8-core monster crunchers

tng*
tng*
Joined: 6 Sep 05
Posts: 13
Credit: 66,244,577
RAC: 0

RE: RE: $4,000+? If you

Message 80715 in response to message 80708

Quote:
Quote:
$4,000+? If you mean US$, you can get a dual-quad crunchbox for far less.

Yes, I would agree with you if you build/spec the machine as more of a server than a general purpose/gaming/etc. computer.

I was mostly joking around a bit at intel's one and only offering in the 8-core "enthusiast" PC (not server) market. I personally think you'd have to be as nutty as a fruitcake to build a dual quad gaming rig since so few games can utilize more than 2 cores still, but some folks have more money than brains. ;)

As for "$500 for 8GB of FB-DIMMS?" I figued 2x4Gb products. One must leave room to expand afterall! ;) (again tongue in cheek since I think the board has 8 slots)

Ah. I was a little slow on the uptake there, and didn't catch on that this was tongue in cheek. IMHO, Skulltrail is a product which is superbly designed for its intended purpose: extracting money from the wallets of people with more money than brains.

tng*
tng*
Joined: 6 Sep 05
Posts: 13
Credit: 66,244,577
RAC: 0

RE: RE: RE: Dell is the

Message 80716 in response to message 80711

Quote:
Quote:
Quote:

Dell is the way to go....picked up several with v8 setups for 1000-1400 bucks

quiet and easy on electric


Which model?

Dell 690's ...all have v8 platforms and some Dell 490 have v8 motherboards. Have picked them off e-bay as cheap as 900.00 with 1 chip . I look for the E5320 @ 1.86 chips and do bsel mod to them to e5320 @ 2.33..or E5340 @ 2.4 mod to X5340 @ 3.00

bsel mod? A quick search shows this to be connecting a couple of pins to boost FSB. I've got a PowerEdge 1900 with E5320s that might benefit from this. Any links to detailed info, advice on gotchas?

tng*
tng*
Joined: 6 Sep 05
Posts: 13
Credit: 66,244,577
RAC: 0

RE: I'm a bit puzzled by

Message 80717 in response to message 80714

Quote:

I'm a bit puzzled by this Nehalem. According to Intel marketing, it's a brand new microarchitecture and as such the successor to the Core µ-architecture. But .... why??

Are the differences really that big that it's justified to claim it's something new?
And look how long the P6 architecure was around (all the way from Pentium Pro to the Pentium M or even the Yonah core). The core microarchitecure is still quite new and has since its introduction outclassed the competition, so why already proclaim it outdated?

CU
Bikeman

I'm no expert on processor architectures, but what I get from reading trade rags is that the big differences are an on-chip memory controller (AMD had this long ago, and it may have been very significant in the performance advantage they had over Intel for a while) and dedicated vs. shared cache (again, what AMD has had for some time). Killing the core microarchitecture this quickly looks to me to be a preemptive strike against AMD.

Paul D. Buck
Paul D. Buck
Joined: 17 Jan 05
Posts: 754
Credit: 5,385,205
RAC: 0

RE: I'm a bit puzzled by

Message 80718 in response to message 80714

Quote:
I'm a bit puzzled by this Nehalem. According to Intel marketing, it's a brand new microarchitecture and as such the successor to the Core µ-architecture. But .... why??


There are always two reasons for internal architecture changes ...

a) Marketing, and
b) The design has hit a design limitation that was not apparent when the initial architecture was first proposed.

Back in the OLD days ... when each computer was essentially hand crafted there was the debate between the single store and the dual store ... the von Neuman vs the Harvard architectures ...

In recent history, the external to the chip has always been von Neuman while the caches closest to the actual CPU has been Harvard ... giving the best of the two architectures...

In fact, for quite some time the actual instructions of the external world are run on a RISC type internal core... the problem again is that if the uOPs are poorly chosen or designed, well, eventually you run into a brick wall and have to start over ...

The problem peering into a company like Intel is that you have no idea of which is the real reason that the decision is made ... This could be pure marketing, or it could be an engineering necessity ...

Like the architecture where the user was only to plug in more CPUs and they would self-organize ... the trouble was when you added the 5th and later CPU the perfomance decreased as the CPUs spent all their time coordinating tasks that they got nothing done ... I can't for the life of me remember the chip number ... cool idea that did not work ...

Donald A. Tevault
Donald A. Tevault
Joined: 17 Feb 06
Posts: 439
Credit: 73,516,529
RAC: 0

RE: RE: I'm a bit puzzled

Message 80719 in response to message 80717

Quote:
Quote:

I'm a bit puzzled by this Nehalem. According to Intel marketing, it's a brand new microarchitecture and as such the successor to the Core µ-architecture. But .... why??

Are the differences really that big that it's justified to claim it's something new?
And look how long the P6 architecure was around (all the way from Pentium Pro to the Pentium M or even the Yonah core). The core microarchitecure is still quite new and has since its introduction outclassed the competition, so why already proclaim it outdated?

CU
Bikeman

I'm no expert on processor architectures, but what I get from reading trade rags is that the big differences are an on-chip memory controller (AMD had this long ago, and it may have been very significant in the performance advantage they had over Intel for a while) and dedicated vs. shared cache (again, what AMD has had for some time). Killing the core microarchitecture this quickly looks to me to be a preemptive strike against AMD.

