lesser-equity

Magazine
Go Back   Computer Juice > Computer Hardware > CPUs, Motherboards & RAM


Register


Reply
 
Thread Tools
  #1  
Old 15th Feb 2009, 05:01
New Member Group
 
please help me on this guys , how many instruction can new processor execute at the same time

like Pentium 4 has 126 Reorder buffer entries , so
how many Pentium D, Intel Core 2 Duo , Core i7 , Amd Phenom has

AND how it is calculated
  #2  
Old 15th Feb 2009, 05:37
Donor Group
 
You can use a unit of time inside a processor called the multiplied clock cycle. A single instruction takes several of these. The processor might be optimized to have pre-fetched the material needed for the next instruction but even if there's an overlap an instruction involves the loading of one or more data elements, an execution and a save, each bit of which takes one or more units of time. The very most a processor can handle at once is one instruction.

A processing chip might have several processors designed into it and a board might have many processor chips. All of those are running in parallel. A high-end graphics card might have 250 parallel processors, the i7 has 4 at the moment, Pentium D has 1, Phenom has 2 or 3 or 4 depending on the model. Each processor is executing at most one instruction at one instant though it may be working on preparing or storing other instruction data.

http://en.wikipedia.org/wiki/Cycles_Per_Instruction might help.

Reorder buffer entries just manage the fetching and carrying of the data and addresses, they optimize the processor's ability to get through one instruction in the fewest time units. It's true that the entire chip is working on several instructions at once but each instruction is still only executed one at a time in the correct order for a given processor. The only meaningful question is how many processors can work on different data sets in parallel and those are the numbers I instanced. The point of threads is that the result of one isn't allowed to interfere with the data stream of another and consequently the processing solution is the same regardless of the order the threads are performed in.

Breaking a task down into parallel threads is a headache for a programmer and a complicated automated business for an parallel-thread-optimising compiler.
__________________

My System: Tim

Processor(s):
Athlon 64 3500+
Motherboard:
Asus A8N-VM CSM
RAM Memory:
Corsair PC3200 CL2 DDR-400 2GB
Graphics Card(s):
nVidia Geforce 6600 512MB
Sound Card:
Cherry RS 6000 M keyboard
Hard Drive(s):
Barracuda.7+ 2x200GB 58MB/s sustain
Optical Drive(s):
Samsung DVD-ROM TS-H352
Case / PSU:
Thermaltake Soprano
Cooling:
Stock
Network / Internet:
Telewest 2x20Mb/sec
Monitor(s):
SXGA flat panel
Operating System(s):
Slackware (2.6.27.7) (Fluxbox)
  #3  
Old 15th Feb 2009, 10:47
New Member Group
 
look at this and tell me how the new processor fit in


CPU
Instructions executed
at the same time
AMD K6-II
24
Intel Pentium III
40
AMD Athlon
72
Intel Pentium 4
(first generation)
126


what this table shows , this what i have to know about the new processors , that i have mentioned
  #4  
Old 15th Feb 2009, 20:30
Donor Group
 
http://www.firingsquad.com/hardware/...mance_preview/ has a good discussion of the i7 architecture and explains why trying to simplistically multiply pipeline length by the number of available pipeline threads isn't meaningful with today's processors. Those tricks are in the museum compared with the last five years. The i7 has 8 threads but what's done with them doesn't let you just multiply by 8.

The Pentium 4 pushed pipeline length to 31 micro-instructions because nobody was predicting the cap in processor speeds caused by excessive heat generation much past 3GHz. What they've done is actually dropped pipeline lengths - for example, in the core duo, to a 14-stage pipeline - and introduced words like hyperthreading, Quick Path Interconnect, Turbo Mode to describe alternate ways of pushing instructions through the micro-instruction expansion around the fetch:decode:execute:write cycles.

The other reason you can't easily just multiply up threads by pipeline length to get the numbers in that table for new processor designs is that some processors now have different-length pipelines to deal with a particular set of operations. Integer and floating point instructions are broken onto different length threads to optimise their throughput, for example.

Your table's repeated on http://www.karbosguide.com/books/pca.../chapter30.htm and that even has a section warning of these approaching developments:
The Pentium 4’s architecture must therefore be seen from a longer-term perspective. Intel expects to be able to scale up the design to work at clock frequencies of up to 5-10 GHz.
Well, it didn't happen did it. That's why the extrapolation of your table isn't a good description of the way a modern processor like i7 or Phenom works.
Reply

Register
Thread Tools




Arabic Bulgarian Chinese (Simplified) Chinese (Traditional) Croatian Czech Danish Dutch English Finnish French German Greek Hebrew Hungarian Italian Japanese Korean Latvian Lithuanian Norwegian Polish Portuguese Romanian Russian Serbian Slovak Spanish Swedish Thai Turkish Ukrainian

Copyright ©2006 - 2009 Computer Juice.

Powered by vBulletin® Copyright ©2000 - 2009 Jelsoft Enterprises Ltd. SEO by vBSEO ©2009, Crawlability, Inc.