Potentially the longest thread in history...

[Trotter]

Swimming would be great, but I've got no access. besides, my geek skin fries in no time. I got sunburned today staking out the rows in the garden and planting the tomatoes. Still got squash, cucumbers, onions, corn, and okra to plant in the morning. We plan on being up and planting by 8 a.m., though. Should have it knocked out in 2-4 hours.

I like the flexing idea. I sort of do that with my legs already, but I can incorporate it better.

The butt flex reminds me of the gay Mr. T joke Eddie Murphy has.

 

[Apokalipse]

Actually, I'm not expecting that much out of the 980 (for SuperPI runs anyway). Because of the larger cache there's much more delay than normal. For lightweight programs like SuperPI which aren't really going to take advantage of either the extra cores OR the larger cache, the 980 isn't really going to be anything that special.

That's why you set core affinity in task manager.

The only chance I'll have is when I decide to get some WC.

And a new case to fit it in to?

On my OC run last night I only took the cpu vcore to 1.39v, I reckon I'll try it up to 1.43v tonight

How much is a big number for i7's?
They don't seem to tolerate as high volts as Phenom II - I've had mine set up to 1.7V for superpi. Some people use 2.0V with LN2/LHe.

edit: Does anyone know if the a1 FAH cores should only be using 50% of the CPU resources even when I've set them to 100%? Limitation of the a1 core?

It's probably only using 2 threads.

 

[S0ULphIRE]

That's why you set core affinity in task manager.

yeah, all cores will be *available*, BUT, due to how the actual program is written, it won't be able to utilise them.

And a new case to fit it in to?

nah, already blown a heap of cash on this build, don't want to spend any more

How much is a big number for i7's?
They don't seem to tolerate as high volts as Phenom II - I've had mine set up to 1.7V for superpi. Some people use 2.0V with LN2/LHe.

well stock voltage is (I think) 1.25v for the 920. Any voltages above 1.32ish and you'll need more than a stock cooler.

It's probably only using 2 threads.

Yeah, solved that problem.

 

[Apokalipse]

yeah, all cores will be *available*, BUT, due to how the actual program is written, it won't be able to utilise them.

That's not what I meant.
I meant setting affinity so it only uses one core, so it doesn't keep switching between different cores.
That way it doesn't have to keep shifting data to different cores/areas of cache, which means cache latencies have less effect, and you get better performance.
And also set real time priority so that all clocks of that core go to superpi.

Also, in the case of Phenom II's, you can just clock one of the cores high and keep the rest at low clocks to reduce the possibility of crashing (also reduces heat, which Phenom II's really like), then run superpi on the one core you clock high.

It also helps to use an OS cut down as much as possible (heavily nlited XP - I made one that's about 100MB on the install CD) and close every single program (including explorer)
Unplugging the mouse can shave a few milliseconds too (mouse movements can use more CPU time than you think, especially if it's a high DPI mouse)

nah, already blown a heap of cash on this build, don't want to spend any more

At least if you have a good case and a good watercooling system, it's something you don't need to keep upgrading to keep up with performance (just get a new mounting plate once in a while if the socket changes)

 

[Baez]

What do you know... I DO have two hands.

 

[S0ULphIRE]

Wow. Folding at home REALLY loves my 980x I'm getting an average of about 2:23 tpf, and that's with only a tiny OC on.

That's not what I meant.
I meant setting affinity so it only uses one core, so it doesn't keep switching between different cores.
That way it doesn't have to keep shifting data to different cores/areas of cache, which means cache latencies have less effect, and you get better performance.
And also set real time priority so that all clocks of that core go to superpi.

I already do that

Also, in the case of Phenom II's, you can just clock one of the cores high and keep the rest at low clocks to reduce the possibility of crashing (also reduces heat, which Phenom II's really like), then run superpi on the one core you clock high.

Not sure if you can do that with i7's. You can disable cores and only run one, never tried that though.

It also helps to use an OS cut down as much as possible (heavily nlited XP - I made one that's about 100MB on the install CD) and close every single program (including explorer)
Unplugging the mouse can shave a few milliseconds too (mouse movements can use more CPU time than you think, especially if it's a high DPI mouse)

I just open up msconfig in win7 and disable nearly everything. I don't unplug the mouse tho just a crappy microsoft one but hey I'll give it a shot and see what happens.

At least if you have a good case and a good watercooling system, it's something you don't need to keep upgrading to keep up with performance (just get a new mounting plate once in a while if the socket changes)

I already have a HAF 932, and a watercooling system is en route, so it appears I've got all bases covered!

 

[Apokalipse]

I saved these nlite config files:
http://sites.google.com/site/apokalipse/nanoXP.rar
loading these in nlite will create an XP install CD of close to 100MB (down from 589MB of original XP SP3 disc).
You might want to integrate your own SATA/chipset drivers though (otherwise it might not detect your hard drives)

There is also MicroXP, which is slightly smaller. But I discovered there are a couple of services it doesn't have, which futuremark Systeminfo needs (which means you can't use ORB validation in 3dmark06/vantage)

 

[Hefemeister]

^ Do they still use the same melody?

 

[S0ULphIRE]

LAAAAAAAAAAAAAWL!

 

[Yami]

LULZ!