Here is my Super Cruncher 2009 PC:
- two Nehalem CPUs = 32 threads
- 4 x 2 GB DDR3 ECC RAM
- 1 TB Hard Disk Drive S-ATA-III
- one or two PCI Express 2.0 3D cards (1 GB / card)
- AGEIA PhysX
- two Ninja Scythe CPU coolers
- Tagan PSU (> 500 W)
- 64-bit Windows Vista Ultimate with Service Pack 1
- Maybe PCI/PCI Express FCPGA-card to crunch something (if available)

So, 32 threads to crunch 32 WUs at once. One or two 3D cards for Folding@home GPU app. AGEIA PhysX card for games and/or future BOINC apps(??).

Comments?

Can 64-bit Windows Vista Ultimate SP1 handle 2 CPUs (32 threads)?

Maybe I will break the RAC 700 with the Super Cruncher 2009. LOL.

---

Manned mission to Mars in 2019 Petition <-- Sign this, please.

I'd suggest the name "Super Crusher" instead of "Super Cruncher" :) And, yes, this SC will certainly break the RAC 700. lol

---

Physics capabilities will be added to most future generation GPUs (AMD/ATI has already claimed they will add physics processing before their merger), so the PhysX card will be unnecessary.

---

Just 1 thought.

Hope you have a nice bank manager. Your going to need a nice amout of money. Vista will run just as poorly on 32 as it does on 2.

Vista will run just as poorly on 32 as it does on 2.

Now that's an unfair comment. Vista runs beautifully on a standard dual core machine (albeit with a decent amount of RAM).

---

So, 32 threads to crunch 32 WUs at once.

How many cores is that? I assume there is only one FPU per core, not thread (similar to P4 w/ HT). So you probably won't get the full performance out of each thread.

Still....I'll take one! ...with OSX please. =;^)

---

Dublin, California
Team: SETI.USA

So, 32 threads to crunch 32 WUs at once.

How many cores is that? I assume there is only one FPU per core, not thread (similar to P4 w/ HT). So you probably won't get the full performance out of each thread.

Still....I'll take one! ...with OSX please. =;^)

Two Nehalems = 2 x 8 cores = 16 cores.

---

Manned mission to Mars in 2019 Petition <-- Sign this, please.

>750W seems closer to the mark, I suspect. Don't you think Vista will be closer to service pack 2 at that point?

Will Nehalem support HT? Can't remember

---

>750W seems closer to the mark, I suspect. Don't you think Vista will be closer to service pack 2 at that point?
Will Nehalem support HT? Can't remember

Ok, maybe 750 W.

Vista SP2? Hmm. Maybe?

Nehalem supports multi-threading. I do not know, is it called HyperThreading Technology anymore.

---

Manned mission to Mars in 2019 Petition <-- Sign this, please.

How many cores is that? I assume there is only one FPU per core,...

By 2009 I'd be hoping to be playing with at least 65536 cores all busily crunching away.

We can do some pretty good tricks with PCIe cards offering 256 cores and more already...

Happy crunchin',
Martin



(Hint: GPU arrays or FPGAs...)

---

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)

Here is my Super Cruncher 2009 PC:
- two Nehalem CPUs = 32 threads
- 4 x 2 GB DDR3 ECC RAM
- 1 TB Hard Disk Drive S-ATA-III
- one or two PCI Express 2.0 3D cards (1 GB / card)
- AGEIA PhysX
- two Ninja Scythe CPU coolers
- Tagan PSU (> 500 W)
- 64-bit Windows Vista Ultimate with Service Pack 1
- Maybe PCI/PCI Express FCPGA-card to crunch something (if available)

So, 32 threads to crunch 32 WUs at once. One or two 3D cards for Folding@home GPU app. AGEIA PhysX card for games and/or future BOINC apps(??).

Comments?

Can 64-bit Windows Vista Ultimate SP1 handle 2 CPUs (32 threads)?

Maybe I will break the RAC 700 with the Super Cruncher 2009. LOL.

