Ninja Copper: Scythe's 5th Year Celebration
-
- Moderator
- Posts: 7681
- Joined: Mon Dec 09, 2002 7:11 pm
- Location: Maynard, MA, Eaarth
- Contact:
@MikeC: I had a look at an overlooked contender for the 'best heatsink with low airflow', the HR-01 you reviewed here: http://www.silentpcreview.com/article290-page5.html
It would have been nice to include it in the comparison table, since the old review shows it to be 1-4 degrees worse than your precious Ninja sample.
Which means it's about as good as the Ninja CU, but a lot lighter, and cheaper too. Of course, that review was on Socket 478, things may be different on 775 or AM2.
There is a new version of the HR-01 with 6 heatpipes, called the Plus, it would be quite interesting to read a SPCR review on it.
It would have been nice to include it in the comparison table, since the old review shows it to be 1-4 degrees worse than your precious Ninja sample.
Which means it's about as good as the Ninja CU, but a lot lighter, and cheaper too. Of course, that review was on Socket 478, things may be different on 775 or AM2.
There is a new version of the HR-01 with 6 heatpipes, called the Plus, it would be quite interesting to read a SPCR review on it.
-
- Site Admin
- Posts: 12285
- Joined: Sun Aug 11, 2002 3:26 pm
- Location: Vancouver, BC, Canada
- Contact:
You're right, the rec. list should be revised to reflect the Rev 2 performance as the original has not been available for some time.kaange wrote:Given how poorly the Rev 2 Ninja performs vs the original version used by SPCR (even with the bolt through kit), the recommended list of CPU coolers need to be revised as the Ninja rating is now highly misleading.
Was there an obvious fault with the Rev 2 Ninja, like a highly convex base?
No, no obvious flaws, as mentioned a couple times in the review, the bases of all three were flat and very smooth.
-
- Site Admin
- Posts: 12285
- Joined: Sun Aug 11, 2002 3:26 pm
- Location: Vancouver, BC, Canada
- Contact:
yeah, we have a new one to review.Tzupy wrote:@MikeC: I had a look at an overlooked contender for the 'best heatsink with low airflow', the HR-01 you reviewed here: http://www.silentpcreview.com/article290-page5.html
It would have been nice to include it in the comparison table, since the old review shows it to be 1-4 degrees worse than your precious Ninja sample.
Which means it's about as good as the Ninja CU, but a lot lighter, and cheaper too. Of course, that review was on Socket 478, things may be different on 775 or AM2.
There is a new version of the HR-01 with 6 heatpipes, called the Plus, it would be quite interesting to read a SPCR review on it.
I did the Ninja Cu - Mini bracket swap today. They are indeed interchangeable, so I now have the Cu bracket system holding the Mini, and the Thermalright bolt-through kit on the Cu, using the bracket that previously held the 775 pins. Clearly the bolt through is superior, particularly in the ease of installation. I shimmed the springs with a few flat washers as well to account for the Cu's heavier weight compared to the Thermalright heat sinks.
Unfortunately, I didn't take any before/after temperature measurements. This article came out in the middle of my build so I made this change before I could do any comparison.
Unfortunately, I didn't take any before/after temperature measurements. This article came out in the middle of my build so I made this change before I could do any comparison.
Fanless?
The Ninja Cu is known to be the heatsink of choice for PC systems without a dedicated CPU fan. The theory is, that the widely layed out fins better support evacuation of wasted heat by convection.
SPCR, could you elaborate on this? Did you do some testing without a CPU fan attached?
I have an E8400 running under a Ninja Cu without a CPU fan. Below the truly massive copper tower the northbridge and a 8800GT are convecting some additional heat up to the Ninja. A rear-side Noctua NF-S12 at 800rpm and a quite quiet BeQuiet DarkPowerPro PSU are the only air stirring devices within the Lian-Li PC7 case. Unfortunately the readout of the E8400 internal temperature diodes seems to be a little bit fuzzy, but I never see CPU temps above 60 degree celsius. (BTW the loudest noise in my PC is the f*ing whine of the 8800GT under 3D load, and I cannot return it because I already mounted an S1. Sorry for getting off-topic here...)
