Folding Forum

My dad finally gave in and got himself an up-to-date laptop to replace his 9 yr old Dell desktop with a Pentium 4 3.0 ghz.

Is this even worth considering for folding?

Thanks!

Nopers
not worth the power being used for the incredibly small amount of points it gets

Does it have a PCI-e slot (and enough power) to put a modern video card in it then absolutely GPU fold with it. Otherwise, the power usage makes CPU folding on a P4 a hard sell compared to a modern machine. That being said, any WU completed by your P4 is still useful...

No, there's no pci-e slots.

Thanks for validating my thought that it might be a high point to power usage situation.

Another thought, is there any place worth donating this computer to?

I have an old Pent 4/HT 3.2 GHz that I have kept in service, not because of the CPU (which will get around 150 ppd running Uniprocessor WUs), but because I have been able to run a highly overclocked non-cooled GT430 GPU in one of its PCI non-Express slots. This gives me consistently about 6000 ppd on the non-zeta Core 15 WUs. It will run the Zeta work units, but since it only has 96 CUDA cores and has a slow bus, it does them very inefficiently - getting maybe 2500 ppd.

In the grand scheme of things, its probably better to get a modern generation Kepler card for a PCI-e desktop - many more cores, much better power and thermal performance - much better ppd/watt. But, what I can get with that GT430 constantly surprises me.

GreyWhiskers wrote:I have an old Pent 4/HT 3.2 GHz that I have kept in service, not because of the CPU (which will get around 150 ppd running Uniprocessor WUs), but because I have been able to run a highly overclocked non-cooled GT430 GPU in one of its PCI non-Express slots. This gives me consistently about 6000 ppd on the non-zeta Core 15 WUs. It will run the Zeta work units, but since it only has 96 CUDA cores and has a slow bus, it does them very inefficiently - getting maybe 2500 ppd.

In the grand scheme of things, its probably better to get a modern generation Kepler card for a PCI-e desktop - many more cores, much better power and thermal performance - much better ppd/watt. But, what I can get with that GT430 constantly surprises me.

Interesting thought, totally forgot about the video card upgrade that was done to it. Have to look into it and see if its worth pursuing.
Thanks!

eagles_fly wrote:

GreyWhiskers wrote:I have an old Pent 4/HT 3.2 GHz that I have kept in service, not because of the CPU (which will get around 150 ppd running Uniprocessor WUs), but because I have been able to run a highly overclocked non-cooled GT430 GPU in one of its PCI non-Express slots. This gives me consistently about 6000 ppd on the non-zeta Core 15 WUs. It will run the Zeta work units, but since it only has 96 CUDA cores and has a slow bus, it does them very inefficiently - getting maybe 2500 ppd.

In the grand scheme of things, its probably better to get a modern generation Kepler card for a PCI-e desktop - many more cores, much better power and thermal performance - much better ppd/watt. But, what I can get with that GT430 constantly surprises me.

Interesting thought, totally forgot about the video card upgrade that was done to it. Have to look into it and see if its worth pursuing.
Thanks!

If it has no PCI-E slot then there's no card it can run that would be supported by F@H, and even if there were they would be too slow to even be worth it

Best thing you can do with it is donate it to a church or give it away for free

Read GreyWhisker post again. He has a PCI non-e GT430 that works well with FAH, so it must be supported. And for about $50, you can extend the life of that P4, make pretty good PPD and then start saving for a PC upgrade.

FWIW, if you are really interested in acquiring a Fermi card for a PCI non-Express slot, the Zotac ZT-40605-10L is the model number of the GT430 I have. I've seen it available in some of the usual online venues yesterday. I've looked at other manufacturers, and other PCI bus cards, like the GT520 (ZT-50610-10L) and GT610 (ZT-60606-10L), but those both have only 48 CUDA cores vs the 96 cores of the GT430. And, as I said, the GT430 has been able to be overclocked to 910 core clock, 780 memory clock, with complete stability. In many cases, this card is the equal of the 700 MHz clock GTX560M (192 cores) in my laptop. And, remember, the platform is this 2006-era COSTCO special Pentium 4/HT 3.2 GHz desktop with 400 watt power supply still running Win XP, with Nvidia drivers 301.42.

This card is passively cooled - no built in fan, but a massive heat sink on one side of the card. I've been using a Thermaltake Mobile Fan 12 External USB Cooling Fan 120mm AF0007 that I can place on the floor of my old HP case (with the side panel off) blowing right on the heat sink.

Right now, I'm running Core 15 Project: 8074 (Run 11, Clone 36, Gen 101) with a 9:26 TPF, and an estimated 5914 ppd (from the v7.3.6 FAH Control GUI). The GPU temp is about 66 deg C

There are very few possibilities for either PCI non-e GPUs or AGP GPUs, but the ARE a few that are supported and are ideal to extend the life of pre-PCIe systems, especially if they're also still being used for something other than FAH. PPD is low compared to more modern systems but power requirements are pretty low too.

bruce wrote:There are very few possibilities for either PCI non-e GPUs or AGP GPUs, but the ARE a few that are supported and are ideal to extend the life of pre-PCIe systems, especially if they're also still being used for something other than FAH. PPD is low compared to more modern systems but power requirements are pretty low too.

I don't think that there are any viable AGP bus cards any more. I installed the GT430 in the PCI non-e slot when my ATI HD4670 AGP was depreciated by F@H - when this class of ATI cards no longer was assigned any Core 11 work units. At the time, the 4670 was the highest performing AGP bus GPU on the market. I haven't found any CUDA core AGP cards out there. I saw this thread CUDA-enabled AGP card? in the Anandtech forum that seems to categorically dismiss the availability of AGP bus CUDA cards.

As it turned out, the GT430 is getting ~~6,000 ppd vs the max of ~~1,800 ppd for the AGP bus HD4670.

In addition to the GT 430, they made an AMD HD 5450 PCI (non-e) GPU, but with only 80 Stream processors, it would be low performing as well.

I have both an HD5450 and and a GT430 (PCIe versions, but that's not the point). They are really slow GPUs. Since AMD takes one CPU away from SMP (assuming you plan to avoid FahCore_17), the 5450 often produces a net negative PPD, depending on how important that CPU is to your CPU slot. The GT430 does not have this disadvantage.