Thermalright SI-128: Evolution of a Past Master
-
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
It doesn't make sense to me.
1. Many other Thermalright units are screw in with a back plate, although in comparision only, this unit is a lighweight.
2. There doesn't seem to be a good way to mount it on a vertical motherboard. The heat pipes are either going to be like this \/,
or they will be like this /\. Either way the evaporated liquid is suppose to head downhill at some point. And do you see a clear path for the condensed liquid to return?
Surely you can't use it on a vertical mother board with the heat pipes facing out the sides like this < or >.
And stop calling me Surely.
Aris
1. Many other Thermalright units are screw in with a back plate, although in comparision only, this unit is a lighweight.
2. There doesn't seem to be a good way to mount it on a vertical motherboard. The heat pipes are either going to be like this \/,
or they will be like this /\. Either way the evaporated liquid is suppose to head downhill at some point. And do you see a clear path for the condensed liquid to return?
Surely you can't use it on a vertical mother board with the heat pipes facing out the sides like this < or >.
And stop calling me Surely.
Aris
Hmm good review, though I got to say that the SI-128 is working well for me passively in my Antec Fusion. I am using the S-Flex 1200 RPM models and I think that I could easily go down to the 800 RPM models. Of course my processor is an older model, and Athlon 64 3700 s939.
I will also mention that the included bracket likely will not work on many K8 boards. My board and quite a few others that I have encountered use a single screw in the center of the two sides to hold down the bracket. Of course I had no issues using the stock bracket.
I will also mention that the included bracket likely will not work on many K8 boards. My board and quite a few others that I have encountered use a single screw in the center of the two sides to hold down the bracket. Of course I had no issues using the stock bracket.
-
- Moderator
- Posts: 7681
- Joined: Mon Dec 09, 2002 7:11 pm
- Location: Maynard, MA, Eaarth
- Contact:
Looking at some of the recent crop of densely finned heatsinks like this one, I can't help but wonder if low-flow performance would be improved by surgically removing every 2nd or 3rd fin.
Either way, I'm still looking for an aftermarket HS for my 2400 based HTPC, this isn't it. No hurry, because even the stock fan at about 7v is adequate for cooling my X2 - 3800EE, and inaudible across the room.
Either way, I'm still looking for an aftermarket HS for my 2400 based HTPC, this isn't it. No hurry, because even the stock fan at about 7v is adequate for cooling my X2 - 3800EE, and inaudible across the room.
All modern heat pipes use capiliary action within a "wick" on the inside of the pipe to return the working fluid to the hot part of the pipe (actually to keep a uniform "depth" of fluid across the entire inner surface of the pipe, which has the same effect as fluid boils away from the hot spots). This mechanism is almost completely independent of orientation. For example the Condor in my system has the radiator below the heat block and it works just fine.aristide1 wrote:There doesn't seem to be a good way to mount it on a vertical motherboard. The heat pipes are either going to be like this \/,
or they will be like this /\. Either way the evaporated liquid is suppose to head downhill at some point. And do you see a clear path for the condensed liquid to return?
The wicking is typically a sintered layer, or a large number of very fine grooves.
The advantage of a fatter heat pipe (in this case 8 mm) is more surface area for wicking, which makes it more efficient. It should be possible to boil off more liquid in a fatter pipe and thus increase its potential heat load, as compared to a smaller pipe. But don't ask me how much!
Hi, I was looking for a heatsink for my new system that I just put together and as my case isn't wide enough for any of the 120mm fan tower sinks were out. I looked at the Si-128 and concluded that the fins would be too dense for effective very quiet cooling. Looks like I was right! Also, IIRC, it's ~90mm tall plus 25mm of fan and, preferably at least 25mm of space above for air to get in to the fan and you're looking at 130-140mm, not short at all. I think the Si-120 with wider fins would be better for quiet cooling, it's still as tall and availability is a problem. I might have gotten an XP-120 if I could have found one easily as it has wide fin spacing and low height. I ended up with a Zalman 9500AT as, as best I could work out, it has best thermal performance at low flow of any heatsink that would fit in my case. That the minimum speed the (PWM) fan would run was ~1300rpm and too loud wasn't so good. I've now gotten round that and it's cooling my over-clocked Core 2 Duo at ~1000rpm, good enough
Seb
Seb
Thanks to SPCR for a review I've been wanting to see, I've been curious about this HS for a while now. Would've liked to see it compared directly to the recently reviewed Big Typhoon. If the 128's primary advantage is it's form factor, then it makes sense to compare performance and dimensions against the other big "umbrella" heatsink on the market.
