Memory Configuration and Benchmarking Performance

[crazyeddie]

I have an Athlon 64 3000+ Winchester (1.8Ghz) CPU on an MSI RS480M2-IL motherboard and a Radeon X800XL PCI-E video card.

I have been working on a couple of PCs, and I swiped a stick of PC3200 DDR out of my main PC to test another system. I was hesitant to button up my main gaming machine with only half of its RAM, because I wasn't sure I'd want to use it if the missing RAM caused a big performance hit.

I guess I didn't need to worry too much:

4792 3DMark05 - 1Gb (512x2) Dual Channel PC3200 DDR RAM
4714 3DMark05 - 512Mb Single Channel PC3200 DDR RAM

I really expected that cutting my system RAM in half and dropping from Dual Channel to Single Channel would impact performance more than this. A 1.63% drop in performance isn't a whole lot to worry about.

Just curious if anybody else has noticed similar results, because I'm thinking I'll be worrying a whole lot less about optimal memory configurations from now on...

 

[ryanv12]

that test is almost all GPU-bound. If you have the same video card but even compared a pentium 4 2.4 and an AMD FX-55...there would be much less difference than what you'd expect. In that way, it's not very indicative of real game performance, because you'll probably see a much bigger hit if you ran your test with a newer game.

 

[Matthias99]

read this link. Memory speed/timings has very little effect on most real-world apps on the A64 platform.

 

[Fern]

Hi crazyeddie

Just curious if anybody else has noticed similar results, because I'm thinking I'll be worrying a whole lot less about optimal memory configurations from now on...

"Optimal memory configurations" don't have much of a FPS effect on my NF2 rig either.

BUT, cutting your ram in half may well have a BIG negative effect on your actual gaming experience.

I have benched a number of games (not just benchmarks) and I have recorded SIGNIFICANT performance differences between 512MB and 1 gig of ram. Not really a FPS thing, but loading/reloading saves and in generally navigating around the in-game menu's.

The loading times etc in games like Far CRy and Painkiller is majorly affected by ram. For example, I have documented that the loading time in PK can go from 20 seconds (1 gig of ram) to 2.5 minutes or more with only 512MB.

What a PITA (2.5+ minute reload) when stuck at the stupid "jumping" puzzle in the coloseum level.

Timings? I agree. But, I would NOT go back to 512MB of system memory with today's games.

Fern

 

[Pete]

Dual channel only really helps if you have an IGP which has to split the system RAM bandwidth with the CPU. Otherwise, there are bigger bottlenecks in the system than the CPU-RAM conxn.

 

[ddogg]

yup 3DMark05 is entirely GPU bound so it doesnt really matter how much ram u have