TR Forums

notfred wrote:

You're going to need a separate USB drive per VM, and note that it can get tricky under Linux hosting to get the VM to pick up the drive so that the guest VM can use it. I don't run Windows so I don't know if it suffers from the same temperamental behaviour
I am working right now on an "Official" diskless VMWare image that will use a .vmdk virtual disk to run from and save and restore the WUs, keep your eyes open for the next release.

That sounds like the perfect solution

theMASS wrote:

You can get the same or better PPD with 2 clients. Right now @ 3.6GHz I have a 2653 and 3064 running and am getting 5800PPD. A single SMP 2653 can do ~4200PPD if you've got your rig setup right.

Linux SMP or WinSMP? And what's your secret

cob wrote:

notfred wrote:

You're going to need a separate USB drive per VM, and note that it can get tricky under Linux hosting to get the VM to pick up the drive so that the guest VM can use it. I don't run Windows so I don't know if it suffers from the same temperamental behaviour
I am working right now on an "Official" diskless VMWare image that will use a .vmdk virtual disk to run from and save and restore the WUs, keep your eyes open for the next release.

That sounds like the perfect solution

theMASS wrote:

You can get the same or better PPD with 2 clients. Right now @ 3.6GHz I have a 2653 and 3064 running and am getting 5800PPD. A single SMP 2653 can do ~4200PPD if you've got your rig setup right.

Linux SMP or WinSMP? And what's your secret

@notfred good news to hear about .vmdk virtual disk. I briefly tried to set one up and kept getting host/guest driver compatibility errors.

@cob no secrets... A lot of reading, 2 weeks of trial and error, quality components (RAM gets finicky @ 1200MHz), and having a motherboard that allows tRD adjustments is a requirement. Luck also plays a role, WU assignment can very PPD over 2000 The box I'm talking about is running notfred's 1 SMP/2cores (2 SMP Clients on a Q6600) In general however higher PPD is easier to obtain running 2 SMP clients under VMWare Server (tends to get a better WU mix) It was really a fluke that I got this machine running so fast I was trying to build a energy efficient box but it was running so well I just kept cranking it up. It still does quite well on a PP/Watt basis as it runs about 150W under full load. I don't know if this plays factor in performance (maybe from a stable power perspective or non restricted airflow) but this box has NO HD, NO CD, NO Video...Just RAM, CPU, HSF, and USB stick.

theMASS wrote:

but this box has NO HD, NO CD, NO Video...Just RAM, CPU, HSF, and USB stick.

Ah well that's explains a lot Vista and VMWare like to steal a few resources on my system

I've mentioned the diskless method to a few of the members running dedicated crunchers on my team. It's a superb idea and something I'll be keeping a very close eye on

Hi guys,

I'm getting the following problem whilst trying to set this up on a VMWare Server. The USB drive I have is being detected correctly, and I've downloaded the iso a couple of times just in case it was something I did....



Any ideas?

It's because I was half way through doing a VMWare appliance (save and restore from a virtual disk - /dev/hda1) when Stanford switched the URL and I had to release early. Don't worry about the message, it's just saying that it can't find the VMWare virtual disk.

Ah ok, thanks I look forward to the release where you can save to a virtual disk then

Would this also be the reason why FahMon claims that the diskless instance is *Hung*?

Thanks for a great program.

Yeggstry wrote:

Ah ok, thanks I look forward to the release where you can save to a virtual disk then

Would this also be the reason why FahMon claims that the diskless instance is *Hung*?

Thanks for a great program.

FahMon has trouble monitoring VM clients. If you have a windows box FahSpy does much better monitoring clients running in VM's.
http://fahspy.com/index_eng.shtml

Cool thanks I didn't know this one existed so I'll try it out tonight.

Well I finally got around to running this tonight on an E6600. And, as was to be expected, the single SMP client allocated to the E6600's two cores was only using ~90% of the CPU. So with a little edit of the fold.iso it was simple enough to persuade it to download and run two SMP clients, thus maxing out the CPU




Frame times will ofcourse be ~90% longer than they would be with a single client, but the ppd should be a good 10% higher. And every little counts


Thank-you for an awesome application

Yeggstry wrote:

Ah ok, thanks I look forward to the release where you can save to a virtual disk then

Would this also be the reason why FahMon claims that the diskless instance is *Hung*?

Thanks for a great program.

You'll know this by now, but I'll pop the solution up anyway.

Ticking the 'Ignore asynchronous clocks' option in Fahmons options solves the 'hung' message.

Hi

2 machines are running the USB stick version fine (love it by the way!)
I'm trying on a 3rd machine but it just keeps rebooting after the trying to start core part....overclock is minimal and was folding in windows just before trying this.
This one however is a 4gb usb flash drive ,the others were 2gb,all same brand though.

I have three boxes all hard wired into a linksys wet54g ethernet bridge which is getting our internet wireless from upstairs.


I tried formatting it again using HPUSBFW utility as FAT (not NTFS or FAT32) then unzipped the files directly to the usb stick
Then I go Start>Run>CMD and type syslinux D: and it copies onto the USB drive a file called ldlinux (system file)

I then unplug and put it in and boot to it.

Its here where this PC goes into an endless loop of rebooting after the trying to strat core part.