Two (unrelated) questions about storage and hypert
 

ran out of space in the title. hyperthreading was the other topic.

I am buying a i7-2600k Sandy Bridge processor and Z68 mobo hopefully in the next week or so. My first problem is, my current processor is an AMD Phenom II X6 1055T Thuban and my AMD motherboard, which between them are worth about $200 I would imagine. I would like to know the best way to store these after I install the Intel hardware. I tried Googling it, but since I'm storing both, it seems convenient to me to just leave the processor on the mobo and so save the thermal paste. What do I have to be worried about? Static? Dust? One search result mentioned storing cpu's upside down, but again, I'll just keep mine on the mobo.

The second question is: How can I best make use of the hyperthreading on the new Sandy Bridge cpu? I've seen the affinity options in undoc.txt; if I set a core and its hyperthread using the affinity, how does M/Prime/95 use those two logical cpu's? Can I have two workers run on two cores without h/t'ing, and have up to 4 workers on the remaining 2C/4T? (Obviously with the performance hit to those workers for using the h/t'ing)

[QUOTE=Dubslow;267865]rWhat do I have to be worried about?[/QUOTE]Very little, in my experience. Don't do anything obviously stupid, otherwise just put them somewhere they'll stay clean and where you can find them again later! I'd leave the processor in place on the mobo. The less you fiddle with things, the less likely they are to break

Paul

Would a small cardboard box do?

[QUOTE=Dubslow;267872]Would a small cardboard box do?[/QUOTE]

I'd put it in plastic, then in the box, to keep down the dust.

Thanks. Now, for a rephrasing of my second question:
Suppose cpu 0 and 4 are the same physical, 1 and 5, 2 and 6, 3 and 7. So then undoc.txt says to put the affinity as 04152637 so it knows which are the actual physical cores. How does it use that to actually run 4 workers? It assigns a 'helper thread', correct? So that one worker runs on two threads? Does that mean the worker gets all the cycles of one physical core without interruption, as if it was not hyperthreaded?
Or what if I set the affinity as 0415 and ran 6 workers? Would two workers get 0 and 1 all to themselves while 4 other workers split the resources of 2 and 3? I'm just trying to figure out how exactly M/Prime/95 handles HT'ing.

[QUOTE=CRGreathouse;267898]I'd put it in plastic, then in the box, to keep down the dust.[/QUOTE]

Agreed.

Additionally, place the de-humidifier packet (usually a small white sealed envelope) which (hopefully) comes with your new MB in with your old, and seal the plastic as best you can.

Beyond dust, moisture can be a real killer for electronics which is not run for an extended period of time.

[QUOTE=Dubslow;267865]
The second question is: How can I best make use of the hyperthreading on the new Sandy Bridge cpu? [/QUOTE]

I have a 5.0 GHz i5-2500K, which has no hyperthreading.

In my experiments, I have found that in using only 2 cores with LLR, LLR can "steal" processor time from my unused cores, and performance is very high.

When I run four instances of LLR, one on each physical core, the slowdown on all cores is such that I can't even keep pace with the 2 cores doing work.

Now LLR finds Proth and Riesel type primes and is not used for Mersenne Prime hunting like Prime95 is.

My guess would be, though, that if you try to use all 8 hyper-threads as 8 cores, you will experience a slowdown that might mirror my same experience with LLR.

I would recommend testing Prime95 with hyper-threading turned off in your BIOS. Take notes on how long the iterations are taking.

Restart the box with hyper-threading on, and note the speed of each of your 8 workers. Work out algebraically if 8-but-slower will do more work than 4-but-faster.

[QUOTE=chalsall;267906]Agreed.

Additionally, place the de-humidifier packet (usually a small white sealed envelope) which (hopefully) comes with your new MB in with your old, and seal the plastic as best you can.

Beyond dust, moisture can be a real killer for electronics which is not run for an extended period of time.[/QUOTE]

Thanks!! I highly doubt that would have occurred to me, even if I was staring at the packet.

L-Nitro, I'm wondering if I can get a worker to run on a core as if there were no hyperthreading, yet have hyperthreading enabled. That's why I'm asking about affinity and such. As for 8-slow vs 4-fast, my own experience with my current Phenom X6 is much the same with LL/Prime95, as in the more LL, the slower they each go. (See [url]http://mersenneforum.org/showthread.php?t=15771[/url] for details about my current cpu.)

I also would prefer 4-fast over 8-slow, because I'd like to see if I can get one of them to run a 100M digit test. Have you tried that with yours yet? Admittedly, I would prefer to activate hyperthreading to only be active on two cores and have the other two just one-threaded, but I can't do that, hence all the affinity questions.

[QUOTE=Dubslow;267917]
L-Nitro, I'm wondering if I can get a worker to run on a core as if there were no hyperthreading, yet have hyperthreading enabled. That's why I'm asking about affinity and such.[/QUOTE]

Yes that would be nice. But I found hyper-threading produces a tremendous amount of heat on each core when it is active and being used.

Download this (better do it fast, mods have a tendency to treat every link I post as "advertising" for whatever reason, then they delete them)

[URL]http://www.cpuid.com/softwares/hwmonitor.html[/URL]

This is a free Hardware monitoring tool. It will show the temperatures of each of your CPU cores. Watch how the temps compare at idle, then when you are running without hyper-threading, then with hyper-threading.

When you consider that heat = death in the world of electronics, you might think twice before letting hyper-threading kill your system sooner for such a noble, volunteering cause.

[QUOTE=Dubslow;267917]I also would prefer 4-fast over 8-slow, because I'd like to see if I can get one of them to run a 100M digit test. Have you tried that with yours yet? [/QUOTE]

Not on my Sandy Bridge system, but I did on my 5.27 GHz i7-980X which has 6 cores/12 threads. Again, I turned off hyper-threading and my Prime95 workers did more work. There is something wrong with my Ethernet though, and I can't get the box to communicate to the server. I will have to transfer my work to another system while I swap out the motherboard.

Why don't they just include that in CPU-Z? Also, how long would it have taken the 100M test, = how long were the iterations?

[QUOTE=Dubslow;267919]Why don't they just include that in CPU-Z? Also, how long would it have taken the 100M test, = how long were the iterations?[/QUOTE]

if I remember correct from previous reading on here a 100 M test takes about 3-4 years on a cpu and still at least 1 year on a GPU by my math.