|
I borrow 240 streaming processors from my nVidia... worse, I regularly borrow two sets of them. My work machine has 8 cores, I have access to 64 or more depending on if I activate a grid. Still, when I pull in the nVidia CPUs I can outdo the older 64 node super computer.
|
|
|
|
|
LOL I just knew there's be a "more than 100" answer from you.
|
|
|
|
|
just wait until I get my 4 GPU Tesla... I am already doing fluid dynamics and physics calculations at speeds that exceed any super computer we have ever owned. And I only have one card at work, and two at home. When I get four?
|
|
|
|
|
I bet. Sounds like a fun rig - are you going to put a CB and go-faster stripes on it?
|
|
|
|
|
Anna-Jayne Metcalfe wrote: are you going to put a CB and go-faster stripes on it?
oh no... but I might have to calculate the RF communication distance of the CB through rough terrain as well as the ambient reflection of the racing stripes.... does that count?
|
|
|
|
|
I'm sure it does.
|
|
|
|
|
The moment I saw the poll in the Insider, I *knew* you'd be boasting here
Cheers,
Vıkram.
I've never ever worked anywhere where there has not been someone who given the choice I would not work with again. It's a job, you do your work, put up with the people you don't like, accept there are probably people there that don't like you a lot, and look forward to the weekends.
- Josh Gray.
|
|
|
|
|
That's not boasting... that is just "drowning my sorrows" with a Tesla. I find it interesting that the last two home machines have "embarassed" work into upgrading my machines at work. It is kind-of sad that it is remarkably that easy. My bosses like to have bragging rights on machines: origin 2000, CM5, quad single core, 64 node Enterprise Linux box, dual quads, etc. When an employee owns a computer faster than the company... they get a bigger computer.
Of course that means they don't have the R&D risk associated with new machines, they simpy one-up my designs directly. or try. The last one (dual quad core xeon, single nv280 is doing great, but design wise it is functioning in all respects save one, at half the speed of my home system. My RAIDs are 3 times faster, my memory bench twice as fast, IO almost twice as fast. They have 8 cores vs my 4 on the main chip, but even so, my 4 at 3.2ghz seem to outbench their 8 at 2.8ghz because of the slower BUS and memory. But when I demonstrate CUDA processing... when you can out process the sum total of all the standard CPUs in the building BEFORE you kick in the second card...
but... of course I am drowning sorrows this week.... so pass the Tesla to me... I'm ready for it.
|
|
|
|
|
Ah, you still have to pull in the results from NVidia into main RAM and then to CPU if you going to do anything other than weak C++ more like C-style mess CUDA compiler is.. have you measured the penalty there? It should be significant and I believe it should be a part of any benchmark or results (at least half-way
|
|
|
|
|
In a world were so many cores exist, we should be thinking seriously into developing more applications that take advantage of it._
|
|
|
|
|
that is why parallel coders exist.... and why other programmers scream that "thinking" in parallel (stretching their brain) is impossible.... one wants to grow, the other doesn't.
Now if only I could have a direct brain hookup.
|
|
|
|
|
I want one of those interfaces that Hamilton described, wired direct into the brain.....
Never underestimate the power of human stupidity
RAH
|
|
|
|