hi

i just built my new system, and it's real cool. my mobo has 4 dimm slots, but i was reading somewhere that using 4 dimms can decrease performance.

i have 4 512mb sticks in there now (2gb of ram), and i was wondering: how much does it slow it down?

i'm running windows vista, and i do some light gaming and programming. would i notice the difference between 2 dimms and 4 dimms?

Toshiba M50-MX2:
Celeron M 1.5GHZ 1MB L2
512 DDR2 PC4200 RAM
60GB 5400RPM HDD
DVD+/-RW CDRW DVD-RAM

Yes 4DIMM can decrease performance due to when the CPU look for information in the memory, it will first start at DIMM 1 and 2, 3.... so on.. For example, if every time the CPU look information in RAM by each DIMM cost 1 cycle...

1 stick = 1 cycle
2 stick = 2 cycle
3 stick = 3 cycle
4 stick = 4 cycle

n stick = n cycle

So basically that means the more DIMMs you have in your motherboard, your CPU will take longer to search through the RAM other than fewer ones, but the performance difference is marginial < 1% max. If you are doing gaming and usually get 60FPS average, by using 4x dimms, you probally lose 0.6fps max so it will be like 59.4 FPS. Unless you really want to brag about the 1%, I suggest you to buy the lowest latency RAM and make up for the 1% difference..

TMP-Man

Asus P5P800-SE
P4 506 2.66Ghz @ 4100Mhz 1.525v
Thermaltake CLP0024 w/ 2000RPM FAN
1024MB DC Corsair DDR400
40GB 5400RPM/120GB 7200RPM HD
128MB Radoen 9500 mod 9700 @ 360/585

"but the performance difference is marginial < 1% max"

that is not true on amd64 systems. By using four memory modules you can loose a 5% to even a 10% decrease in perfomance. Why? Because with four memory modules, 2t command rate has to be used. You cannot have four sticks of ram and use 1t command rate, therefore, you take something like a 7% perfomance hit comparing to 1t command rate. However, because 2t command rate is giving less speed than what each module can handle, it can let you overclock a bit more than 1t command rate, so if you are lucky you can reduce greatly the perfomance hit of 2t command rate by overclocking.

Free

http://www.xbitlabs.com/articles/cpu/display/athlon64-e3-mem.html

Free

If you have the same timing for 1DIMM, 2DIMM and 4DIMM motherboard i.e 3-3-3-8 @ DDR400. How much performance will you lose by going there? It is true that when you use all 4 DIMM, the motherboard has limitiation now not able to run at DDR400 at lowest timing, but the 7-10% increase you are looking at is related to synthetic benchmark. Compare the games

Doom 3
108.9 vs 107.4

Farcry
181.43 vs 174.69

HL2
80.45 vs 78

Q3
471.4 vs 456.4

UT2004
180.87 vs 175.04

The largest difference I see is <4% in games... Unless you want to brag about ur memory bandwith in synthetic benchmark using less than with 10% increase over max dimms, feel free to do so...

TMP-Man

Asus P5P800-SE
P4 506 2.66Ghz @ 4100Mhz 1.525v
Thermaltake CLP0024 w/ 2000RPM FAN
1024MB DC Corsair DDR400
40GB 5400RPM/120GB 7200RPM HD
128MB Radoen 9500 mod 9700 @ 360/585

Well, the performance of 2t command rate can vary from dim to dim, even if having the same timings. When I run 2t command rate my minimum framerate in f.e.a.r at maximum quality drops from 38 fps to 34, which is in fact more than a 10%. Synthetic benchmarks are not to be overlooked. The games tested in the article I gave you are not even half demanding for the ram comparing to f.e.a.r. or quake 4 in ultra quality (note that in the article doom3 is tested in medium quality only). If you run quake 4 in ultra quality and you have a 256mb video card and one gig of ram you'll get some big slowdowns from moving from one area to another (going through doors, etc). With 2t command rate the "waitings" are way longer. I've been testing this cause I also wanted to go for 2 gigs by adding two more 512mb modules but I'm feeling like selling my ram and getting two 1024mb modules to keep command rate 1t. However, as I said before, 2t command rate can let you overclock some more and can help minimazing the perfomance hit, but that was not the case with my ram :(.
Tmp-man, even if the perfomance hit is not that much in some older games it is still to be considered, since the more ram demanding the application is the slower 2t command rate will behave. That's why some synthetic benchmarks show some very dissapointing results with 2t command rate.

Well... It's clear that the slowdowns of using 4 ram modules are not a "marginal 1%"...

Free

"When I run 2t command rate my minimum framerate in f.e.a.r at maximum quality drops from 38 fps to 34"...

Can you please justify your claim by showing a screen shot? Out of the benchmarking # you gave me show <4% increase... Also I'm talking about when both 2x DIMM and 4x DIMM are using the same timing, how much performance will you lose? Not in terms of 4DIMM have high latency than 2DIMM due to chipset limitation.

TMP-Man

Asus P5P800-SE
P4 506 2.66Ghz @ 4100Mhz 1.525v
Thermaltake CL2042 w/ 2000RPM FAN
1024MB DC Corsair DDR400
40GB 5400RPM/120GB 7200RPM HD
128MB Radoen 9500 mod 9700 @ 360/585

lol
the thing is that 4 dimms CANNOT use the same timings that 2 dimms in A64 systems because of 2t command rate.

2dims: 3-3-3-8-1t
4dims: 3-3-3-8-2t

If you switch manually to 1t with for dims the system won't boot. If it's auto, the bios automaticly asigns 2t.

