Informal test results CPU 350 vs Apogee GTZ

Quote:

---

Originally Posted by twwen2

Perhaps because the 350 has been shown to be one of the best on i7? Just a thought...

---

Also in the stickies theres a review of them versus the GTZ. http://www.xtremesystems.org/forums/...d.php?t=198681 Albeit with a C2Q.

:rofl: mcoffey gave it a 5 dead horses award! Oh man, it's like very thread I click on is hilarious lately. :ROTF:

Quote:

---

Originally Posted by nikhsub1

Swiftech's pump is uber powerful, think RD-30 or so. I am assuming with the flow rates Gabe posted that his pump was MAXED out. You do realize that to get 50% more flow you would need 2.5 RD-30's? Let me ask you, who will run that much pump? What purpose would that much flow serve? If anything feed the blocks about 50% LESS and see the results.

Now onto some meaningful commentary, I have 100% confidence in skinnee... I know he will do a great job, I do believe he will more than fill Martin's shoes. And ranker, thanks for the kind words.

---

To be honest I didn't realize that. Then I went over and read Skinee's review of the Koolance top (which reviewed well) and the flow rates at max speed....

Honestly it looks like I was completely wrong in my assumption here and that Gabe was actually running at a higher pressure than the vast majority of users in the field would ever see.

Frankly I'm a bit embarassed but I'd rather learn the hard way than not learn at all :)

Quote:

---

Originally Posted by santiagodraco

To be honest I didn't realize that. Then I went over and read Skinee's review of the Koolance top (which reviewed well) and the flow rates at max speed....

Honestly it looks like I was completely wrong in my assumption here and that Gabe was actually running at a higher pressure than the vast majority of users in the field would ever see.

Frankly I'm a bit embarassed but I'd rather learn the hard way than not learn at all :)

---

I have the utmost respect for people who can admit when they are wrong, learn from it and move on. I do it all the time. I've been wrong and when I am, I own up to it. Your clout with me just went up a whole lot. :up:

Interesting, I was reading another thread and stumbled on something:

http://www.xtremesystems.org/forums/...d.php?t=201501

Hellcamino seems to be passionate about his dislike for Swiftech. I now see why he's so adamant about proving Gabe/Swiftech wrong.

Quote:

---

Originally Posted by ranker

Interesting, I was reading another thread and stumbled on something:

http://www.xtremesystems.org/forums/...d.php?t=201501

Hellcamino seems to be passionate about his dislike for Swiftech. I now see why he's so adamant about proving Gabe/Swiftech wrong.

---

Show me where in this thread where I have tried to say Gabe was wrong please. Yes I have certainly had a bad experience with Gabe but I saw no reason to dig it up here. I just don't think much of the test in question.

Let me say this one more time for the cheap seats....What I personally like to see is real world tests only. The reason behind that is simple...you get to see real world results.

Quote:

---

Originally Posted by mcoffey

He he, I almost forgot that one....and a highly coveted award too I might add.:rofl::rofl: I learned alot during that round of testing. Primarily that 1.6 c spread between the blocks doesn't amount to squat:D I busted my damn ass over 1.6 c:rofl: Like that was going to be some revolutionary thing. That's when the light bulb sort of came on for me.

You gotta love it man.

andyc

---

Lemme tell 'ya. I've kinda thought about what this testing entails. For one thing before you start you'd have to get a box full of TIM...no joke. Then, you have to start thinking about the variables and how to isolate all of them. Then after that buy the equipment (which ain't cheap) and set it all up and get it working and checked. The after that you'd have to do "test tests" to make sure the tests made sense...in other words "trials". Then after that you have to do all of those blocks with multiple mounts (so carefully) which would be a marathon session all the while watching ambients and trying to control variables while juggling the computer and making sure the test is going like it's supposed to and the software is working. Then, after all of the tests are done, go back and recheck the equipment and make sure something did not skew the results. Then compile all the data, look it over, and start writing, proof read, and post it.

Whew! I know I didn't cover half of it either. It ain't that simple with stuff like this until you actually do it. I've done many tests before on rocket motors, and it is by no means easy. Any kind of testing like this takes hours, days and sometimes weeks to do.

Heck I spent months one time just doing small scale tests for a large rocket firing. The actual firing was awesome, but it lasted 10 seconds. Oh it was a rush, but you can pour work into testing stuff.