Best Regards,
Dirk
SPCR, could you elaborate on this? Did you do some testing without a CPU fan attached?
I have an E8400 running under a Ninja Cu without a CPU fan. Below the truly massive copper tower the northbridge and a 8800GT are convecting some additional heat up to the Ninja. A rear-side Noctua NF-S12 at 800rpm and a quite quiet BeQuiet DarkPowerPro PSU are the only air stirring devices within the Lian-Li PC7 case. Unfortunately the readout of the E8400 internal temperature diodes seems to be a little bit fuzzy, but I never see CPU temps above 60 degree celsius. (BTW the loudest noise in my PC is the f*ing whine of the 8800GT under 3D load, and I cannot return it because I already mounted an S1. Sorry for getting off-topic here...)
Best Regards,
Dirk
-
- Moderator
- Posts: 7681
- Joined: Mon Dec 09, 2002 7:11 pm
- Location: Maynard, MA, Eaarth
- Contact:
-
- *Lifetime Patron*
- Posts: 1288
- Joined: Sat Oct 25, 2003 3:21 pm
- Location: 15143, USA
- Contact:
Sigh... It figures that this would come out while my first Ninja was still on the UPS truck. When did HSFs become obsolete faster than CPUs?MikeC wrote:You're right, the rec. list should be revised to reflect the Rev 2 performance...
After examining the new Rev. B, I'm confident that it will work well with a solid mount. The adapter clips are respectably stiff, so replacing the pushpins should be enough.
We'll see.
@MikeC: in order to determine why the Ninja+B is such a poor performer, without dissecting it, you could measure the heatpipe temperatures.
And maybe fin temperatures too, to check for insufficient contact pressure on the heatpipes (they are press-fitted IIRC).
It was reported in the forums (and the Xbitlabs review) that the contact of the heatpipes with the base is ... suboptimal, to put it gently.
Such a manufacturing flaw could explain sample variation between Ninja+B's, maybe some of them are decent.
And maybe fin temperatures too, to check for insufficient contact pressure on the heatpipes (they are press-fitted IIRC).
It was reported in the forums (and the Xbitlabs review) that the contact of the heatpipes with the base is ... suboptimal, to put it gently.
Such a manufacturing flaw could explain sample variation between Ninja+B's, maybe some of them are decent.
-
- Moderator
- Posts: 7681
- Joined: Mon Dec 09, 2002 7:11 pm
- Location: Maynard, MA, Eaarth
- Contact:
Hello,
Here's a proposal: if anyone has a Ninja that they suspect is not up to snuff -- it may be possible to affect the heatpipe/base connection by adding solder. If someone can do this, and if they see a big/noticeable improvement, that could point the way for folks to repair/improve their Ninjas.
This is my working hypothosis, as well. There was a poster (named "andaca") who violently ripped apart his Ninja and posted pictures that showed sub-optimal (to use your word) soldering between some of the heatpipes and the base. (Those pictures are no longer on the site, for reasons to do with the poster's behavior, though I wish the pictures themselves had been salvaged.)Tzupy wrote:It was reported in the forums (and the Xbitlabs review) that the contact of the heatpipes with the base is ... suboptimal, to put it gently.
Such a manufacturing flaw could explain sample variation between Ninja+B's, maybe some of them are decent.
Here's a proposal: if anyone has a Ninja that they suspect is not up to snuff -- it may be possible to affect the heatpipe/base connection by adding solder. If someone can do this, and if they see a big/noticeable improvement, that could point the way for folks to repair/improve their Ninjas.
