| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
| World Community Grid Forums
|
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 25
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I hope your case is Wide cuz the NH14 is huge. ( NH-D14 )
----------------------------------------Mine measures from the surface of the motherboard to tips of heatsink 6-5/8" (169 MM) I can just barely close the side of the case and it just touches the side... Cools good but it looks more like a Volkswagen radiator than a CPU heatsink. ![]() [Edit 2 times, last edit by Former Member at Dec 15, 2011 7:22:39 PM] |
||
|
|
Hypernova
Master Cruncher Audaces Fortuna Juvat ! Vaud - Switzerland Joined: Dec 16, 2008 Post Count: 1908 Status: Offline Project Badges:
|
tomast you are right it is huge. In the past I used it on the 1366 socket and had no problem to fit it in terms of height, my standard box is large enough. But I ran into collision problems with the huge 140mm fans and the memory sticks. I had to use only one fan instead of two and there was immediately a loss of performance. This time I have low profile sticks and there is more distance between the two groups of ram slots and the CPU socket. I plan to use it bottom up and not sideways.
----------------------------------------Cross fingers ![]() ![]() |
||
|
|
OldChap
Veteran Cruncher UK Joined: Jun 5, 2009 Post Count: 978 Status: Offline Project Badges:
|
This is beginning to look like a water cooling only monster.
----------------------------------------With the fans/ram clearance issue is it not possible to slide the one fan upwards and re-attach??? Just thinking that degraded performance from that fan would be better than no fan at all. Or maybe fit a 120mm fan on that side?? ![]() |
||
|
|
sk..
Master Cruncher http://s17.rimg.info/ccb5d62bd3e856cc0d1df9b0ee2f7f6a.gif Joined: Mar 22, 2007 Post Count: 2324 Status: Offline Project Badges:
|
Hypernova, any change in task performance since the RAM upgrade?
----------------------------------------PS. Congrats on your 150Year milestone and runtime rank of 150 [Edit 2 times, last edit by skgiven at Dec 16, 2011 9:18:39 PM] |
||
|
|
Hypernova
Master Cruncher Audaces Fortuna Juvat ! Vaud - Switzerland Joined: Dec 16, 2008 Post Count: 1908 Status: Offline Project Badges:
|
No results yet. I need a few more days for a significant statistic. A little more patience
----------------------------------------![]() ![]() |
||
|
|
Hypernova
Master Cruncher Audaces Fortuna Juvat ! Vaud - Switzerland Joined: Dec 16, 2008 Post Count: 1908 Status: Offline Project Badges:
|
I have taken away that Corsair H100 cooler and replaced it with a Noctua NH-D14 SE2011, which is a CPU cooler for the socket 2011. A pretty large cooler with two fans, a front 120mm and a middle 140mm. I used for thermal paste the excellent Spire SP802 Bluefrost and not the one included in the Noctua box.
----------------------------------------The result has been immediately improved. With the Noctua fans running at very moderate speed I can crunch at 4.3 GHz with CPU core temps stabilizing in the sixties degrees celsius. With the H100 at 4.2 I was immediately at 101-105 degrees probably near TJ limit. Why the H100 did behave poorly. - First the large radiator mounting with the two fans was not optimal. Beware if you buy such a cooler, you must have a casing that allows to fix the two fans and the radiator on the top of your casing. - Second the cooling head that comes on the CPU, is not perfectly well designed. The four metal arms that are drilled to the 2011 socket spec are not perfect. There is not enough clearance and I had to enlarge the holes otherwise it was impossible to position it and press it correctly. Poor design here. - The cooling head comes already with a thick layer of thermal paste, and probably not of very high quality. I used it as it was. I think those three elements are the cause. On my ANTARES rig, I use the Corsair H80, that runs perfectly well since nearly a year. I run a 980X at 4.2 Ghz and have core temps around 58 Degrees. ![]() |
||
|
|
oldDirty
Cruncher Joined: Mar 10, 2009 Post Count: 21 Status: Offline Project Badges:
|
When the 2011 Box is set up well, you will get better performance than the 980x System for sure.
----------------------------------------![]() |
||
|
|
Hypernova
Master Cruncher Audaces Fortuna Juvat ! Vaud - Switzerland Joined: Dec 16, 2008 Post Count: 1908 Status: Offline Project Badges:
|
After having let the 3960X stabilize with its new parameters here are the latest results averaged over eight days, and compared to my reference rig ANTARES.
----------------------------------------MERCURY (3960X) 4.3 Hz, 8GB DDR3 2000 MHz RAM, 300 GB Raptor HD Score: 53'237 ppd ANTARES (980X) 4.2 GHz, 6GB DDR3 2000 MHz RAM, Crucial M3 SSD (250 GB) Score: 53'540 ppd The results are now very similar. The two rigs crunch exactly the two projects, GFAM and C4CW. I can tweak a little more the MERCURY rig but the results are there. If your only reference is crunching power then the 3960X is not better that a 980X, it is even a little worse. If power consumption is a major factor for you then 3960X beats the 980X by a clear margin. I have to upgrade three rigs that are running with 950s. I have a spare 980X and can have two 990X at a good price. I will not go through the hassle of changing motherboard and socket for the 3960X. Just upgrade the BIOS and there you go. We will see when the 8 or 10 cores come mainstream. Then the 980X/990X leadership will start to wobble seriously. P.S. You have noticed that ANTARES has an SSD that is btw connected through a SATA 6GB high speed link. This may give an advantage to ANTARES for very frequent checkpointings. My seetings for both rigs is at 800 seconds for checkpointing frequency, so this should not be a major impact. ![]() [Edit 1 times, last edit by Hypernova at Dec 30, 2011 5:56:03 AM] |
||
|
|
Dark Angel
Veteran Cruncher Australia Joined: Nov 11, 2005 Post Count: 728 Status: Offline Project Badges:
|
The SSD will give a substantial advantage only if the threads are IO bound, in other words probably nothing substantial on anything except CEP2.
----------------------------------------And seriously, someone suggested spreading a layer of thermal compound? That's a guaranteed way to get air trapped in the mount and make it fail early. Nobody who knows what they're doing has used that method for fifteen years and for good reason. Thank you for doing the comparison, Hypernova, plenty people have wanted to know but not everyone has the resources to try both. ![]() Currently being moderated under false pretences |
||
|
|
sk..
Master Cruncher http://s17.rimg.info/ccb5d62bd3e856cc0d1df9b0ee2f7f6a.gif Joined: Mar 22, 2007 Post Count: 2324 Status: Offline Project Badges:
|
Thanks for your testing and analysis Hypernova. Good to know it's not worth updating to the 3900's.
Part of the reason the 980 performs slightly better is the OC'ing methods; on the 980X you overclock much of the system, and on the 3960X you just increase the turbo. The 2011 architecture is more energy efficient generally, but several overclocked components vs one is also a factor. Presumably, when Intel stick another 2 cores into a desktop processor that will make it the latest and greatest, but even then it would only be ~30% more productive, and would still require an entirely new system, at substantial expense. There are 10core Xeon's, so there is room for a 20thread desktop processor, but I suspect these might only come with a drop in frequency, undermining core count. You might want to consider that power usage reflects heat output. Buy more 980 or 990's and your heat problem will remain. Wait until April or May for 22nm's and rest assured the temps will drop with these CPU's. This would allow you to run cooler and longer into the summer. GL |
||
|
|
|