There's usually a bit of back-and-forth discussion between the SPCR staff before an article goes live, mostly to catch silly typo's and grammatical errors, but also to discuss the content of the article. This quote is from one of those emails regarding the Ninja Cu that I think bears being repeated here:

X-bit labs noted that the Copper Ninja didn't have any sort of thermal interface between the heat pipes and the base (http://www.xbitlabs.com/articles/coolers/display/scythe-ninja-copper_2.html#sect0). I wonder if your original Ninja is soldered or otherwise thermally coupled, while the rev. b and copper are not?

I think you're right on the whole clips vs. bolt through thing. I have a an FX-60 s939 with the old style Ninja. I find that it over time cantilevers itself off the cpu, just slightly. If I leave it too long, the gap between HS base and CPU will increase and the thermal paste will run down ever so slowly, until it leaves the top part completely bare. I was shocked the fist time I took off the and found the base looking like this:



Top part looking burnt with no thermal paste and bottom part looking okay. I was all "Ahh, so that's why the cores were running at more than a 10° C difference!" when I saw that. (this was after one year - original thermal paste was Zalman Super Thermal Grease, ZM-STG1. I used the apply a thin layer to IHS surface-method.)

Too bad that the new copper ninja doesn't have a good a mounting as the rev. B. It would be interesting to see how it performed against a regular Ninja Plus under similar conditions. But great review overall - it's not your fault that the conclusion was that mounting matters more than whether the material it's made from is copper or aluminum!



Review rating: 5/5 smilies!

One question remains: given 10+ samples each of Ninja Cu, Ninja Rev B, and Ninja original, how would their temperature averages and variations compare?

I am using Ninja Rev B because it allows use of the Thermalright LGA775 bolt-through kit.

Great review, thank you very much SPCR! I was considering a Ninja Plus Rev. B for my next build, but now it's out of the question.

A few questions still need answers IMHO, I hope to get these answers from the SPCR staff if possible.
1) Would it be possible to test the original Ninja with push-pins, 478 clip + NinjaCU plate, and Thermalright bolt-through?
2) If your original Ninja sample still performs better than the CU with the SAME attachment method, investigate further WHY?
3) A possible investigation method involves dissecting the NinjaCU and Ninja+B, to check heatpipe / contact quality, etc.
4) I wouldn't dare to ask for dissection of the original Ninja. Instead I am cheeky enough to ask if you ever plan to sell that very sample?

If your investigation on the differences between your original Ninja and the rubbish Ninja+B proves fruitful, then maybe you could ask Scythe for a generous sponsorship in return.

Excellent review that unfortunately revealed some very disturbing facts.

I doubt the difference in performance is due to heatpipe variance alone. In my opinion the design of the copper base does not allow for a consistently good contact between it and the heatpipes. Thermalright's snuggly fit heatpipes (and probably carefully soldered) is a good example of how it should be done.

As for the mounting system there is nothing I can say that hasn`t been discussed already. While the current one is an improvement, why not go with the proven bolt through method? The impact on performance has been shown many times.

Skythe have an excellent design which is compromised by poor decisions. I wish they had addressed the main flaws in the 5th anniversary model, if they did, the thermalright ultra might not be the top performing heatsink today.

Some comments:
On page 6, under conclusion #2, the link to the bolt-thru kit is wrong. Maybe you mean this picture?

Nice to read, lots of comments (a good thing).

But, as I notice in several reviews, sometimes you can be a little more critical. There are cheaper heatsinks with almost as good performance (and better mounting clips) on the market. That could be said a little louder.

But, great review nonetheless.

Excellent review and comments on heatsink performance.

I have a question for the group. I own (a) a Ninja Cu (used in a tower with stock 775 mounting bracket), (b) a Ninja Mini (used in a 2480 with 775 pushpins), and (c) a currently unused Thermalright bolt-through kit (bought and never used because the pushpins seemed sufficient for the Mini).

My question: are the brackets that screw to the heatsink bases interchangeable between the Ninja CU and Ninja Mini? I ask because if so, I will switch the Cu mounting bracket to the Mini, and remove the pushpins from the other bracket and use the bolt-through kit on the Cu.