Also, nitpick time: The article states the SI-128 is evolved from the SI-120, which in turn comes from the XP-120. Does the SI-97 not seem like a more direct ancestor than the XP-120? Not that it matters much.
Also, re: fan+airspace height- I wonder if the overall height could be cut by removing the "support wires", and mounting the fan under the heatsink. Seems like I've seen a picture where somebody used undermounting on the SI-97.
Also, nitpick time: The article states the SI-128 is evolved from the SI-120, which in turn comes from the XP-120. Does the SI-97 not seem like a more direct ancestor than the XP-120? Not that it matters much.
Also, re: fan+airspace height- I wonder if the overall height could be cut by removing the "support wires", and mounting the fan under the heatsink. Seems like I've seen a picture where somebody used undermounting on the SI-97.
I am a bit skeptical about the SPCR conclusion on the SI-128. I have mine on an E6600 with a Nexus fan and right now the fan is not spinning and the CPU temp is 34C. In idle, after some heavy load, the fan would be spinning at ~450 rpm and CPU temp drops to 24C (I touch the SI-128 and it feels cool). It is by far the quietest part of my computer, definitely quieter than the S12E+650 (highly recommended by SPCR) and 7900 GTX (stock cooler) that I have.
I'm using an XP-120 with the heatpipes vertical (bends down) on a vertical motherboard, as shown here. It works just fine. For comparison, I tried it with the MB horizontal and saw no difference in CPU cooling under sustained 100% load.aristide1 wrote:2. There doesn't seem to be a good way to mount it on a vertical motherboard. The heat pipes are either going to be like this \/,
or they will be like this /\. Either way the evaporated liquid is suppose to head downhill at some point. And do you see a clear path for the condensed liquid to return?
I decided to stay away from the Ninja or Ultra-120 because many people reported mounting problems, that adversely affected temps.
My SI-128 seems to make very good contact with the CPU, after some hassle with the backplate on my Gigabyte DQ6.
I had to loosen the backplate screws to be able to mount the SI-128, then I tightened the screws back.
On the DQ6 mounting a Ninja or Ultra-120 would have meant a lot more trouble, due to this backplate.
I chose the DQ6 because it has the best stock chipset cooling for Skt 775, and it has slightly better sound than the DS4 (same cooling).
My SI-128 seems to make very good contact with the CPU, after some hassle with the backplate on my Gigabyte DQ6.
I had to loosen the backplate screws to be able to mount the SI-128, then I tightened the screws back.
On the DQ6 mounting a Ninja or Ultra-120 would have meant a lot more trouble, due to this backplate.
I chose the DQ6 because it has the best stock chipset cooling for Skt 775, and it has slightly better sound than the DS4 (same cooling).
Sorry ... have I missed something ?
Those 2 articles :
- http://www.silentpcreview.com/article646-page3.html
- http://www.silentpcreview.com/article719-page4.html
give a completely opposite view of how the top contenders (Ninja and Ultra 120) fare against one another !
How is that possible ?
Those 2 articles :
- http://www.silentpcreview.com/article646-page3.html
- http://www.silentpcreview.com/article719-page4.html
give a completely opposite view of how the top contenders (Ninja and Ultra 120) fare against one another !
How is that possible ?
-
- Site Admin
- Posts: 12285
- Joined: Sun Aug 11, 2002 3:26 pm
- Location: Vancouver, BC, Canada
- Contact:
Yes, those results changed, and the Ninja's results were better than the Ultra120's this time around. It does not change our opinion of these HS, however: They are both excellent for low airflow, quiet cooling, they are at the top of this niche category. (Of course, neither of the versions we refer to are actually "current" anymore... )jjr wrote:Sorry ... have I missed something ?
Those 2 articles :
- http://www.silentpcreview.com/article646-page3.html
- http://www.silentpcreview.com/article719-page4.html
give a completely opposite view of how the top contenders (Ninja and Ultra 120) fare against one another !
How is that possible ?
Do consider these many factors:
1) completely different HS test platform with a different CPU and a different way of assessing CPU heat. Note that all the result temperatures went up with this platform.
2) There are differences in where the CPU thermal diode is, in the secondary heat paths for the CPU, in the proximity of other heat sources on the motherboard, etc. The thermal linearity of the CPU/MB could also be different from the first test platform, which would magnify thermal differences at the extremes (read: low airflow).
