Prime95 on a 5.0Ghz Processor
 

Apparently IBM has a chip that can deliver close to the 5.0Ghz realm... How would Prime95 run on such a beast- of course putting one instance per core. I had no idea that such chips existed. I wonder what one of these would cost?

See this link:

[url]http://www.theregister.co.uk/2008/04/08/ibm_595_water/[/url]

[quote=Primeinator;176428]Apparently IBM has a chip that can deliver close to the 5.0Ghz realm... How would Prime95 run on such a beast- of course putting one instance per core. I had no idea that such chips existed. I wonder what one of these would cost?

See this link:

[URL]http://www.theregister.co.uk/2008/04/08/ibm_595_water/[/URL][/quote]

The PowerPc has no MMX / SSE(2) as far as I know. It would be slower ... you can overclock an Intel 5200 Dual Core up to 4,3 Ghz (with two cores). Should be running very nice (need probably around one week for an actual exponent) ...

I'll pretend like I just understood that first part of what you said. Is there someway to build a chip that supports features that would run an LL at 5.0Ghz?

[quote=Primeinator;176430]I'll pretend like I just understood that first part of what you said. Is there someway to build a chip that supports features that would run an LL at 5.0Ghz?[/quote]

It would be interesting when the graphic card companies would create gfx with double precision in the gpu. Then you could calculate on your high tech graphics card.

[QUOTE=joblack;176434]It would be interesting when the graphic card companies would create gfx with double precision in the gpu. Then you could calculate on your high tech graphics card.[/QUOTE]
AFAIK newer graphic cards like NVIDIA GeForce GTX 280 are able to calculate double precision.

[quote=rudi_m;176439]AFAIK newer graphic cards like NVIDIA GeForce GTX 280 are able to calculate double precision.[/quote]

They are?

Hmmm ... that would be the perfect machine for a new prime95 version. Could decrease the calculation time by the factor 10 or more.

By a factor of 10? That is rather impressive... Or do you mean 10%?

Based on data here:
[url]http://blogs.sun.com/bmseer/entry/linpack_hpc_benchmark_sun_sparc[/url]

It seems each core of the dual core power6 processor is capable of 4 FLOPs per cycle, this means 4 FLOPs * 2 cores * 5 GHz = 40 GFLOP/s theoretical peak for a dual core power 6 CPU at 5 GHz.

The i7 CPU is also capable of 4 FLOPs per cycle, but it's a quad core CPU and is sold at a maximum clock of 3.33 GHz, therefore it is capable of 4 FLOPs * 4 cores * 3.33 GHz = 53.3 GFLOP/s theoretical peak.

Of course, not all of the theoretical performance can be realised for various reasons, but the i7 does have a leg up with hyperthreading. While one thread is sitting idle waiting for data, the other can step in and use up some of the clocks available that would otherwise be wasted.

So while 40 GFLOP/s peak is nice, 53.3 GFLOP/s with HT is better.

Wow, very nice. I am tempted to get one but I have neither the expendable resources at the moment nor will my laptop have sufficient cooling power. Plus, I do not think my motherboard supports it.

[quote=Primeinator;176441]By a factor of 10? That is rather impressive... Or do you mean 10%?[/quote]

It was the factor 10 for md5 password checking. I'm not sure how well you can parallelize the work of prime95. That could answer the programmer of the program.

Hello,
This is my first post here and hopefully I shall not regret it (you all seem like true Ladies and Gentleman).

I agree with the initial statement that since this 5GHz chip does not have MMX and SSE etc, it would calculate less efficiently (probably a hell of a lot slower), but if a version of Prime95 was written and compiled for these chips and any special features it has, it would run just as efficiently.

Take for example Linux running on a PS3. It is a specially "tweaked" version for that architecture which is totally different to anything us "IBM Compatible" people are used to, yet it runs just as well. I know is has three cores etc, but the point is the core code is the same with just a few changes to make it compatible with that architecture. The compiler does most of the work when changing chip types.

Another example is the older PocketPC/Windows Mobile devices (most new ones run ARM based cores). They had three or four different chip architectures yet the compiler would take a basic program and compile it for all devices, unless you used specific features.

Now if I am wrong in this please correct me. I probably have made an error here somewhere.

--
Millsie