-
- Site Admin
- Posts: 12285
- Joined: Sun Aug 11, 2002 3:26 pm
- Location: Vancouver, BC, Canada
- Contact:
Having examined the soldering of the heatpipes to the base of the mini-ninja and the 3 ninja samples on hand, I have to report that only the copper version's secondary aluminum heatsink was not soldered to the heatpipes at the base. It fell off when the screws around the perimeter were removed -- on the other three, the secondary aluminum heatsink did not budge.
Screws are off on all 3 heatsink bases. Notice the "fallen" secondary aluminum heatsink on the CU version on right. (Click the image for a super large version ~4000px wide.)
Also, I can't see how anyone can access the base heatpipes to add more solder. The secondary aluminum heatsink is trapped in the tight space between the base and the bottom fin. No soldering iron could get in there. I believe the manufacturing process goes something like this: The base is soldered (welded?) to the 6 u-shaped heatpipes in some kind of clamped jig. 3 in one direction, then the 3 across the top of them. The two sets of heatpipes are directly affixed to each other. Then the secondary heatsink is screwed on, and perhaps the entire base assembly heated up again to make the solder flow to permanently affix the secondary heatsink...???? Only after that are the fins inserted.
I see two possible defects in the poor performing Rev B sample -- there seems to be more solder between the base and the heatpipes directly connected to it. In other words, less of the heatpipe is making contact with the base than in the other ninjas. Maybe. Hard to tell for sure. Also there's a fairly hefty depression (from a bump?) in one of the heatpipes as it exits from the base -- but if this affects anything it would only be one of 12 heatpipe runs.
Bottom line: I think the construction of the Ninja is sufficiently complex, and there are too many variables (which includes the quality consistency of the heatpipes themselves) for users to fix or figure out what to do. About the best that could be done is that if the secondary aluminum heasink does come loose when unscrewed, squirt in some good TIM in there before you screw it up again.
Screws are off on all 3 heatsink bases. Notice the "fallen" secondary aluminum heatsink on the CU version on right. (Click the image for a super large version ~4000px wide.)
Also, I can't see how anyone can access the base heatpipes to add more solder. The secondary aluminum heatsink is trapped in the tight space between the base and the bottom fin. No soldering iron could get in there. I believe the manufacturing process goes something like this: The base is soldered (welded?) to the 6 u-shaped heatpipes in some kind of clamped jig. 3 in one direction, then the 3 across the top of them. The two sets of heatpipes are directly affixed to each other. Then the secondary heatsink is screwed on, and perhaps the entire base assembly heated up again to make the solder flow to permanently affix the secondary heatsink...???? Only after that are the fins inserted.
I see two possible defects in the poor performing Rev B sample -- there seems to be more solder between the base and the heatpipes directly connected to it. In other words, less of the heatpipe is making contact with the base than in the other ninjas. Maybe. Hard to tell for sure. Also there's a fairly hefty depression (from a bump?) in one of the heatpipes as it exits from the base -- but if this affects anything it would only be one of 12 heatpipe runs.
Bottom line: I think the construction of the Ninja is sufficiently complex, and there are too many variables (which includes the quality consistency of the heatpipes themselves) for users to fix or figure out what to do. About the best that could be done is that if the secondary aluminum heasink does come loose when unscrewed, squirt in some good TIM in there before you screw it up again.
MikeC, I think you may have spotted the flaw there.
As you mentioned, there might be more solder between the base and heatpipes in the Rev B. With the Copper, they have just applied pressure to make the contact, probably squashing the heatpipes a bit in the process and thus making more surface area touch the base, as well as evening it out.
It would be interesting to try applying a bit of thermal paste to the heatpipes on the copper, to see if performance could be improved even further.
It seems then that the Rev B has two major flaws - the socket hold down clips which are not tight enough and the poor contact between the base and the heatpipes.
To be fair on the first point, the plastic clips are probably well within the Intel spec. The Thermalright springs are probably right at the limit of it, but even that isn't enough as evidenced by the improvement seen when adding a shim/coin. On that front it's probably as much to do with the CPU heatspreaders being concave as anything - extra pressure flattens them out a bit.
