load anyways? I mean, maybe that's how the CPU heat actually comes! Also, I seem to remember that the internal diode is capable of reading temp changes up to 40 deg C per second. It might be interesting to run this test at a higher frequency (than 4 Hz), but I'm not sure that the diode reader can handle it.

What's "External 1" and "External 2"

webmedic · 08-13-2002, 10:28 AM

On http://www.benchtest.com/calc.html he talkes aboutthis verry thing. He also reconds using a progy k7burn. The author of witch says it will only get about 88% cpu usage. From what I've heard this one of the most loading programs there is. I personaly have been using toast because I get about c2 higher with it than prime 95.

pHaestus · 08-13-2002, 10:28 AM

Those numbers are from a Maxim 6655 evaluation system. It came with monitoring software (the reason I purchaed it), a pcb with a 6655 on it and a second one that performs smbus-> parallel. The software won't work without a parallel interface unfortunately.

As for temp monitoring equipment; go with the best accuracy that you can afford. The Fluke 2190 seems to be a good bargain, but I have no experience about the cost of type T tcs or other costs of ownership.

I use YSI 700 series linear thermistors (0.01C res and 0.1-0.2C accuracy) on D5180 readers (have 6 probes and two thermometers). No software monitoring/serial out on these; they are ancient. I have one spare and another I am bidding on just for parts.

If you want better than 0.1C accuracy then things get really expensive...

webmedic · 08-13-2002, 10:32 AM

Thats why I was looking at the fluke. It's not to bad and I cah get decentthermistors for it. So I can get to .01c. To note it would be nice to have one that prints out the temps or logs them to a windows box.

Read my post right above your last one. I must have posted it at the same time as your last one.

pHaestus · 08-13-2002, 10:32 AM

Sorry. External1 is the AMD diode and external2 is a diode placed 1" above the center of the heatsink's intake fan. internal is the local temp of the 6655 IC.

That is the essence of the problem, Bigben. The definition of "full load" is dependent not only upon the program used, but often the amount of RAM, RAM timings, and fsb. So how can I possibly make a calculator?

Yes CPU burn is a good program. I need to get my testbed back up and running. I was planning on monitoring the CPU temps remotely with the 6655 via a notebook PC and running the test system by booting to DOS and running CPUBurn in real time

pHaestus · 08-13-2002, 10:34 AM

Quote:

Originally posted by webmedic
Thats why I was looking at the fluke. It's not to bad and I cah get decentthermistors for it. So I can get to .01c. To note it would be nice to have one that prints out the temps or logs them to a windows box.

Read my post right above your last one. I must have posted it at the same time as your last one.

Accuracy and resolution aren't the same thing. I think the Fluke is 0.2C accuracy with type Ts. Those will provide you a fair amount of flexibility though. Talk to JoeC at o/cers; he uses type ts for his setup.

webmedic · 08-13-2002, 10:42 AM

Sorry I know the difference between accuracy and resolution. Just used the wrong words. The resolution is good. The accuracy may be lacking. If this is the case I've seen some newer all in one little handheld units that are just as acurate. I think it was resolution of .01c and acuracy of 1%. I'm kind of leary though. Thats why I'm asking arround..

Cova · 08-13-2002, 11:39 AM

pH - you seem to be on webmedics case a lot lately (mostly in regards to his plans for a WB roundup) in regards to accuracy and resolution, and the fact that he needs better temp monitoring equipment. And now you post up a graph for us all of your CPU temp jumping around all over but within a range of about 1.4C and you mention that you have .125C resolution but fail to mention that your sensor is only +/- 3C accurate (stats on chip taken from page 2 of Maxim's pdf on it at http://pdfserv.maxim-ic.com/arpdf/MAX6655-MAX6656.pdf )

Now maybe this should be in that other thread - but I'm putting it here in response to your graph. You now are the one making conclusions that is pushing the limits of your measurement equipment, and yet you continue to tell webmedic not to do the same thing.