Would be a nice solution, but it all hinges on compatibility between these brackets -- otherwise I won't bother disassembling everything. So if anyone knows for sure, I'd appreciate the info!

Thanks!

1) no. The 478 clips on the original are permanent -- ie, "native". There's no other way to mount it.
2) n/a
3) maybe... if the orig can be dissected without destroying it
4) probably not

Finally, I don't think the Ninja+ Rev B is a "bad" heatsink. It sells for $40 or less in US/Canada, and $7 on a 775 or am2 Thermalright bolt through kit is cheap. For sheer cooling, many other HS will do better, espeically with higher airflow. But in silent low flow mode, it's still tough to do better.

No, I don't mean that one. You cannot have any part that goes over the base. I see my link didn't work (it was a pop up image). I meant this one: LGA775 olt-Thru-Kit Retail BOX Bolt-thru Board Mounting kit

I think they are the same. I hadn't examined that because the Minja is still mounted on another test platform... but yes, the base of the Minja and the NinjaCU look identical.

So, btw, does the arrangement of heatpipes. I'm sure the Plus rev.B will get the same treatment -- unless they decide to retire it.

Thanks Mike. I will give it a shot and report back.

I still love my Rev.B + bolt-thru kit. Great value as it only cost me $33CDN total (well plus 12% tax ugh).

In a way I'm glad you found the Ninja Rev. B to be so lacking, because it at least proves my disappointing results were not down to anything I did wrong.

I think it's more than just the clips though. I too used Thermalright kit with mine and while there was some improvement initially, it didn't last. After a few days, things deteriorated. My guess is that it's more to do with the general build quality of the Ninja, in particular the thermal interface between the heatpipes and the base. In a horizontal position I think it tends to settle a bit. Vibration from the fan probably helps loosen the parts too.

In the end I was really disappointed with the Ninja. Japanese stuff is normally very good, and my experience made me doubt SPCR's testing methods. My faith is now restored I think.

Well these conclusions weren't any surprise for us forum readers.

I have a TRUE for a reason. Also I have the "old" Scythe S-Flex fans and I still believe they are best or amongst the best.
Thank you for recommending them to me forum members.

I was warned early (april 2007) that 9 blades fans weren't that good (hype).
So don't be surprised if that is true.
Bluefronts recent test still show that the Scythe S-Flex are best (or at least very good).
Also note his conclusions about Noctuas new 9 blade fans.
Noctua NF-P12 ... vs ten other fans. Done.

Remember older proven designs are sometimes better than new designs.
I would say there is a 50% chance in general.

i think, these 2 pictures are actually the same item, aren't they?? the picture on the box seems to just have the top part removed.. click

there seems to be a mistake in this review, how is it possible to make noise measurement below ambient?

If you're talking about the Scythe fan at low voltage measuring 17 dBA... Our ambient level in the lab varies (mostly on neighborhood activities/conditions), sometimes we forget to change the 18 dBA that's part out our usual test conditions spiel. At 1m, the Slipstream 800, powered at 7V, registered no change in the 17 dBA ambient when it was turned off/on. Put the meter to 1 ft, and yes, it would register. It's a very quiet fan.

i guess it could be another case of subtracting out ambient noise from the measurement? this is not such an absurd thing to do; for the heatsink cooling performance you already do a similar thing by reporting the temperature as a degrees C rise over ambient temperature instead of an absolute temperature. you can do the same for noise measurement, but its not a 'usual' subtraction because it has to be done in the log domain..

for example, you have ambient noise of 17 dBA, and you are finding that measuring the fan at 12V causes an increase in noise level to 18dBA, i.e. a 1dBA rise over ambient. manufacturer/marketing specs say at 12V their fan spins 800rpm at noise level of 10.7dBA. so fan noise plus ambient noise would add up like so :

total = 10 * log( 10^(10.7/10) + 10^(17/10) )
= 17.9 dBA

i.e. pretty much agreeing with the 18dBA measured here. i touched on this in an earlier thread once.. i realise its technically difficult to quantitatively measure noise levels that are really close to ambient. but it's not really valid to say 'this fan is M dBA loud', by direct measurement, unless 'M' is at least something like ~6dBA above ambient (the 'delta' that the ambient level tends to add to the total measurement decreases as the measured object is louder, this ~6dBA above ambient is about where the contribution of ambient itself tends to have added only about 1dBA to the total measurement)

by the way, here is a python func which subtracts out ambient from the total measured level and also prints the aforementioned 'delta'



f(18.0)=11.1, FWIW

I know this is kind of trivial, because we can simply analyze SPCR labs result, but there are ways to measure the pressure applied by the mounting systems. You can buy pressure sensitive paper, it's very thin. The paper changes color with applied pressure, similar to the change litmus paper undergoes. If we had solid numbers on applied pressure, it might be interesting? It would probably take too much time though.