Just you wait, in a month or two someone will start offering pre-lapped Core 2s with lapped and re-welded Ninjas on eBay
As you mentioned, there might be more solder between the base and heatpipes in the Rev B. With the Copper, they have just applied pressure to make the contact, probably squashing the heatpipes a bit in the process and thus making more surface area touch the base, as well as evening it out.
It would be interesting to try applying a bit of thermal paste to the heatpipes on the copper, to see if performance could be improved even further.
It seems then that the Rev B has two major flaws - the socket hold down clips which are not tight enough and the poor contact between the base and the heatpipes.
To be fair on the first point, the plastic clips are probably well within the Intel spec. The Thermalright springs are probably right at the limit of it, but even that isn't enough as evidenced by the improvement seen when adding a shim/coin. On that front it's probably as much to do with the CPU heatspreaders being concave as anything - extra pressure flattens them out a bit.
Just you wait, in a month or two someone will start offering pre-lapped Core 2s with lapped and re-welded Ninjas on eBay
-
- Moderator
- Posts: 7681
- Joined: Mon Dec 09, 2002 7:11 pm
- Location: Maynard, MA, Eaarth
- Contact:
Hello Mike,
Thank you for posting the high rez picture -- there is an awful lot of air between the base plate and the second row of heatpipes on the Rev B! Also, there is a much smaller area of contact between the base and the second row of heatpipes. The original and the Rev Cu ( ) bases appear to be machined (?) -- while the Rev B is pressed.
The three bases are designed differently, and only the original has a decent amount of contact with the second set of heatpipes -- so they are not able to be very useful. When they are in use, are the ends of the two sets of heatpipes different temperatures?
If heat is applied to the center of the base, it will draw the molten solder to it. This is how every copper plumbing connection is made. The solder can be applied at the air spaces next to the heatpipes.
Thank you for posting the high rez picture -- there is an awful lot of air between the base plate and the second row of heatpipes on the Rev B! Also, there is a much smaller area of contact between the base and the second row of heatpipes. The original and the Rev Cu ( ) bases appear to be machined (?) -- while the Rev B is pressed.
The three bases are designed differently, and only the original has a decent amount of contact with the second set of heatpipes -- so they are not able to be very useful. When they are in use, are the ends of the two sets of heatpipes different temperatures?
If heat is applied to the center of the base, it will draw the molten solder to it. This is how every copper plumbing connection is made. The solder can be applied at the air spaces next to the heatpipes.
-
- Site Admin
- Posts: 12285
- Joined: Sun Aug 11, 2002 3:26 pm
- Location: Vancouver, BC, Canada
- Contact:
I'm not convinced the extra air around the 2nd set of heatpipes at the base of Rev B is really a problem. On the others, even if there is less air, there's still some gap -- it can't be a really tight fit, because the secondary heatsink has to be manually fitted over it. So then does it matter if there's 1mm of open air between the heatpipe and the secondary heatsink or 10mm? It's still open air...NeilBlanchard wrote:Hello Mike,
Thank you for posting the high rez picture -- there is an awful lot of air between the base plate and the second row of heatpipes on the Rev B! Also, there is a much smaller area of contact between the base and the second row of heatpipes. The original and the Rev Cu ( ) bases appear to be machined (?) -- while the Rev B is pressed.
The three bases are designed differently, and only the original has a decent amount of contact with the second set of heatpipes -- so they are not able to be very useful. When they are in use, are the ends of the two sets of heatpipes different temperatures?
If heat is applied to the center of the base, it will draw the molten solder to it. This is how every copper plumbing connection is made. The solder can be applied at the air spaces next to the heatpipes.
As for heating up the base to make the solder melt, I thought of that before, but wonder how such heat could be applied without damaging the base in some way. I would be willing to set the rev B sample on an electric stove element on whatever setting to get the whole thing hot enough so threading solder into the gap would make it melt... But it could just turn into a huge mess which would make my honey unhappy... and you wouldn't want that!!