Webmedic - go ahead and do your roundup with what equipment you can afford - don't bankrupt yourself on it trying to please just a couple people. Everyone needs to realize that just because a sensor is accurate to +/- 3C does NOT mean that it will always be off by 3C - with a few test runs it should average to be very close to accurate. I for one at least really look forward to reading your article.

pHaestus · 08-13-2002, 12:25 PM

Accuracy of 3C over the FULL temperature range; the accuracy degrades a good bit at lower temperatures and in the 50-60C range I use this to test with then it is pretty close to 1C. It is also possible to calibrate the diode reader and setup and account for this as the accuracy reflects the variability of testing setups (electrical noise, capacitance, and variability in the diodes used) as well as the intrinsic accuracy of the IC; I am still working on that however. The problem is that you have to get a perfect joint at the CPU socket so that the calibration done by putting CPU with diode reader on pins into water is relevant to the diode reader soldered on the motherboard. Maybe you can assume that both joints are equal? Comments are welcome on this.

I believe if you look around I calculated some uncertainties based upon +/- 1C for a diode reader. That is reasonable at the temps I typically run (50-60C thus far with heatsinks).

The CPU numbers jump around because that is what CPU temps DO; the programs run cycles and if your equipment is sufficiently good then you pick therm up. The only way to deal with that is to take a LOT of measurements, average, and post a std deviation along with the number:

And no, those numbers aren't good enough to quantitatively rate waterblocks. My std deviations are in the 0.4C range. You really need a die simulator with exact power inputs and close to perfect insulation if you want better; a CPU won't do.

I have been critical to webmedic at times because I have done exactly what he is headed towards and I already know that he won't get the results he wants from the testing equipment that he has. I already attempted the same! Just trying to save him some frustration. I don't think I ever told him not to proceed.

In my opinion, the people telling him to go ahead and do all that work with shitty equipment are the ones doing him a disservice; not the ones offering advice on what is needed.

Les · 08-13-2002, 01:17 PM

Quote:

Originally posted by pHaestus
.......... It is also possible to calibrate the diode reader and setup and account for this as the accuracy reflects the variability of testing setups (electrical noise, capacitance, and variability in the diodes used) as well as the intrinsic accuracy of the IC; I am still working on that however. The problem is that you have to get a perfect joint at the CPU socket so that the calibration done by putting CPU with diode reader on pins into water is relevant to the diode reader soldered on the motherboard. ....................

I am still happy with my calibration method for an "AMD cpu diode".
I read/monitor the "AMD cpu Diode" remotely through 2nd PC's SMBus .I calibrate against the temp of the water flowing through a waterblock attached to the unpowered "AMD cpu".
First reported here:
http://forum.oc-forums.com/vb/showth...0&pagenumber=3

bigben2k · 08-13-2002, 01:54 PM

Kewl... I was about to ask if the diode goes dead if the CPU has no power.

So for calibrating it like that, I'd submit the CPU to various temps, wait a fixed amount of time for the temp to propagate through the CPU's... and what? Use a live mobo, with a live CPU, with the reader hooked up to the other CPU that I'm subjecting to various temps?

maskedgeek · 08-13-2002, 03:40 PM

Quote:

Originally posted by Fixittt
Ok who now wants to argue that copper is better then aluminum? Like I have always said, a well designed block in aluminum can preform just as good as a copper block.

Any takers?

Good job jay!

B!T@H`N

i agree that a alu block can do as good as a copper block! i did testing a little while back with my popcan blocks, i made a alu and a cu based one same sizes and all, the cu was .01c/w better than the alu block...

thats a nice block there jaydee one thing to make it work better is if you would fill out all the possible space of the metal with the paths itll perform better...

morphling1 · 08-13-2002, 04:22 PM

pHaestus aren't you useing 6657 chip for reading and the one you send me 6659 both have accuracy of +-1°C with 0.125°C resolution so not exactly +-3°C , also while we're discussing temperature reading (sorry JD to do it here), can I put another sensor on smbus, I was thinking of multichanell MAX1668 so I could read 4 more temps, will mbm recognized both, max6659 and max1668 remote sensors if they are soldered to the same dimm (does it even make a difference if I solder to separate dimms).
Btw my reader work perfectly, it corelate very accurate to cpu frequency change.

jaydee · 08-13-2002, 07:37 PM

I just got home for work so sorry for the late reply. The temp monitoring is close enough. I could care less if it is 1C+/- myself. Remeber we are talking about roughly 30gph to roughly 60gph. I really wasn't expecting any change.

