Lower PCMark05 after upgrading memory

I upgraded from
2x512 DDR2-667 5-5-5
to
2x1024 DD2-667 4-4-4

and my PCMark05 score went down from 4570 to 4347.

Any ideas?

 

Was the 2Gb lower quality than the 1Gb?

 

Well, the 2x512 came with the Dell.
The 2x1024 is G.SKILL from Newegg.

 

Swap out again to confirm and if correct just another reason some people don't trust benchmarks!

 

There's something about Core 2 Duo and memory, something like the more memory you add the less efficient the processor is. Not sure exactly why. The only reason I know this is because I was watching some video on 80 core processors and how they only need something like 6GB RAM. Something like that.

 

Some thoughts:

1. I have seen that range of variability in PCMark05 results during repeat runs on the same hardware.

2. I don't think that PCMark05 has a very big memory footprint so the extra RAM will not be improving the performance during benchmarking.

3. Is your new RAM actually 4-4-4 at 667MHz? Use the memory tab of CPU-Z to check. (It will be the first genuine CL=4 @ 667MHz notebook RAM which I have seen).

SiSoftware Sandra includes a specific memory bandwidth test module. Run that benchmark with both RAM combinations.

John

 

That's some very good information. Here's the CPU-Z results.




I noticed the bandwidth is being reported as 333 Mhz. This is what the CPU-Z website had in their FAQ.

I wonder if the Santa Rosa chipset is ramping down the memory, or if the manufacturer fudged the numbers?

 

CPU-Z reports the true RAM speed (333MHz) the DDR-2 function means that it is effectively 667MHz. So you do have genuine CL=4 667MHz RAM (about time some showed up). I attach the CPU-Z result for typical CL=5 RAM.

John

 

I suggest running the benchmark with your old ram again.

 

I ran PCMark05 a couple of times, including a memory upgrade but it was always between around 2600-3000 marks. The 3000 marks was with 1GB of memory, think that was the first time.

 

Currently your processor has 4 Mb of cashe. When you increase the memory size your cashe becomes less efficient when random access occurs. PC Mark is not a bad test, it just shows how cashe and memory size are related. Try to use SiSoft Sandra test. Your results should increase.

 

Holy thread revival. And the info is wrong.