The biggest problem I see in resoldering or brazing the base to the heatpipes is that the amount of heat required could damage the fins + the secondary heatsink.
To heat the base sufficiently for solder to flow and attach properly could cause the soldering of the heatpipes to the fins to flow as well since the heatpipes are designed to transfer heat.
To heat the base sufficiently for solder to flow and attach properly could cause the soldering of the heatpipes to the fins to flow as well since the heatpipes are designed to transfer heat.
-
- Moderator
- Posts: 7681
- Joined: Mon Dec 09, 2002 7:11 pm
- Location: Maynard, MA, Eaarth
- Contact:
Hello,
My other point is the big differences in the contact areas in the path that the heat has to take:
(I've added dimensions & a note)
The original Ninja has 2X the surface area to transfer heat to the second level of heatpipes as the Rev B, while the Rev Cu has just line contact with the second level of heatpipes.
And the solder is needed to transfer heat from the base plate to the heatpipes -- the secondary extruded aluminum HS is a minor player in the proportion of the heat dissipation, I think.
[Edit: Silicon? No, I think that would harm the heat transfer?]
My other point is the big differences in the contact areas in the path that the heat has to take:
(I've added dimensions & a note)
The original Ninja has 2X the surface area to transfer heat to the second level of heatpipes as the Rev B, while the Rev Cu has just line contact with the second level of heatpipes.
And the solder is needed to transfer heat from the base plate to the heatpipes -- the secondary extruded aluminum HS is a minor player in the proportion of the heat dissipation, I think.
[Edit: Silicon? No, I think that would harm the heat transfer?]
Last edited by NeilBlanchard on Tue Mar 04, 2008 4:27 am, edited 1 time in total.
Hmmm.
I wonder if a ghetto mod could be done by running a bead of silicone across each side of the base, except in each direction to seal up the base into an open box, then filling the base with water before sealing it all in with more silicone on last side of the base. Just a thought.....
Would be messy if it leaked, of course.
I wonder if a ghetto mod could be done by running a bead of silicone across each side of the base, except in each direction to seal up the base into an open box, then filling the base with water before sealing it all in with more silicone on last side of the base. Just a thought.....
Would be messy if it leaked, of course.
I read happiness...NeilBlanchard wrote:... while the Rev Cu has just line contact with the second level of heatpipes.
been awake too long.
thanks for the big review and particularly for the comparison between the retention mechanisms.
Q:
with the revB, I'd have to physically break off the pushpins from the 775 retention plate in order to use the bolt-thru-kit ?
i've had the ninja for some time now but not installed, and am getting ready to get a new mb/cpu combo any time soon. good time to get the kit as well.
-
- Moderator
- Posts: 7681
- Joined: Mon Dec 09, 2002 7:11 pm
- Location: Maynard, MA, Eaarth
- Contact:
Hello,
I would think that filling all the voids with solder would be the best bet.
Another possible problem of the Rev B vs the other two, is that the mounting pressure does not help the secondary extruded aluminum HS press the heatpipes onto the base. Rather, the the pressure that is exerted by the mounting pulls the copper base, and then the four screws pull the aluminum HS against the heatpipes. On the original and the Rev Cu, the pressure exerted by the mounting is transfered to help press the aluminum HS and the heatpipes against the base.
I'll mention again that the second row of heatpipes on the Rev Cu are not flattened, and therefore have only a line of contact with the base, and therefore the heat transfer is probably less effective.
So, the design of the later revisions is lacking, and if there is less than ideal soldering (or no soldering!), then the heat transfer to the heatpipes is probably the weak point.
I would think that filling all the voids with solder would be the best bet.
Another possible problem of the Rev B vs the other two, is that the mounting pressure does not help the secondary extruded aluminum HS press the heatpipes onto the base. Rather, the the pressure that is exerted by the mounting pulls the copper base, and then the four screws pull the aluminum HS against the heatpipes. On the original and the Rev Cu, the pressure exerted by the mounting is transfered to help press the aluminum HS and the heatpipes against the base.