3) the resolution of the thermal measurement tools really can't be trusted <2C or so. Even with the same heatsink on the same platform, removing and reinstalling it for a second test often gives different results. The differences are usually small (<2C) but they come up over and over again.
4) Another test on the same platform could lead to yet another flip of the results.
IS there a way to eliminate the vagaries and create a 100% accurate high resolution HS test instrument? Probably.
Will we do it? Maybe, maybe not.
It would have to be a custom made "heater" based device like the kind used by http://www.frostytech.com/testmethod_mk2.cfm -- the challenge with such devices is in duplicating the HS retention systems on motherboards. It would not be cheap.
Having said that, I have to admit Frostytech may have found a solution. It looks like they use two difference heater elements and top panels for Intel vs AMD processors. Perhaps they would be willing to share their knowledge with us...
does SPCR know a top mount wire fan clip is made by thermalright?
http://www.thermalright.com/a_page/main ... sories.htm
Just wanted to make sure this is common knowledge as every review states you need open flange fans. I know its not included w/ the heat sink (and it really should be), but FYI anyways.
http://www.thermalright.com/a_page/main ... sories.htm
Just wanted to make sure this is common knowledge as every review states you need open flange fans. I know its not included w/ the heat sink (and it really should be), but FYI anyways.
I missed that mistake in the review, but I already mentioned this twice in the forums, with the SI-128 you don't need an other clip or open corner fans:
http://web.clicknet.ro/opticode/SI-128% ... us%202.jpg
http://web.clicknet.ro/opticode/SI-128% ... us%202.jpg
Thanks for the answer. I went (rapidly) through the test methodology article and realised that you had picked that up too.MikeC wrote: Yes, those results changed, and the Ninja's results were better than the Ultra120's this time around. ...
One question : I didn't see any mention of mobo orientation. (Hope I haven't missed it).
Could that be an parameter in this case ?
Ninja has differently laid out heatpipes and I could see the ones on the side and at the bottom being less efficient when mounted vertically. (That's 75% of the heatpipes with variable loss)
On the Ultra 120 : only 50% ?
-
- Site Admin
- Posts: 12285
- Joined: Sun Aug 11, 2002 3:26 pm
- Location: Vancouver, BC, Canada
- Contact:
All heatsinks are tested with the motherboard sitting flat on the test bench as shown in the photographs. In a few rare cases, a vertical orientation (of the mb) was tried as well. There should be no difference either way with either HS. As cmthomson already pointed out earlier in this thread, modern heatpipes are essentially multidirectional.jjr wrote:One question : I didn't see any mention of mobo orientation. (Hope I haven't missed it).
Could that be an parameter in this case ?
Ninja has differently laid out heatpipes and I could see the ones on the side and at the bottom being less efficient when mounted vertically. (That's 75% of the heatpipes with variable loss)
On the Ultra 120 : only 50% ?
-
- Posts: 414
- Joined: Wed Jan 17, 2007 7:55 am
- Location: EU
That's the theory, however I wonder whether things do really work always that way. Just think of the Arctic Cooling Freezer & Freezer Pro which feature a preferred orientation.MikeC wrote:As cmthomson already pointed out earlier in this thread, modern heatpipes are essentially multidirectional.
On the other hand, most manufacturers say nothing about it, but I wouldn't be surprised if they optimized their tower coolers for vertical mounting (and I believe it would make sense)
Well, believe what you want.J. Sparrow wrote:That's the theory, however I wonder whether things do really work always that way. Just think of the Arctic Cooling Freezer & Freezer Pro which feature a preferred orientation.MikeC wrote:As cmthomson already pointed out earlier in this thread, modern heatpipes are essentially multidirectional.
On the other hand, most manufacturers say nothing about it, but I wouldn't be surprised if they optimized their tower coolers for vertical mounting (and I believe it would make sense)
Modern heat pipes work just fine in any orientation because they use wicking, not gravity, to move the fluid around.
-
- Site Admin
- Posts: 12285
- Joined: Sun Aug 11, 2002 3:26 pm
- Location: Vancouver, BC, Canada
- Contact:
It's not just theory. I know we've tested a few heatpipe HS in both vertical and horizontal positions and generally found little or no difference. The very first ones (going back some 3 years to the Thermalright sp94/97) actually came with cautions about making sure the condensor end of the hetapipes were positioned higher than the evaporator side, but heatpipes appear to have improved significantly since then.J. Sparrow wrote:That's the theory, however I wonder whether things do really work always that way.MikeC wrote:As cmthomson already pointed out earlier in this thread, modern heatpipes are essentially multidirectional.
