30 day free trial of Pro/E!

4 TeraFlops on Your Desktop. The nVidia Tesla SuperComputer. How Fast Will Your Computer Be?

by Josh on February 5, 2009 · View Comments

nvidia-tesla-3d-cadAre you ready to elevate your heart rate a couple notches about extremely powerful computer tech? How does the word ‘Supercomputer’ make ya feel?

While the nVidia Personal SuperComputer is bound to have your models streaking across the screen, don’t shove your old workstations down the stairs just yet… Ok, go ahead, but you’re gonna need some cash and a new jacket that says, ‘Back OFF, I’m supercomputin’.”


Watercolor Memories
Do you remember when 10GB was more than enough storage? Or when a couple Gig of memory was HUGE? Technology that went perfect with a steamy cup of hot coffee and hours of 3D CAD modeling… until those CAD models became so unmanageable that we shook our cubicle walls and angered the IT department with constant request for upgrades.

That’s pretty much the standard mode of operation. Souped-up systems that ‘make everything easier’ are always gonna cost more. However, the big difference is, the systems that use to take the warehouse space of underground government facilities are now small enough to move in next to your China Cabinet. They’re calling them ‘Desktop Supercomputers’ and the ones coming out in the present are sure to be our normal ol’ laptops and workstations in the future.

That nVidia SuperComputer Thingy
While it’s not the first or only supercomputer on the market, the nVidia Personal SuperComputer is coming in strong with a configuration and price that makes it one of the most powerful, yet affordable system available.

It uses nVidia’s CUDA parallel computing architecture powered by up to 960 parallel processing cores, which allows it to achieve a performance up to 250 times faster than standard PCs. This all basically means instead of having a bunch of computer linked together for more processing power, you’ll just need this one. No more rooms dedicated to housing racks of CPUs.

The Tesla options. via <a href=

Example of Complete 4 Tesla C1060 System Configuration
Here’s a quick run-down on the standard specs. The price is approximate as different vendors will have different configurations.

4 Tesla C1060 Configuration
Motherboard Foxconn Destroyer nForce 780a
Tesla GPUs 4x Tesla C1060
CPU AMD Phenom 9850 2.5 GHz quad-core
Memory 16 GB (4x 4GB) DDR2 DIMMs from G.Skill
Power Supply Coolmax CUQ-1350B 1350W
Case Lian Li PC-P80
Hard drive 640 GB
DVD drive DVD burner
CPU fan, heat sink For the AMD Phenom
Total Retail Price ~$ 8500*

* Prices obtained from online retailers in the US

For a beefy machine the price doesn’t seem all that bad. I recall people building computers with less for the same price or more.

It’s interesting to think how fast, how small, how powerful our computers will be. On top of that, it’s interesting to think how much software and 3D apps will advance, since they always seems to stay slightly ahead of what the mainstream hardware is capable of handling. What do you think?

Interesting Fact:
As of November 2008, the most powerful supercomputer is the Cray XT Jaguar, at the U.S. Energy Department’s Oak Ridge National Laboratory, with 1.64 Petaflops and 362 Terabytes of memory, more than 50% faster than IBM’s previous record. – UPI

nVidia via DeskEng.com [BIG hat tip to Mike O'Sullivan]

{ 17 comments }

bigmikeo February 5, 2009 at 10:30 am

Told my IT guy I'll wait on the upgrade on my current system so I could get this. He had that “your kidding” blank stare on his face.

Sonicson February 5, 2009 at 11:04 am

I am hoping this beast would be the last desktop you'd ever have to buy!

MadMatt32171 February 5, 2009 at 12:21 pm

But can it run Crysis?

Josh M February 5, 2009 at 12:25 pm

It would be… until the next day when the hardrive storage, RAM and processing power was doubled for cheaper. ahhhy, technology.

Josh M February 5, 2009 at 12:29 pm

yeh, the true test. i bet it gets all pixely when you move the mouse real fast like.