Also for a small update I am running (since 4pm yesterday)2.0Vcore 1523mhz at 145fsb on the XP1600+ and running 41C load, 30C water, 26C room. I am very happy with it for its size.

Also I must point out that Fixittt is not saying two identical blocks in copper and al will have the same results. he is saying a well designed al block will perform as well as other not so well designed copper blocks. At least that is how I read it. I made a couple identical ones (thank god for the CNC) and they where about 1C different. yeah yreah yeah that is with my half assed temp monitoring.

But I think the results would change with different flow rates ect.. As I am sure they both react different to different flow rates and what not. In any event AL does a plenty fine job. I also have a lead on a good anodizing outfit localy. Will be looking into it later this year.

Anyway carry on with the other topic if you wish. I am pretty much done with this block anyway.

08-13-2002, 09:42 AM	#26
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	So... are you going to come up with a new/revised Radiate? __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

08-13-2002, 10:00 AM	#28
webmedic Cooling Savant Join Date: Dec 2001 Location: Cheney, Wa Posts: 367	Ph what are using to monitor your temps? I've been looking for a digital thermometer. I've been looking at the fluke 2190a. Any other recomendations? __________________ www.water-cool.com

08-13-2002, 10:28 AM	#30
webmedic Cooling Savant Join Date: Dec 2001 Location: Cheney, Wa Posts: 367	On http://www.benchtest.com/calc.html he talkes aboutthis verry thing. He also reconds using a progy k7burn. The author of witch says it will only get about 88% cpu usage. From what I've heard this one of the most loading programs there is. I personaly have been using toast because I get about c2 higher with it than prime 95. __________________ www.water-cool.com

08-13-2002, 10:32 AM	#32
webmedic Cooling Savant Join Date: Dec 2001 Location: Cheney, Wa Posts: 367	Thats why I was looking at the fluke. It's not to bad and I cah get decentthermistors for it. So I can get to .01c. To note it would be nice to have one that prints out the temps or logs them to a windows box. Read my post right above your last one. I must have posted it at the same time as your last one. __________________ www.water-cool.com

08-13-2002, 10:42 AM	#35
webmedic Cooling Savant Join Date: Dec 2001 Location: Cheney, Wa Posts: 367	Sorry I know the difference between accuracy and resolution. Just used the wrong words. The resolution is good. The accuracy may be lacking. If this is the case I've seen some newer all in one little handheld units that are just as acurate. I think it was resolution of .01c and acuracy of 1%. I'm kind of leary though. Thats why I'm asking arround.. __________________ www.water-cool.com

08-13-2002, 09:54 AM	#27
pHaestus Big Player Making Big Money Join Date: Aug 2001 Location: irc.lostgeek.com #procooling.com Posts: 4,782	Nope. I have a feeling that the whole concept of "heat watt calculators" are missing the boat. Ever had Prime95 crash due to aggressive memory timing? I have. The power that your CPU is capable of producing under 100% software load is influenced by other components. So if you set a CPU to say a low multiplier and a really high FSB, you may change the heat load under programs like Prime95 by changing memory bandwidth. Interesting questions, but difficult to test/prove. Here are some numbers from my diode reader while idling via WPCredit CPUHLT: (0.125C res and 4Hz) See that there's still a regular oscillation in CPU temp? No such thing as 100% idle. Not so sure there is 100% load either. Such is the challenge of having monitoring equipment that is decent (but still perhaps not accurate).

08-13-2002, 11:39 AM	#36
Cova Cooling Savant Join Date: May 2002 Location: Canada Posts: 247	pH - you seem to be on webmedics case a lot lately (mostly in regards to his plans for a WB roundup) in regards to accuracy and resolution, and the fact that he needs better temp monitoring equipment. And now you post up a graph for us all of your CPU temp jumping around all over but within a range of about 1.4C and you mention that you have .125C resolution but fail to mention that your sensor is only +/- 3C accurate (stats on chip taken from page 2 of Maxim's pdf on it at http://pdfserv.maxim-ic.com/arpdf/MAX6655-MAX6656.pdf ) Now maybe this should be in that other thread - but I'm putting it here in response to your graph. You now are the one making conclusions that is pushing the limits of your measurement equipment, and yet you continue to tell webmedic not to do the same thing. Webmedic - go ahead and do your roundup with what equipment you can afford - don't bankrupt yourself on it trying to please just a couple people. Everyone needs to realize that just because a sensor is accurate to +/- 3C does NOT mean that it will always be off by 3C - with a few test runs it should average to be very close to accurate. I for one at least really look forward to reading your article.

