New RAM results in poorer performance

LONEWOLF · Post by **LONEWOLF** » 18 Dec 2003, 08:28

My friend recently purchased a new PC. It came fitted with 2 x 256 MB Legend DDR400 DIMMS configured in dual channel array. He encountered many problems and tested the RAM with Memtest86. The program found 1002 errors so he returned the RAM for replacement modules. Strangely, afterwards, his 3DMark2001 score decreased by 400 points.

After my friend's experience I decided to test my RAM as well. I did not have any reason to suspect faults as my PC was running perfectly. However, Memtest86 encountered numerous errors as well so I exchanged the RAM at Rectron. Initially my PC came equipped with no name OEM Nanya RAM but Rectron exchanged it for similar spec'd, name branded Transcend RAM. Because of the "better" RAM I expected my 3DMark2001 score to increase, but it also decreased - by 150 points

Question: why did both our systems perform poorer after we swapped our RAM? Considering that neither of us made any changes to our BIOS's or OS configurations? And my friend swapped his RAM for EXACTLY the same modules while I, supposedly, got "better" RAM.

RuadRauFlessa · Post by **RuadRauFlessa** » 18 Dec 2003, 08:44

Go and run 3DMark2001 again ad see what you get. You could get that flux cuz of temp inside the box. Leave your pc to cool down before starting it up and then try run the test. You could possibly get better scores.

LONEWOLF · Post by **LONEWOLF** » 18 Dec 2003, 09:03

After I installed the new DIMM and ran Memtest86 for 24 hours to ensure that the new DIMM was not faulty, I left my PC for another 24 hours before I actually ran 3DMark2001. I ran 3DMark immediately after Windows loaded; five times. Where I usually received a humble score of about 11150 points on average, I could not surpass the 11000 mark anymore with a new average of about 10998. Not being able to achieve 11000 points anymore hurts the ego

Post by **OnlyOneKenobi** » 18 Dec 2003, 09:03

The drop in performance in 3Dmark might not be related to the ram, as Rua said it might be temperatures or maybe your hdd just needs a defrag. 150 and 400 3dmarks isn't that much after all.

nutcase · Post by **nutcase** » 18 Dec 2003, 12:32

DDR has a few specs few people take note of. Besides the frequency, you have to look at the following three properties as well.

SDRAM CAS# Latency
Selects the number of clock cycles required to
address a column in memory.

SDRAM RAS# to CAS# Delay
Selects the number of clock cycles between
addressing a row and addressing a column.

SDRAM RAS# Precharge
Selects the length of time required before accessing
a new row.

In most cases you will get a 3,4,3 (The slowest) module.
You probably had something faster (e.g. 3,3,3) first and now you are sitting with slower modules (e.g. 3,4,3). Check these settings in your BIOS.

RuadRauFlessa · Post by **RuadRauFlessa** » 19 Dec 2003, 07:29

That is propably some verry good advice.....

LONEWOLF · Post by **LONEWOLF** » 19 Dec 2003, 08:36

My 512 MB DDR333 DIMM Nanya RAM (part no. NT512D64 S8HAB6-6) was rated 2.5-3-3 and the Transcend replacement (part no. TS64MLD64V3F5) is rated exactly the same. I also checked under "Advanced Chipset Features" in the BIOS and all the values for the RAM remained the same (I compared it with a previous printout of when my Nanya RAM was installed). I also checked under "Voltage Control" and the RAM is correctly detected as "166 MHz-DDR333". During POST and in Windows the full 512 MBs of RAM is also correctly detected.

Therefore, the only setting under "Advanced Chipset Features" that may have an impact is where you have the opportunity to set certain values like CAS latency manually. I have always left this option "Auto" and haven't changed it for the new RAM. Problem is you never know what the BIOS assigns if you leave it "Auto". I say this because initially the CAS latency is shown as 2.5 but if you change it manually, say, to 2 and set the option to "Auto" again, it still says 2 while I'm sure it was reset to 2.5.

So maybe although on paper the Nanya RAM and Transcend RAM are rated the same, but in real world terms the Nanya RAM was faster

But then again the Nanya RAM did show up lots of errors in Memtest86 while the Transcend RAM was clean. And 150 3DMark2001 points in hindsight isn't all that much like Only1 said.

Fishzn · Post by **Fishzn** » 19 Dec 2003, 13:56

Certain pieces of hardware differ from each other almost always even though they have the same rating, its like having a motor vehicle of the same brand but they almost always never give the same performance!

You RAM sticks might have the chips rated at the same speed but the fact remains certain types of ram are faster than others, many say it is negligible but on certain systems it does rear its ugly head! Funny though, don't you have Duall DDR capabilities on your mobo, that would get you that extra 500 points maybe even a 1000 extra, dual ddr doubles the memory bandwith, so if your mobo has it, enable it!

Other than that the score decrease is only slight and i'd put that down to manufacturers hardware differences!