It looks easy when these reviews and tests come out, but it ain't easy at all. That doesn't begin to fully describe it I know.

Quote:

---

Originally Posted by hellcamino

Let me say this one more time for the cheap seats....What I personally like to see is real world tests only. The reason behind that is simple...you get to see real world results.

---

And here lies my disagreement with you. It is always best to ISOLATE the part that you are testing, remove as many variables as possible. You say just the opposite, you would like testing done with all the variables in the world. This makes for bad data. Ever heard of the word 'GIGO'?

Quote:

---

Originally Posted by nikhsub1

And here lies my disagreement with you. It is always best to ISOLATE the part that you are testing, remove as many variables as possible. You say just the opposite, you would like testing done with all the variables in the world. This makes for bad data. Ever heard of the word 'GIGO'?

---

Yeah I'm familiar with the term garbage in garbage out.

I really do understand the idea behind isolated testing but I feel end users would be better served by real world testing. As it sounds that Gabe has a very elaborate lab it just doesn't sound too difficult to me to use standard water cooling gear for at least a secondary test. I would think it would even be in Gabe's interest to do so!

By standard gear in Gabe's case I mean using a mcp355, mcres, GTZ and an MCR320. I would think it would be a real selling point for him (and useful info for everyone) to run a test (in this case) with both blocks then add an MCW30 to the nb and run it again. I am betting that it would just about be a toss-up on the first test and the GTZ would slaughter the koolance block in the second. As many people like to run cpu-nb loops I would think that would be one hell of a demonstrative test in his favor and untainted.

Would that not make for a solid test for you as well?

Hellcamino.... you made your point about wanting more tests. Thats fine. Gabe has eyes and he'll read them and maybe he might. Maybe he won't. The tests done are exactly what some of us wanted, so that we isolated the block. Its one thing to say you want more tests, its another to discredit the ones done. Any more *attempting* discrediting and we'll consider it trolling.

Been uber busy all week working on major project, and I am shocked to see this thread reached 7 pages already.

Still working hard, and too busy to respond individually, it would take hours! Here is some additional info.

Stephen told me that he ran the chiller pump on full because he noticed how restrictive the 350 was, and he wanted to keep his measurements out of the asymptotic area of the curve (0-0.5 GPM - see for example http://www.swiftech.com/assets/image...e-TR_VS_FR.PNG). He likes to take his temp measurements in the "flat" area.

For comparative testing, it is not practical to try to reproduce a user environment with the chiller, because the PQ curve of our pump is too different from that of commonly used pumps (655, 350's etc). Knowing that we need to eliminate the single biggest variable by keeping the coolant temp constant, one needs to understand that if we wanted to mimic a user setup fairly in a comparative test, we'd have to measure the flow with an arbitrarily chosen loop setup, then disassemble the block, and match the flow on the chiller, then do that again for another block, then adjust the chiller pump again, etc.. it's simply too time consuming.

Absolutely. I'm gonna be one pissed off individual if this forum gets overrun by garbage tests, and the people that do real scientiffic method testing stop posting these tests.

Tests with all kinds of variables thrown in there are meaningless. For the folks who do these tests, please continue, and ignore the naysayers. Don't let them win and destroy this forum. I've seen some of the most talented folks just get to the point of not caring anymore, and I get furious when I see that. What can we do to stop this? That information is too valuable to let it slip away because of naysayers and trolling. Name it.

I don't wanna see these guys that have worked their tails off to build these labs and that do these tests get ran off by this trolling. Let me know what I can do...just name it, and I'll do it.

Quote:

---

Originally Posted by nikhsub1

I think you don't understand. They have an industrial pump (similar in specs to an Iwaki MD-30Z IIRC) that they can vary the flowrate with valves... the water is kept at a constant temp. Your comments are irrelevant.

Gabe it would be interesting to see the GTZ at 1.33GPM as well... I realize the 350 is way more restrictive but still.

---

see http://www.xtremesystems.org/forums/...&postcount=162

Quote:

---

Originally Posted by Pellepel

How much worst? That Is what I’m curious about? ~1.5C temp difference at whatever Gabe’s flow rate was set at the pump. I want to know how much difference would it make with D5 or DDC 3.2’s pumping power. Therefore I can base my judgment whether the temp difference is negligible or not. Like Webby mentioned we won’t know until it is tested.