One thing is certain about that machine Henri... I would pity your electric bill! :P

Seriously, the machine looks good but for one thing... Windoze. I would run some other OS on it, perhaps a nice 64-bit linux, or as someone else suggested, OSX (if possible, I wish Apple would release it for non-apple computers, have they?)

Physics capabilities will be added to most future generation GPUs (AMD/ATI has already claimed they will add physics processing before their merger), so the PhysX card will be unnecessary.

The SSE4 instruction set has already all you need to do physics.
An average game today use around 20% of the CPU to do physics, of a single core.
It is not obvious that you ll need more than a Core Core to do physics.


who?

Physics capabilities will be added to most future generation GPUs (AMD/ATI has already claimed they will add physics processing before their merger), so the PhysX card will be unnecessary.

The SSE4 instruction set has already all you need to do physics.
An average game today use around 20% of the CPU to do physics, of a single core.
It is not obvious that you ll need more than a Core Core to do physics.


who?

Yes, but the whole point is to unload that from the CPU and let it do something else..

But then I suppose you couldn't squeeze another Intel advert in ;)

---

Yes, but the whole point is to unload that from the CPU and let it do something else..

The whole point is to improve or move to all of:

• Faster (and cheaper) processors
  
• Greater parallelism to help beyond the faster sequential processing (CPU + GPU + FPGA/array processor or 'hardware configurable' processor)
  
• Better software and optimisation to make good use of the hardware
  
• Better algorithms to better use the computing system
  
• And all for lower power consumption!
  

Quite a good engineering challenge!

Happy crunchin',
Martin

---

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)

Or it was just another opportunity to slip another advert in.. ;)

---

Or it was just another opportunity to slip another advert in.. ;)

Speaking of plugs- Your team's website has a very good interview with Dave Orton of ATI/AMD.

---

Boinc Button Abuser In Training >My Shrubbers<

Or it was just another opportunity to slip another advert in.. ;)

Why?

Do you have a hang up with that happening?

I am aware there is not the same level of plugs for AMD, but so what?

---

It's good to be back amongst friends and colleagues

Or it was just another opportunity to slip another advert in.. ;)

I don't now about it being an advertisement... but I thought it was very informative. I'm glad to hear that the new SSE4 instruction set will support PhysX type instructions that will add even more functionality for gamers.

If SSE4 and PhysX-enabled GPUs can work in conjunction with one another, then I think we're in for some pretty cool looking games.

---

... but I thought it was very informative. I'm glad to hear that the new SSE4 instruction set will support PhysX type instructions that will add even more functionality for gamers.

You don't specifically need SSE4 for 'PhysX type' instructions.

Sure, the SSE4 SIMD instructions can help with performance for certain programming. There's still other 'high performance' ways of doing the calculations and on current CPUs.

It's all down to what you want to do and how fast.

My view is that the 'PhysX type' of add-on card has missed its window of opportunity. Current GPU cards and the latest CPUs already give more compute for less costs.

The only really successful add-on compute card that I know of is The GRAPE which is specialised for just the one calculation of "1/x". (OK, so it does it very fast!)


SSE4 is just the next (Intel) step in CPU design evolution. AMD (and others) have their fast features also.

If anything, AMD is championing lower cost and higher efficiency computing. (Just to balance with an AMD bit of Marketing! ;-) )

Happy crunchin',
Martin

---

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)

You don't specifically need SSE4 for 'PhysX type' instructions.

Oh, I'm very aware that it isn't a necessity, but you already illustrated my point in your post. I didn't mean to imply (nor do I think I did imply) that it was a requirement or prerequisite to decent performance. I merely wanted to state that, as you already put it, PhysX isn't necessary before it even began. The concept was great, but it can be easily implemented at a much cheaper cost in other ways.

It's simply that Intel recognized this and reacted accordingly. That's good business. AMD will probably license SSE4 and receive the same capability.

I'm not stating this as some sort of "huzzah for Intel", but more for the fact that we will all benefit from it eventually.

---