• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Barcies dual FPU

manveruppd

manveruppd

manveruppd

Associate
Joined
24 Feb 2004
Posts
1,083
Location
Leeds/Cyprus
A friend of mine's looking to build himself a new PC, was gonna spec him a C2D rig, but then recently I read about the dual FPUs in the Barcelona cores. Thing is, he mainly wants it to do some hardcore Mathematica calculations (not sure what kind, but fairly processor-intensive he says). Is he likely to see a massive improvement from a Barcelona core than from a Conroe? Shall I get him to wait a bit more for them to become available?
 
manveruppd said:
A friend of mine's looking to build himself a new PC, was gonna spec him a C2D rig, but then recently I read about the dual FPUs in the Barcelona cores. Thing is, he mainly wants it to do some hardcore Mathematica calculations (not sure what kind, but fairly processor-intensive he says). Is he likely to see a massive improvement from a Barcelona core than from a Conroe? Shall I get him to wait a bit more for them to become available?
Click to expand...

Benchmarks from the first 'Barcy' servers dont indicated enough speed increase to confirm the rumor that they would have dual FPU's*. Seems like its just 1 fpu per core as with the K8 and C2D's. Performance is 'good' but nothing amazing.

The desktop chips may be optimized differently, and there could still be some surprises when the chip's officially released, but it doesnt look like it will be earth shattering.

Just get a Conroe, dual, or quad if the software is able to make use of the additional cores.

*Or it could be that current software only expects 1 FPU per core, and hasnt been optimized to send multiple parallel floating point instructions to the cpu.
 
Yeah, there's always something better round the corner though...

He says he's mostly using Mathematica. I don't know which version he has, or whether it supports multiple cores. I saw those benchies too and I don't think the tests they ran would benefit from dual FPUs. The only way to quantify benefits like that would be to run dual instances of SuperPi I reckon. Also, do we know that applications will need driver support or patches to make use of the dual FPUs? I hope they won't tbh.

I would have to wait for the desktop chips obviously, which I think aren't due till November so it might simply not be worth the wait.
 
manveruppd said:
Yeah, there's always something better round the corner though...

He says he's mostly using Mathematica. I don't know which version he has, or whether it supports multiple cores. I saw those benchies too and I don't think the tests they ran would benefit from dual FPUs. The only way to quantify benefits like that would be to run dual instances of SuperPi I reckon. Also, do we know that applications will need driver support or patches to make use of the dual FPUs? I hope they won't tbh.

I would have to wait for the desktop chips obviously, which I think aren't due till November so it might simply not be worth the wait.
Click to expand...

Dual superpi wouldnt help. It would send one copy to each core. All the dual, and quad core cpus already have FPU per core... Barcy was supposed to have 2 fpus' per core (IE 8 FPU's). And I believe most software would need recompiling to benifit from that. It was the same when P4 was new, as they have two integer units... but so much software couldnt benifit from the second one until it was recompiled.

Still, I would have thought there would have been some sign of improvement even with 'current' software.
 
Oops, yes, you'd need octuple then!:D

I guess before making a final decision he needs to find out whether his software would even benefit from it. I presume there's a forum for Mathematica users somewhere, I'll tell him to go on there and ask, but tbh if he'll have to stay on his current 1GHz Athlon while he waits another 2 months for the Phenom to appear I think he'll just choose to go with a Conroe.
 
Apparently, there was an issue with all the chips bar the 2.5GHz one shipped to Anand - in the B1/BA spin of the silicon, something went wrong with a part of the FPU meaning it wasn't working properly.

In the B2 spin (i.e. the 2.5GHz chip Anand was playing with) this was apparently fixed.

No idea what this constitutes performance-wise but something to bear in mind when looking at such benchmarks.
 
You must log in or register to reply here.
Back
Top Bottom