---

see http://www.xtremesystems.org/forums/...&postcount=162 and take a look at the example of curve you will understand quickly!

Quote:

---

Originally Posted by santiagodraco

It has been a bit entertaining and frankly I'm probably pushing too hard. I tend to get overly passionate about things like this and I should probably back off a bit :)

So, in the interest of fairness and all that, I have a suggestion that I think might resolve the issue and make Gabe's tests more valid.

I suggest that the tests be conducted like this. First, take the GTZ block and install and test as Gabe mentioned. The key here is to find the BEST temperature possible with the same load and same water temperature by only varying the water flow. In other words change the water flow over time, pausing at various steps and observing the resulting temperature, until you find the absolute lowest you can get.

Next repeat the test for the CPU-350. Keep all test conditions the same except vary the flow rate until you obtain the absolute lowest temperature for the Koolance block.

If the theory is right you might have a best low temp for each block at varying speeds/flow for the pump. If true it might indicate that each blocks "sweet spot" is different, so choosing one over the other may depend on the type of pump, or flow rate, you plan on using.

Of course we may see that one block is always lower across all flow rates. If true then great, we have a king. But I think it's too early to tell that.

Sound reasonable?

---

please read: http://www.xtremesystems.org/forums/...&postcount=162 it explains how difficult what you are asking for is.

Quote:

---

Originally Posted by eXa

Your point is valid, your math is not. What if you reduce the gpm on the gtz to 1 ;) , cant just shave of 1gpm. The flow rate would probably be just above 1gpm for the 350.

---

if you set the flow rate of GTZ to 1GPM, and use the chiller's pump at same setting to test the 350, flow will be higher in the 350 due to chiller's pump PQ. please read: http://www.xtremesystems.org/forums/...&postcount=162 it explains how time consumming it would be to do that fairly in the context of regular W/C pumps.

too bad you are in MN Skinnee, We'd be delighted to have you visit our lab.

Thanks gabe! I will take you up on that lab visit when I am out in CA. I have a feeling that before this time next year I will over on the west coast for a few days.

Quote:

---

Originally Posted by gabe

see http://www.xtremesystems.org/forums/...&postcount=162

---

Well, then eliminate the chiller all together and use a radiator. I'm sure your lab has a good constant room temperature.

lols, it never ends, huh?

Easy killers!
In this test the only variable is the CPU block.
Restriction difference for just about any type of pump used
should be proportional.

Let me ask the question a different way (and I may be misunderstanding Gabe's reply (the one he's linked).

It seems that by using the chiller certain testing peculiarities are introduced that would make the use of a "normal" pump inappropriate in some way (correct Gabe?). If that is the case, and the use of the chiller is to eliminate the variable of room temperature (another assumption on my part) then if we assume the room temperature is very constant (maybe not) wouldn't the use of a radiator eliminate that and allow the use of a readily available pump?

I guess what I'm getting at is the good old real world test scenario. All roads seem to lead back to the beginning :)

But let me say this. I'm not saying Gabe's tests don't have merit (and yes I am very aware of what the scientific method is and variables) but... I just can't seem to get it out of my mind that the lab you use Gabe seems to introduce it's own set of variables that might not be representative. Maybe I'm completely off base, and I can't quantify them well, it's just a feeling (and yeah feelings aren't very scientific :) )

Any sort of extended testing will dump heat into the room, which will slowly raise the ambient temperature. It's almost impossible to avoid this - if you look at some of Martin's testing outlines, you'll notice that he actually used an AC unit to keep ambients steady, but even his testing had an ambient range of several degrees, courtesy of the AC kicking on and off. Really, what a chiller does is exactly what the AC would do under ideal circumstances, and the chiller can do it without having to worry about variables such as air circulation in the room, or what path the radiator exhaust takes.

I would second Niksub1's suggestion (originally Cathar's idea) in plotting c/w relative to "Hydraulic Pumping Power". This would be a easy way to completely eliminate the pumping power variable as you would clearly see the relative performance differences between both blocks and they wouldn't be a great representation of performance. C/w vs Flow rate skews the results to make more restrictive blocks appear better, however if you use hydraulic horsepower it's comparing a straight thermal performance relative to pumping power.

Thanks for testing, very nice:up: