• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

RE: Conroe going from 3 Complex to 3 Simple plus one Complex...

thestain

Senior member
Hi,

I have been reading up on the new Conroe and noticed it has 3 simple decoders (instructors?) and one complex one... please forgive my simple use of these terms.. and also the ability to add an extra set of combined smaller instructions to the cycle.

I am guessing that having the ability to generate four or five instructions in a single cycle is going to boost performance. It appears Intel has went with more simple instructors in order to fit more into a cycle.

Can we expect something like this from AMD and.. in the current benchmarks we see where Cache has really helped the Intel Pentium D's with the smaller and simpler applicaton benchmarks.. and appears to help even more with the 4 GB cache Conroe, how much will the extra, although simple instructors boost performance on these benchmarks?

Seems like a formidable one-two benchmarking wise from Intel..

And.. if only one complex instructor per four... and with the large Cache, what kind of performanc hit can Conroe be expected to take as the testing shifts from simple and small applications to larger and more complex??

Thanks,

The Stain
 
Originally posted by: thestain
Hi,

I have been reading up on the new Conroe and noticed it has 3 simple decoders (instructors?) and one complex one... please forgive my simple use of these terms.. and also the ability to add an extra set of combined smaller instructions to the cycle.

I am guessing that having the ability to generate four or five instructions in a single cycle is going to boost performance. It appears Intel has went with more simple instructors in order to fit more into a cycle.

Can we expect something like this from AMD and.. in the current benchmarks we see where Cache has really helped the Intel Pentium D's with the smaller and simpler applicaton benchmarks.. and appears to help even more with the 4 GB cache Conroe, how much will the extra, although simple instructors boost performance on these benchmarks?

Seems like a formidable one-two benchmarking wise from Intel..

And.. if only one complex instructor per four... and with the large Cache, what kind of performanc hit can Conroe be expected to take as the testing shifts from simple and small applications to larger and more complex??

Thanks,

The Stain


Well, lets see where to begin. First of all this belongs in the CPU/Processor section, not the highly technical section

The P6 had 2 simple decoder and 1 complex decoder, Croe just adds another simmple decoder. There is really no need for another comple decoder since the vast majority of instrction can be decoded by the 'simple' decoders. Also, keep in mind that while the decode bandwidth might be 4 instuction/cycle your gonna be lucky to actually execute an average of 1 instruction/cycle (maybe Core is a little better, but still nowhere even clsoe to 4). The micro/macro ops fusion is a good thing because it allows some instuctions to be combined which is helpfull. Also, Conroe has been benchmarked on a very wide range of benchmarks and has shown to be superior in the vast majority of them. Also, don't think that a 'simple' program and 'simple' instuction are related, larger programms jsut have more instuction, not more complex ones. Also, Conroes cache (4MB, not GB) is certainlly a factor in its efficiency, but don't forget that Presler also as 4MB of cache and still stinks, not to mention that doubling the chace on may processors has only yields moderate (single digit percents) increases in performance. Don't believe this silly "the whole program fits in cache" nonsense, thats completely untrue.
 
Thanks.. will move to appropriate forum.. sorry about my typos... I meant 4 MB..and I will try to stay away from some of those arguments I have heard. So.. single digit help with larger cache in your opinion.. maybe a little help with the extra decoder?, but then.. ??

I have only seen some pretty controlled test.. impressive, but for smaller apps test results so far.. if you come accross the larger and more complex apps, please share them here..

Thanks again.
 
Back
Top