According to who? Nothing in our experience suggests that the AC Freezer works better one way vs the other, and neither does any of AC's own documents on those products. The only comment on orientation refers to the fan -- ideally, it should be set up so that the fan blows from front to back.Just think of the Arctic Cooling Freezer & Freezer Pro which feature a preferred orientation.
On the other hand, most manufacturers say nothing about it, but I wouldn't be surprised if they optimized their tower coolers for vertical mounting (and I believe it would make sense)
It is in tower coolers that heatpipes independent of orientation are most useful, which makes your conjecture highly unlikely. For the purpose of illustration here, there are only two types of heatpipes -- the ones where at least some assistance from gravity is needed to return the coolant back to the evaporator, and the ones where gravity is not needed (because of excellent wick action). With a tower heatsink, it's only when the motherboard is vertically mounted (as in a tower case) that any difference between the heatpipe types would show up. In a HTPC case, a tower HS could use either kind of heatpipe w/o any effect on performance, because the condensor ends would always be higher.
-
- Posts: 414
- Joined: Wed Jan 17, 2007 7:55 am
- Location: EU
When I was about to buy the Freezer Pro I've read several tests and reviews, and some of them stated that it worked better when the cooler is installed in a tower case (which I previously referred to as "vertical mounting" and probably wasn't the best choice of words )
After a quick googling this Madshrimp article is the only reference I could promptly find. The text says:
The Freezer features quite a fast fan, so I think we can safely assume there is little to none natural convection happening: it's all forced. Moreover, the sample of the Freezer I own, tends to exhaust more air from the sides than from the back. In a tower setup this means that air exhausted from the left side (facing down) will rise back up towards the cooler.
For these reasons I would expect better performance in a desktop setup, but since it's not so, it seems reasonable to me to attribute any difference to different cooling efficiency from the heatpipes.
PS: It might be only a wrong feeling, but your answer seemed a little brash to me. I don't feel like I have The One Truth in my hands and if it seems so, then it must be only because my english grew a bit rusty over the years.
After a quick googling this Madshrimp article is the only reference I could promptly find. The text says:
The measurements seem to agree, too.In the documentation Arctic Cooling provided with the Freezer64 they mentioned that the heatsinks performs best when it sits vertical, so we’ve included results with the heatsinks orientated like that.
The Freezer features quite a fast fan, so I think we can safely assume there is little to none natural convection happening: it's all forced. Moreover, the sample of the Freezer I own, tends to exhaust more air from the sides than from the back. In a tower setup this means that air exhausted from the left side (facing down) will rise back up towards the cooler.
For these reasons I would expect better performance in a desktop setup, but since it's not so, it seems reasonable to me to attribute any difference to different cooling efficiency from the heatpipes.
PS: It might be only a wrong feeling, but your answer seemed a little brash to me. I don't feel like I have The One Truth in my hands and if it seems so, then it must be only because my english grew a bit rusty over the years.
Yes, that would happen if the CPU cooler was the only thing in the case, but I suspect in reality the case fan extracts this warm air before it has a chance to rise back up, ie inertial forces are stronger than buoyancy forces in this instance.In a tower setup this means that air exhausted from the left side (facing down) will rise back up towards the cooler.
Nah, don't worry, that's just Mike's style. He only gets really harsh if we mention SilenX (joke!).PS: It might be only a wrong feeling, but your answer seemed a little brash to me.
-
- Posts: 414
- Joined: Wed Jan 17, 2007 7:55 am
- Location: EU
Yes that's true, but many testbeds are made just that way: a bare board outside the case.jaganath wrote:Yes, that would happen if the CPU cooler was the only thing in the case, but I suspect in reality the case fan extracts this warm air before it has a chance to rise back up, ie inertial forces are stronger than buoyancy forces in this instance.
However, the point I'm interested in, is that any difference in cooling performance cannot be explained by means of different convection happening in different orientations, because it's always forced.
If so, then it must look for another explanation. And all being the same, the first thing I would think of, is those heatpipes work more effectively when they are horizontal.
no S****X stuff herejaganath wrote:Nah, don't worry, that's just Mike's style. He only gets really harsh if we mention SilenX (joke!).
Out of curiosity, what temperature range do you get?PPGMD wrote:Hmm good review, though I got to say that the SI-128 is working well for me passively in my Antec Fusion. I am using the S-Flex 1200 RPM models and I think that I could easily go down to the 800 RPM models. Of course my processor is an older model, and Athlon 64 3700 s939.