Pentium 4-M 3.06Ghz SL7DU - 100 PPD, lolwut (ChasR help)

[ThePerfectCore]

So I had a craptop 1.8 Celeron in my junkpile that I recently "upgraded" (for $10, heh) with the cpu in the thread title.

This particular Pentium 4 has HT support and I've been using notfred's client to run two single-core instances on the machine.

With the 1.8 Celeron, I got 60 ppd.

With the new 3.06Ghz chip, I get 100 ppd (50 ppd for each instance).

It appears I'm another victim of the "omg more megahurtz!" race. Would disabling HT and running a single instance increase the efficiency of this machine?

Right now the laptop is pulling a consistent 85W out of the wall. My desktop can do 15000 ppd on 400w, or 37.5 ppd/watt. This crap laptop is doing 1.17 ppd/watt, which is HORRIBLE.

I'm not expecting thousands of ppd, but I was expecting at least 120 ppd after increasing the cache size and nearly doubling the clock speed. Even then the machine might not be worth using. Is there anything I should try beside throwing it off my balcony?

tldr - netburst is crap

[Zerix01]

Are you using notfred's VM? That could be the problem as it might not like the overhead from the VM. For single core you should just run two single core console clients and you might see better results.

[ChasR]

No reason to run a VM. THe Windows and Linux uniprocessor clients have nearly identical performance so the VM overhead is a loss. You should be making about 150 ppd everyday and range up towards 200, but that's about it for that rig. Is it throttling? The Linux mix of WUs may be the cause, but its been years since I even ran one.

[AmbientFiction]

I run a P4 3Ghz HT.....After doing research I found that running 1 uniproc was the best was to fold on it. Which the info on it is here (http://fahwiki.net/index.php/Running_multiple_FAH_clients_%28not_SMP%29#Hyper-Threading). I'm getting 238ppd which is about every 3 days for turn over.

[ThePerfectCore]

The hard drive failed a while ago and has been removed, so I'm using a notfred bootable CD. There is no OS permanently installed, so I'm not running VMs inside of VMs, nor can I tell if it's throttling.

I'd be happy with 200+ ppd. With only one instance running the power consumption will drop down and maybe make this thing a little more efficient to use.

Time to go find an old ATA6 44-pin drive, I guess.

[ChasR]

notfred's is fine as long as it isn't in a VM. I've tested HT over and over and every time I do, with exceptions on very few WUs, I find a 25% improvement in ppd. Its use to run two instances on the uniprocessor client slows down the science though and is discouraged by Stanford. Problem with notfreds is it's too light a distro. I would suggect you run cat /proc/cpuinfo, but I'm not sure notfreds will do that.

The benchmark machine is a P4C @ 2.8 GHz. Benchmark ppd is 110, running a single instance. There are some bonus WUs out there that double that number, but you won't always get one. My P4C @ 3.2 has one unit running @ 171 ppd and the other at 93 ppd. There is a lot of WU variability.