Sounds like the cards are "stuck" in low power mode. Is BOINC crunching happily on all GPU cores - ie does BOINC see 4 GPUs?

I'd also close CCC and use MSI Afterburner.

I have a 5970 and a single 4870 crunching happily away and only see them downclock when they run out of work.

Still.. even 'sadder'... all that time lost.. all the energy spent over 17 years... completely blown away in less than a week from one, single, solitary GPU.

My mobo is an ASUS P5Q3 Deluxe/WiFi-AP@n motherboard. Quote from a review is "Another remarkable thing is three PCI Express x16 slots. However, don’t expect any wonders here: if you use two graphics cards, the first two slots will share 16 PCI Express 2.0 lanes. The third PCI Express x16 slot supports PCI Express x4 logical interface and is connected to the chipset South Bridge.

I was under the impression that a 5970 would be doulbe the speed of a 4870x2, turns out that one GPU in a 5970 is roughly 3 times as fast as a 4870 GPU, so overall that should give out a rough 6x increase.

Based on my results I'm seeing each OC'd GPU on the 5970 doing just over twice the throughput of my 4870. See here.

Power usage with 5970 @ 876/1000MHz & 4870 @ 801/900MHz on MW with 1 WU per GPU and Q9450 @ 2.66GHz running Aqua = 608W to 625W.

Idle = 245W with the 5970 fan at 40%.

At full load before the card has warmed up, so the fan is still at 40%, the power draw was 580W.

As the card temps leveled out at 67°C & 76°C, my fan profile has the fan running at 56% and the power draw is back to 608 to 625W. Glad my PSU is 850W.

Wow, the power bill is going to be BIG, BIG!

With that in mind, I experimented a bit yesterday. I dropped the boxes down to running 1 WU at a time (n1)and set w1.01 (w1.15 on the old P4). Hey presto, they have all been running happy and smooth.

Bad news, it's not about insufficient system memory. I ran four long 4x MW WUs with 8GB of system RAM, only one GPU installed and all unnecessary services and processes disabled on my Win2003 Server Enterprise Edition with PAE enabled and it still finished with "fitness: 1.#QNAN000000000000000". I ran them with network disabled so don't be surprised that they have not been reported, yet.

I have a screen shot of the memory usage here.

My next idea was to try an optimized app but there is no opt app for CUDA. Why? Is the stock CUDA app so well optimized that it's not necessary?

Any stats on mortality rates between vid cards and cpu's? Pure speculation here but I would think vid cards would fail more frequently since gpu's run warmer while cpu heatsinks tend to be beefier. Or maybe gpu's are simply designed to run hot for sustained periods without adverse effects?

Another question, do the admins participate in the forum or are we just talking to ourselves?

I have a 4850 and a 4870 in the same box and will be staying that way for a while unless I win the lottery....

But I will never purchase cards in batch lots so I will always have an asymmetric installation as I upgrade one card etc.....

Then best complain so I am not the only voice ... because UCB is going to take silence as consent ... and will never consider reality because I don't know of anyone that can always afford to upgrade in full sets.

Even if they can, like on my next upgrade cycle I have two GTX260s and if they are both running and I get a MB with three PCI-e slots ... will I be able to find GTX260s? By then the fuller line of GTX300 cards may be filling the niches and I am supposed to junk two perfectly good GPUs?

Anyway, silence is consent ...

@ Ironwork16

Actually there is no driver for the 58x0 and WinXp from the ATI/AMD website. But with my Asus 5870 i got a cd with a driver for WinXp. This driver works but i believe it is less stable than the driver for Vista/Win7. Crunching unter Win7 is fun with the 5870...

That's why other projects are "whining" about the overpay at MW. For most of them it's just impossible to utilize the hardware to this extent. It's not just about being lazy programmers - the algorithm and the problem itself don't allow it. They could never achieve the same Flop/s even if they made an ATI app.

MrS

That is very true.
The MW algorithm is really perfect for a GPU. Take vast amounts of parallelism (millions of threads), no branching (except you want to call a loop with a counter checked each iteration a branch), a very compute intense algorithm with only a few memory accesses, minimal communication between the threads (the values are just added in the end), and what you get is virtually the peak performance of a given GPU for the instruction mix of the algorithm. It's not all about multiply-adds, so you won't get exactly peak performance. But the v0.20 has cut all the overhead down to a minimum so you really arrive within 10% of what is theoretically possible with the algorithm's instruction mix. That's better than any current CPU achieves, even relative to its peak performance.

And that will continue to scale, the new ATI HD5870 should easily double the performance of a HD4890 at Milkyway. And when the next nvidia generation arrives, I'm quite sure it will do much more than to double the DP performance of a GTX285 ;)

You're right, I was talking about a 4870 and it's maximum dp performance is indeed 240 GFlops at 750 MHz. Mine runs at 800 MHz (256 GFlops) and achieves ~190 GFlops at MW. That's a really really good optimization done by CP, so even achieving something close to these numbers is challenging.