08-13-2002, 12:25 PM	#37
pHaestus Big Player Making Big Money Join Date: Aug 2001 Location: irc.lostgeek.com #procooling.com Posts: 4,782	Accuracy of 3C over the FULL temperature range; the accuracy degrades a good bit at lower temperatures and in the 50-60C range I use this to test with then it is pretty close to 1C. It is also possible to calibrate the diode reader and setup and account for this as the accuracy reflects the variability of testing setups (electrical noise, capacitance, and variability in the diodes used) as well as the intrinsic accuracy of the IC; I am still working on that however. The problem is that you have to get a perfect joint at the CPU socket so that the calibration done by putting CPU with diode reader on pins into water is relevant to the diode reader soldered on the motherboard. Maybe you can assume that both joints are equal? Comments are welcome on this. I believe if you look around I calculated some uncertainties based upon +/- 1C for a diode reader. That is reasonable at the temps I typically run (50-60C thus far with heatsinks). The CPU numbers jump around because that is what CPU temps DO; the programs run cycles and if your equipment is sufficiently good then you pick therm up. The only way to deal with that is to take a LOT of measurements, average, and post a std deviation along with the number: And no, those numbers aren't good enough to quantitatively rate waterblocks. My std deviations are in the 0.4C range. You really need a die simulator with exact power inputs and close to perfect insulation if you want better; a CPU won't do. I have been critical to webmedic at times because I have done exactly what he is headed towards and I already know that he won't get the results he wants from the testing equipment that he has. I already attempted the same! Just trying to save him some frustration. I don't think I ever told him not to proceed. In my opinion, the people telling him to go ahead and do all that work with shitty equipment are the ones doing him a disservice; not the ones offering advice on what is needed.

08-13-2002, 01:54 PM	#39
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	Kewl... I was about to ask if the diode goes dead if the CPU has no power. So for calibrating it like that, I'd submit the CPU to various temps, wait a fixed amount of time for the temp to propagate through the CPU's... and what? Use a live mobo, with a live CPU, with the reader hooked up to the other CPU that I'm subjecting to various temps? __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

08-13-2002, 04:22 PM	#41
morphling1 Cooling Savant Join Date: Sep 2001 Location: Slovenia Posts: 468	pHaestus aren't you useing 6657 chip for reading and the one you send me 6659 both have accuracy of +-1°C with 0.125°C resolution so not exactly +-3°C , also while we're discussing temperature reading (sorry JD to do it here), can I put another sensor on smbus, I was thinking of multichanell MAX1668 so I could read 4 more temps, will mbm recognized both, max6659 and max1668 remote sensors if they are soldered to the same dimm (does it even make a difference if I solder to separate dimms). Btw my reader work perfectly, it corelate very accurate to cpu frequency change. __________________ [My ftp, with lots of pics, hope to be home page someday

08-13-2002, 07:37 PM	#42
jaydee Put up or Shut Up Join Date: Dec 2001 Location: Spokane WA Posts: 6,506	I just got home for work so sorry for the late reply. The temp monitoring is close enough. I could care less if it is 1C+/- myself. Remeber we are talking about roughly 30gph to roughly 60gph. I really wasn't expecting any change. Also for a small update I am running (since 4pm yesterday)2.0Vcore 1523mhz at 145fsb on the XP1600+ and running 41C load, 30C water, 26C room. I am very happy with it for its size. Also I must point out that Fixittt is not saying two identical blocks in copper and al will have the same results. he is saying a well designed al block will perform as well as other not so well designed copper blocks. At least that is how I read it. I made a couple identical ones (thank god for the CNC) and they where about 1C different. yeah yreah yeah that is with my half assed temp monitoring. But I think the results would change with different flow rates ect.. As I am sure they both react different to different flow rates and what not. In any event AL does a plenty fine job. I also have a lead on a good anodizing outfit localy. Will be looking into it later this year. Anyway carry on with the other topic if you wish. I am pretty much done with this block anyway.

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)