IPC/MIPS

[PsharkJF]

Trying to find some reliable listings of the IPC/MIPS of a (533) Northwood P4 against the A64/A64 FX. Anyone care to help?

 

[PsharkJF]

Once, twice, three times a bumperr......

 

[Ronin]

I can do a LGA775 3.2GHz Prescott with a FX53, if you'd like.

 

[PsharkJF]

That'll be fine, I'll just take it with a grain of salt that the Presscott will be faster than my 2.53/533 Northwood.

I'm trying to see which AMD chip I should be waiting for to get one that's exactly twice as fast as my current one. 😛

 

[Ronin]

Alrighty. Numbers coming up sometime today. 🙂

 

[Goi]

Remember that IPC/MIPS is heavily dependent on the instruction mix, so it doesn't say a lot and isn't comparable unless running the exact same program

 

[whorush]

yeah i would just check the benches for the stuff u do, when the benches double, then buy.

 

[PsharkJF]

Originally posted by: Goi
Remember that IPC/MIPS is heavily dependent on the instruction mix, so it doesn't say a lot and isn't comparable unless running the exact same program

Well, scientifically, if we are running two different programs, how can we isolate the variable of number of executed instructions per second, given different instructions executed and differing number of overall instructions executed? I figured it was a given we'd be trying out the same programs and comparing them one against the other. Then, using an array of several differing programs [hopefully] we can draw an overall conclusion.

yeah i would just check the benches for the stuff u do, when the benches double, then buy.

I agree, but that can be due to many other things - WD raptors could make a pretty big impact on some benchmarks, having an integrated memory controller helps a bunch, etc. There are a lot of variables in the equation. I understand that getting a new CPU and motherboard does include these variables, but I am choosing not to bring those into the equation at this point.

 

[Goi]

Originally posted by: PsharkJF

Originally posted by: Goi
Remember that IPC/MIPS is heavily dependent on the instruction mix, so it doesn't say a lot and isn't comparable unless running the exact same program

Well, scientifically, if we are running two different programs, how can we isolate the variable of number of executed instructions per second, given different instructions executed and differing number of overall instructions executed? I figured it was a given we'd be trying out the same programs and comparing them one against the other. Then, using an array of several differing programs [hopefully] we can draw an overall conclusion.

True, but once you start accepting numbers from people all over this forum without knowing how they're obtaining their numbers, it becomes hard to standardize the procedure. Who knows if Ronin is using Sisoft Sandra, AIDA or some other Dhrystone package?

 

[PsharkJF]

Well, if someone could link me to a testing suite for this, I could do it myself. I was under the impression that you couldn't find it "on the fly", that it would be published somewhere.

 

[PumpActionWalrus]

Sisoft Sandra, they have a free version

 

[Ronin]

Which is what I was planning on using.

Sorry, been swamped at the office, and haven't had time to get this done 🙁