With Nehalem, Intel will finally get rid of their old-fashioned front-side bus architecture, and replace it with a Direct Connect Architecture, which will be somewhat similar to AMD's HyperTransport. The current front-side bus design is somewhat of a bottleneck, and is the reason that Core 2 processors have to have such massive amounts of cache in order to have decent performance.

Also, Nehalem will be a native quad-core, whereas the current Core 2 Quad's are really just two dual cores placed in one mounting.

And yeah, this is a preemptive strike on AMD. The Intel guys were caught with their pants down once, and I don't think that want it to happen again.

Winterknight
Winterknight
Joined: 4 Jun 05
Posts: 482
Credit: 179,213,149
RAC: 117,985

RE: RE: RE: I'm a bit

Message 80720 in response to message 80719

Quote:
Quote:
Quote:

I'm a bit puzzled by this Nehalem. According to Intel marketing, it's a brand new microarchitecture and as such the successor to the Core µ-architecture. But .... why??

Are the differences really that big that it's justified to claim it's something new?
And look how long the P6 architecure was around (all the way from Pentium Pro to the Pentium M or even the Yonah core). The core microarchitecure is still quite new and has since its introduction outclassed the competition, so why already proclaim it outdated?

CU
Bikeman

I'm no expert on processor architectures, but what I get from reading trade rags is that the big differences are an on-chip memory controller (AMD had this long ago, and it may have been very significant in the performance advantage they had over Intel for a while) and dedicated vs. shared cache (again, what AMD has had for some time). Killing the core microarchitecture this quickly looks to me to be a preemptive strike against AMD.

With Nehalem, Intel will finally get rid of their old-fashioned front-side bus architecture, and replace it with a Direct Connect Architecture, which will be somewhat similar to AMD's HyperTransport. The current front-side bus design is somewhat of a bottleneck, and is the reason that Core 2 processors have to have such massive amounts of cache in order to have decent performance.

Also, Nehalem will be a native quad-core, whereas the current Core 2 Quad's are really just two dual cores placed in one mounting.

And yeah, this is a preemptive strike on AMD. The Intel guys were caught with their pants down once, and I don't think that want it to happen again.


The other thing being re-introduced with the Nehalem cpu's is HT, although maybe not called that. What they say is each core can process two threads. They have demostrated an 2 * 4 core computer each with 2 threads running, and it shows up in task manager as a 16 cpu computer.

(Ryle)
(Ryle)
Joined: 20 Mar 05
Posts: 8
Credit: 143,587,336
RAC: 9,280

Thanks for the clarification

Thanks for the clarification Paul.

It sure does look nice with all those cores in Task Manager, winterknight. But it would most likely necessitate a 64-bit OS. The 4 GB barrier would quickly be broken.

I can barely see the numbers in the screenshot, but it looks like 16 GB ram. Impressive.

ML1
ML1
Joined: 20 Feb 05
Posts: 339
Credit: 66,354,585
RAC: 0

Are there any PCIe CPU cards

Are there any PCIe CPU cards that could be added in to make up a (cost effective) super-cruncher?

Happy crunchin',
Martin

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

Paul D. Buck
Paul D. Buck
Joined: 17 Jan 05
Posts: 754
Credit: 5,385,205
RAC: 0

RE: Are there any PCIe CPU

Message 80723 in response to message 80722

Quote:

Are there any PCIe CPU cards that could be added in to make up a (cost effective) super-cruncher?

Happy crunchin',
Martin


Well, there was that discussion on one of the boards (I forget where, SaH?) about the use of the GPU on the upper end models of one of the MFGR's video cards ... For my part, I would think that that would make a killer screen saver if they piped the ongoing calculations to the screen ...

I can remember using o'scopes to watch address busses to monitor the "health" of the system if the traces started to look "funny" you could tell the computer was in trouble ...

Same way I could tell when one server was in dire straits by listening to the drive access patterns ...

The guy I worked for could never understand how I always knew ... :)

He did not believe "Magic" as an explanation either ... :)

Still, I think it would be interesting to see what it would look like as processing happened ...

vraa
vraa
Joined: 24 Feb 08
Posts: 3
Credit: 8,978,855
RAC: 0

Hello there Dual quad core

Message 80724 in response to message 80723

Hello there

Dual quad core boxes are fairly popular in the workstation ./ server arena

GPU's have been used in Folding@home for a while (as well as the playstation3)

Good evening

Quote:
Quote:

Are there any PCIe CPU cards that could be added in to make up a (cost effective) super-cruncher?

Happy crunchin',
Martin


Well, there was that discussion on one of the boards (I forget where, SaH?) about the use of the GPU on the upper end models of one of the MFGR's video cards ... For my part, I would think that that would make a killer screen saver if they piped the ongoing calculations to the screen ...

I can remember using o'scopes to watch address busses to monitor the "health" of the system if the traces started to look "funny" you could tell the computer was in trouble ...

Same way I could tell when one server was in dire straits by listening to the drive access patterns ...

The guy I worked for could never understand how I always knew ... :)

He did not believe "Magic" as an explanation either ... :)

Still, I think it would be interesting to see what it would look like as processing happened ...


Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.