mfakto: an OpenCL program for Mersenne prefactoring

[flashjh]

Originally Posted by kracker

What? the 590 IS clocked down, I believe heat issues, so it WILL be slower than 2 580's, plus SLI and mfaktx doesn't work optimally.

Lots of info out there about the 590, but here is a good writeup. The paragraph right before the video talks about the clock speed. Though the 590 can be OCd a little, each onboard 580 still can't reach an individual 580s speed (at least not without flashing a new BIOS to the 590, etc.)

[flashjh]

Originally Posted by diep

CUDAlucas is something *totally* different man.

True, but when I realized I couldn't max it out with TF and that it was slower than my 580s, I tried CuLu so I could but it to work, which didn't work either.

Trial factoring is using integers. The 590 is fast for integers.
CUDAlucas is a FFT. Of course that is no good plan at the 590 as
the FFT's require shared RAM and are lobotomized for double precision.

For double precision number crunching get Tesla's from Nvidia!


I've got a bunch of Tesla C2075's here they are fast for floating point!

I wish I could afford them, it right now - 580s are the next best thing for me.

You mixup trial factoring with DWT?

?

with TF you need fast cpu's and cpu cores and good software to resupply.
That thing can go what is it 800 million/s or so that GTX590?

How the hell to generate that with a quad core 2700 chip?

Do you mean for the 580s?

[diep]

Originally Posted by flashjh

Lots of info out there about the 590, but here is a good writeup. The paragraph right before the video talks about the clock speed. Though the 590 can be OCd a little, each onboard 580 still can't reach an individual 580s speed (at least not without flashing a new BIOS to the 590, etc.)

PCI-e goes till 375 watt so that's a hard limit. nvidia and amd already were going far over that if you looked at peak power usages of the cards.

a 580 on its own is what is it 250-280 watt or so?

edit: so if they would put 2 unmodified 580's at it they'd be consuming nearly 600 watt. A whopping 225 watt over spec of pci-e.
So they have to limit that a tad somehow :)

[diep]

Originally Posted by flashjh

I wish I could afford them, it right now - 580s are the next best thing for me.

I mailbombed Nvidia until they grew sick of me and donated me a bunch.

Note that the DWT job i'll give to them is an idle job, their core job is for the chess to parameter tune.

In this thread i post however with respect to the TF, because if you don't know how to factor, what purpose has life?

[kladner]

Originally Posted by diep

PCI-e goes till 375 watt so that's a hard limit. nvidia and amd already were going far over that if you looked at peak power usages of the cards.

a 580 on its own is what is it 250-280 watt or so?

edit: so if they would put 2 unmodified 580's at it they'd be consuming nearly 600 watt. A whopping 225 watt over spec of pci-e.
So they have to limit that a tad somehow :)

But these cards don't get all their power from PCIe. They have dual auxiliary power connections, as you must be aware. The lower clock on the 590, as well as throttling on at least the other high-end cards is driven by thermal issues. In addition, the nVidia driver (I think) watches for OCCT and FurMark, and starts switching rapidly from full clock speed to ~400 MHz to limit the core temp to ~70C.

[diep]

Originally Posted by kladner

But these cards don't get all their power from PCIe. They have dual auxiliary power connections, as you must be aware. The lower clock on the 590, as well as throttling on at least the other high-end cards is driven by thermal issues. In addition, the nVidia driver (I think) watches for OCCT and FurMark, and starts switching rapidly from full clock speed to ~400 MHz to limit the core temp to ~70C.

There is a design maximum that equipment that sits in the pci-e can max out at 375 watt.

75 watt from the pci-slot itself. 150 watt from the 8 pins thing and 150 watt from a 2nd 8 pins things. A total of 375 watt on paper.

In reality some cards at gpgpu go far beyond that.

AMD is by far worst there nowadays.

Nvidia really slamdunked with Fermi from my viewpoint seen.

Also really amazing is their K20X tesla for gpgpu purposes.

They are wonderful guys and hopefully keep intels price of the Xeon Phi 2013 (larrabee, knights ferry, corner and other medieval tales) a bit down, as without competition next generation Xeon Phi is gonna cost 10 times more than the current one.

[flashjh]

Originally Posted by diep

In this thread i post however with respect to the TF, because if you don't know how to factwhat purpose has life?

Exactly

[kracker]

Yeah, the 6or8 pin power connectors aren't the only source, there are low end gpu's that don't use them, they draw power from the MB, so it's not just all the axillary connectors.

But of course you need those for high end.

[flashjh]

Originally Posted by diep

Also really amazing is their K20X tesla for gpgpu purposes.

What I'd give for one of those, except $3000, of course ;)

[kracker]

Yeah! I was hoping to buy a few dozen.... Ah well.

[diep]

Originally Posted by kladner

But these cards don't get all their power from PCIe. They have dual auxiliary power connections, as you must be aware. The lower clock on the 590, as well as throttling on at least the other high-end cards is driven by thermal issues. In addition, the nVidia driver (I think) watches for OCCT and FurMark, and starts switching rapidly from full clock speed to ~400 MHz to limit the core temp to ~70C.

I'm no expert on gpu cooling, yet from my viewpoint the default cooler that's on those cards is a crap.

If you look on ebay you see some really good cooling designs to keep the temperature down of the videocards.

http://www.youtube.com/watch?v=DGeERB7BDHM

Of course there is no more warranty then on the gpu, but you have to sacrafice something for cool factoring speeds.