I have f.e.a.r. on my new pc and it doesn't have internet yet but I will take the time of switching to 2t command rate again, taking the screenshots and copying them to a Cd-rw
just to prove you wrong xD

The minimum fps I get in f.e.a.r. would probably not be so low with 2t command rate if I put 4x512mb dimms, cause f.e.a.r would take advantage of 2gigs of ram even though with slower timings than 1gig. The thing is that when having 1gig you switch to 2t command rate the performance difference with 1t command rate is as huge as I told you before whith appliacations that use tons of ram. Of course one extra gig of ram could be better even though it is running on 2t command rate, but by itself, 2t command rate can be 5% - 10% slower than 1t, as I stated before.
If going for 4x512mb dimms can increase performance in some applications, it will most likely decrease in most of them since there are not many that use more than 1gig of ram. But if you have 2x1024mb dimms performance will stay the same for not so heavy applications and will increase a lot with heavy applications. The performance difference of 4x512 and 2x1024 (Same timings except command rate) is usually between a 5% and a 10%.

Free

I'd like to see your screen shot and show me the min fps increase over 2DIMM to 4DIMM, max fps and average fps... Then I will be convinced that 4DIMM can have 10% increase over 2DIMM... Unless you going for synthetic benchmarks like measure the memory bandwitdh, then you will see a 10% increase due to command rate. From what I see in the page you gave me, I don't see the 10% you are talking about in games. Like I said, unless you really want to brag about synthetic benchmark score, the 10% you are talking about, then feel free to do do... Remember, you buy the machine to play games and do real-world application. As stated from post #1

"i'm running windows vista, and i do some light gaming and programming. would i notice the difference between 2 dimms and 4 dimms?"

Your game F.E.A.R on the other hand is not light gaming lol...

TMP-Man

Asus P5P800-SE
P4 506 2.66Ghz @ 4100Mhz 1.525v
Thermaltake CLP0024 w/ 2000RPM FAN
1024MB DC Corsair DDR400
40GB 5400RPM/120GB 7200RPM HD
128MB Radoen 9500 mod 9700 @ 360/585

"Then I will be convinced that 4DIMM can have 10% increase over 2DIMM"

You got it all wrong? I never said that 4dimm can have a 10% increase over 2dimm. In fact, I said the oposite, that 4dimm x512mb is slower than 2dimm x1024mb because of the 2t command rate.

I would love to have all kinds of ram to test them but my case is like this: I've got only one gig of ram (2x512 dimms), and when I switch to 2t command rate manually, I get those min 34 fps in f.e.a.r. compared to 38 with 1t command rate. If I added two extra 512mb dimms I would be forced to use 2t command rate but the perfomance drop probably wouldnt be there (in fact, there would probably an increase) because f.e.a.r. will make good use of the extra ram, making up for the performance loss of 2t command rate. However, 2t command rate by itself, is somewhere between the 5% - 10% decrease in performance, as I was able to see with f.e.a.r. (a game which uses some big amount of ram)

I give you the point that he is doing "light gaming". Ok, I had forgotten that. I'm putting internet up there in my new pc this week so then I'll post the screenshots.

"As you can clearly see, the memory-only Sandra tests show a 21% drop in memory bandwidth in the standard buffered test when reducing to a 2T Command Rate. The more real-world Unbuffered Sandra test still shows a 15.5% bandwidth drop. Games and Super Pi, where memory is just one small component of the overall result, show about a 4% performance reduction. These represent the impact of this memory change on the overall system performance, where the Sandra scores measure impact on memory alone."

http://www.anandtech.com/memory/showdoc.aspx?i=2560&p=2

Well, that 4% is no 4% in games like f.e.a.r. and quake 4 in ultra quality running on a 256mb video card (specially on a 128mb one, since lots of ram has to be used). In my case, it's a huge 11% in f.e.a.r.

Free

"Well, that 4% is no 4% in games like f.e.a.r. and quake 4 in ultra quality running on a 256mb video card (specially on a 128mb one, since lots of ram has to be used). In my case, it's a huge 11% in f.e.a.r. "

Please define the defintion of light gaming...

TMP-Man

Asus P5P800-SE
P4 506 2.66Ghz @ 4100Mhz 1.525v
Thermaltake CLP0024 w/ 2000RPM FAN
1024MB DC Corsair DDR400
40GB 5400RPM/120GB 7200RPM HD
128MB Radoen 9500 mod 9700 @ 360/585

Also, this link

http://www.anandtech.com/memory/showdoc.aspx?i=2560&p=2

I only see <4% increase at games. And I do not see the 11% you are talking about.. I don't have to repeat myself over and over again.. Unless you are going for synthetic benchmark like sisoft sandra memory bandwidth, the impact for 4DIMM on games is minimal, dispite some when games take advantage of 2GB over 1GB... I do not know how u get that big of performance difference by using 1T instead of 2T.. Either your machine is screwed up some point, or you have problems with the software you installed.. Out of all the link, benchmark you showed me, games <4% period...

TMP-Man

Asus P5P800-SE
P4 506 2.66Ghz @ 4100Mhz 1.525v
Thermaltake CLP0024 w/ 2000RPM FAN
1024MB DC Corsair DDR400
40GB 5400RPM/120GB 7200RPM HD
128MB Radoen 9500 mod 9700 @ 360/585

If it's a socket 754 3700+, the decrease is dramatic. You go from full speed to 100mhz when using all DIMMs I believe...

Single/double sided also comes into play. I have 2 1GB sticks, and it'll only run 333...
with my overclock they're back up to speed, but at stock it sucks. If I add another stick of anything, it drops to 200mhz.

754 3700+ @ 255x10 HTx4
ASUS K8Ne Deluxe
2GB PQI DDR3200 / 6600GT
Ultra 500w x-connect PSU

Powerbok G4 1.67 1GB DDR 100GB 7200rpm / 9700 vid

AMD64 3400+/768MB/5700VE
1 Toshiba & a Dell