Bradley Grzesiak February 5, 2009 at 12:35 pm

Imagine a beowulf cluster of…. oh, wait. Wrong website. My bad.

blargh February 5, 2009 at 5:25 pm

What about your graphics card where does that fit in? You aren't going to settle for an onborad geforce for CAD are you?

Warren Hadley February 5, 2009 at 6:25 pm

The Tesla is certainly cool. If Nvidia actively supports OpenCL it will be even cooler (I hope Nvidia moves beyond CUDA!).
Check out the demo on YouTube to see what OpenCL really means (a lot more tangible than Terraflop numbers). The demo is for a multi–core CPU, so only a fraction of the potential of a GPU. But still pretty impressive. Now imagine this on a dedicated many-core GPU (which is what AMD and Nvidia produce).

So if CAD vendors start computational coding in OpenCL, a lot of the complex calculations can get offloaded in parallel to the GPU or actually use multi-core CPUs (so my quad core machine actually does more than just get hot!)

http://www.youtube.com/watch?v=mcU89Td53Gg

Josh M February 6, 2009 at 6:47 am

Warren, thanks for bringing this up. not to many know about OpenCL ( I didn't until recently) I think there's a lot of value in the framework, especially for parallel processing potential (ha! PPP) anyway, I'd hope that SW development is looking into variant open frameworks and how well process CAD data in the future. We can only push Moore's Law so far. Still I suspect, with all the hardware/software technologies that are coming out lately that there's that technology is going to shift drastically. Many vendors will have to adapt quick and those already aware of the options will be the ones that survive… in my opinion :)

Thanks for the comment.

ckeen February 6, 2009 at 7:35 am

Being named after Nikola Tesla, the product bears the mark of high-quality with equal measure of insanity. I am due for an upgrade soon. Check Velocity to configure one. The 3 teraflop unit is actually pretty cheap starting around 4k.

ckeen February 6, 2009 at 7:39 am

haha nevermind. The 4k price was without the tesla unit. Still buying one.

Steve Bell April 7, 2009 at 4:12 pm

Hmmm… I remember when a couple of meg of RAM was huge…

Steve Bell
http://www.archiform3d.com

Josh M April 8, 2009 at 4:47 pm

I'm running 2GB RAM and my computer crawls through a lot of models. That use to be the standard, now it's sub-par. I'm guessing some are in the same boat and that there will be a big leap to 8GB+ when Windows7 comes out/Vista goes away :) however you look at it.

paintballer123 March 1, 2010 at 7:57 pm

How many motherboards in totally are there?

Peter March 1, 2010 at 7:58 pm

How many motherboards are there.

BIGDOG76 March 19, 2010 at 3:36 am

There is one motherboard. There are only a few that can handle it. Asus just released one (p7f7?) with usb3 sata3 and tesla support. Nvidia is releasing the 20 series in May '10. You may want to wait a couple months 'til there out and some more mobos are available. I am waiting 'til there's a bit of a selection with usb3 sata3 and ieee1394b available. You need a mobo that has 4 pcie x16 double wide slots and a case that will fit 4 double cards (10 expansion slots). You can also build with 1+ TESLAS, it just won't be a certified supercomputer. If you're working with 3d modeling/animation/rendering with an app that supports cuda it should speed things up quite a bit

@warrenhadley Your post was a year ago, but nvidia does currently support OCL. nvidia and OCL

BIGDOG76 March 19, 2010 at 8:36 am

There is one motherboard. There are only a few that can handle it. Asus just released one (p7f7?) with usb3 sata3 and tesla support. Nvidia is releasing the 20 series in May '10. You may want to wait a couple months 'til there out and some more mobos are available. I am waiting 'til there's a bit of a selection with usb3 sata3 and ieee1394b available. You need a mobo that has 4 pcie x16 double wide slots and a case that will fit 4 double cards (10 expansion slots). You can also build with 1+ TESLAS, it just won't be a certified supercomputer. If you're working with 3d modeling/animation/rendering with an app that supports cuda it should speed things up quite a bit

@warrenhadley Your post was a year ago, but nvidia does currently support OCL. nvidia and OCL

Comments on this entry are closed.

{ 5 trackbacks }

blog comments powered by Disqus