I definitely think the main differences here can be attributed to the heat pipes making contact with the base. If you solve this issue for the Ninja, the article will be very well linked around the web since it is such a great heatsink. It might be time to break out the soldering gun.

Thanks for the review.

I was able to get my Cu at $50 USD on a special pricing sale at one of our sponsors. I still kind of wish I had went with the original Ninja. I know I still haven't posted my build pictures (life has thrown three curve balls and a screwball at me this week), and my PC hasn't left my parents dining room since the build last Saturday. I actually tossed the fan that came with the Ninja on the rear of my Solo and zip-tie suspended the three speed antec fan under my optical drive (I keep it at the lowest setting).

Right know I have some copper braid slipped under the copper pipes and wrapped around the cross-bar in a primitive pulley style support until I can fabricate a bracket for it that hangs from the cross-bar. If anyone has any other suggestions for beating the weight issue, I'd love to hear them.

I bought a Ninja Rev.B on the strength of your original Ninja article and never got decent results out of it. It does seem to me that the mounting mechanism for both AMD and Intel platforms is suspect.

I think bolt through mounting methods are much better, but obviously they are less convenient to fit.

IMHO....the problem with high temps is a connection problem between the base and the CPU, and not a connection problem between the pipes and the base.

I base that assumption on three recent Ninja (al versions) installations, all using relatively hot P4s. Close examination of the plastic stock retention brackets, all of which used plastic push-pins, showed upward movement(stretching), when pressure from the Ninja was applied. To over-come the stretching, I substituted simple metal screws and metal back-plates for the push-pins. The stock brackets were used in all the installs.

So.....Using a completely stock Ninja without any lapping whatever in the last two projects, I am able to keep a 95W CPU well under 50C maximum, with a minimum of airflow, and no fan on the Ninja. Were there any problems with the heat-pipes themselves, they surely would have shown up in these last few projects.

I too am disappointed with the test results from the copper Ninja..... and have no explanation other than sample variation, and connection difficulty.

I think mounting systems in general will be the next big innovation in heatsinks. With very large and heavy heatsinks, no-one has yet found a perfect way of mounting them.

Even the current best performer, the Thermalright Ultra-120 Extreme, doesn't have a perfect mounting system. People who lapped their CPUs (such as myself) can attest that on dual/quad core systems the physically higher up core is always significantly hotter. Putting a finger on the heatsink or running the system on it's side evens out the pressure, equalising temperatures.

If your CPU cooler is loose enough that you can move it easily you need to increase the contact pressure more. Lapping the CPU and cooler base removes some metal so you often need to add an extra shim to compensate.

The Ultra 120 Extreme isn't perfect, but with a shim and Thermalright IFX-10 backplate it keeps the CPU core temperatures matched on my E8500 CPU.


E8500 Core Temperatures with Ultra 120 Extreme (temps aren't stuck)



@MikeC: With the Scythe Ninja CU metal clips in the review did you try bending them to increase pressure?

.

WR304, the cooler isn't loose at all. I use a 5円 coin as a shim which did help. I tightened the screws a lot... maybe some motherboard bend more than others? Mine is an Asus.

If you've got the springs tight it may be that the cooler backplate is bending.

The Thermalright IFX-10 backplate is a lot more solid than the original Ultra 120 Extreme one. The design is different too so you end up tightening the screws down further.

http://www.thermalright.com/new_a_page/ ... fx-10.html

I'm not sure the extra IFX-10 heatpipe cooler makes any difference but the backplate is decent.

You can use the IFX-10 with the Scythe Ninja Plus rev B also. It might be interesting to see if the test results improve using one of those.