I'll mention again that the second row of heatpipes on the Rev Cu are not flattened, and therefore have only a line of contact with the base, and therefore the heat transfer is probably less effective.
So, the design of the later revisions is lacking, and if there is less than ideal soldering (or no soldering!), then the heat transfer to the heatpipes is probably the weak point.
-
- Site Admin
- Posts: 12285
- Joined: Sun Aug 11, 2002 3:26 pm
- Location: Vancouver, BC, Canada
- Contact:
That's not true, they are flattened in every version, including the Rev B. You can see it in the high res photo I posted earlier.NeilBlanchard wrote:I'll mention again that the second row of heatpipes on the Rev Cu are not flattened, and therefore have only a line of contact with the base, and therefore the heat transfer is probably less effective.
I agree that the smaller mass/size of the material in the base surrounding the heatpipes is probably a factor. If you examine the three Ninja versions, the Rev. B actually has the smallest base (the copper piece that makes contact with the CPU).
It also seems like I should try and get another Rev B sample to see if this is an exceptionally bad one... but then the CU version also did not perform as well as the orig.
Maybe the Ninja's time is over and we just move on?
So the Ninja will now have the same fate as HD-DVD and Rear Projection TVs? No way. I'll be waiting for the Ninja Rev. C.Maybe the Ninja's time is over and we just move on?
Actually, I'm quite happy with my $10 Arctic Cooling Alpine 7s now on half my seven computers with more to come. When good enough will do, it's all about Arctic Cooling.
Well as a very happy owner of a Rev A, I'm not ready to move on! However, the new ones certainly seem inferior, and I suppose we should recommend TRUE or HR-01 to new builders.MikeC wrote:Maybe the Ninja's time is over and we just move on?
The aluminum "secondary heat sink" is pretty much useless. Not only does it have poor contact (via about 4 layers) with the CPU, it has pathetic airflow and relatively poor conduction. I'd wager over 95% of the cooling is provided by the heat pipes. So it's unimportant how well the aluminum block is attached.
On the other hand, the pressure between the copper base and the CPU IHS is extremely important, as has been well documented by Rev B owners.
And I think Neil has a very valid point about how well the heat pipes are attached to the copper base. In the Rev A, all six pipes are mashed pretty hard against each other and the base, and appear from his annotated photo to have over twice the contact area with the base compared to the other two versions.
The second set of pipes in the Cu version do not look flattened in the photo; that's what Neil is referring to. But maybe they are and we just can't see it from that camera angle...
Anyway, the combination of inadequate base/CPU pressure and reduced base/pipe contact seem to doom the Ninja Cu to low-performance (but very pretty) systems.
Possibly. Water has VERY good thermal transfer properties as it is liquid at the temps we are talking about and the convection flow would help vs just relying on thermal conductivity as in a solid.NeilBlanchard wrote:I would think that filling all the voids with solder would be the best bet.
But the big advantage of what I am suggesting is that it is very easy to do vs soldering as it looks like you'd have to flow a LOT of solder to fill the voids in the base and the heat required to do this thoroughly is likely to be destructive or even self defeating since solder flowed in from one direction is likely to flow straight out of the other side of the base.
-
- Moderator
- Posts: 7681
- Joined: Mon Dec 09, 2002 7:11 pm
- Location: Maynard, MA, Eaarth
- Contact:
Hi Mike,
I marked up the photo you posted and I pointed to the round heatpipes that do not appear to be flattened. They are the upper set when it is installed on the CPU, and it would appear that very little heat could be transfered to them. Especially, since there is no solder at all on the Rev Cu, which is more than a little surprising, IMO.MikeC wrote:That's not true, they are flattened in every version, including the Rev B. You can see it in the high res photo I posted earlier.NeilBlanchard wrote:I'll mention again that the second row of heatpipes on the Rev Cu are not flattened, and therefore have only a line of contact with the base, and therefore the heat transfer is probably less effective.