Looking for a program to show memory speed

Hi, I was just wandering if there is a program out there that shows memory speed? I would have thought there was but I can't find a program that shows the actual speed my memory is running at. You see, I just built a new system with 1 gig of corsair ULL twinx 1024 PC3200 ram. PC3200 is 400 Mhz, correct? So with dual channel the bandwidth should be equal to that of 800 Mhz, right? Someone please correct me if I'm wrong. Anyway, I can't seam to find a program that will tell me what my memory is running at, if anyone knows a program that can help me, than please let me know. Thanks.

SiSoftware Sandra Standard (Win32 x86 Unicode) has the common buffered and unbuffered memory benchmark.

The nForce2's "dual-channel" memory system is slowed by the fact that it has a 400MHz (200MHz x 2) FSB. So performance would be under 3200MB/S (PC3200 - DDR400). There is a slight boost for it, though, over running with single-channel.

If you meant a program to just show what you have the memory set to, CPU-Z is handy for that. If you want to see your memory bandwidth, such as demonstrating that Dual Channel on an NF2 motherboard yields a data rate improvement of 48% over single channel, then consider good old MemTest86.

Hoot

I believe with the program AIDA32(the best out there), there is an option to test your memory, both read and write speeds. It'll compare it to other machine types to see where you get placed. That program is great for tons of other things too.

I have both AIDA32 and SiSoftware Sandra, but they wouldn't show me what my memory speed was set to. I am still a little confused about what my memory's frequency actually is, CPU-Z was an interesting program I didn't know about though thanks

When I do a benchmark with Sandra, it says my estimated max memory bandwidth is 2672 MB/s. Is that right for dual channel PC3200? WHat is PC133 memory bandwidth?

In Sandra, you can select other similar or different PC setups to compare your results to. Just open up one of the other bar chart sources on the left hand part of the results and select the comparison rig from the pull-down window. There is documentation on how to use Sandra, that comes with it. AKA RTFM

Hoot

I don't know what pc133 bandwidth would be. why not test it and display those results...could be informative.

Hoot said:

If you want to see your memory bandwidth, such as demonstrating that Dual Channel on an NF2 motherboard yields a data rate improvement of 48% over single channel, then consider good old MemTest86.

Hoot

Click to expand...

What are you talking about? 48%? You get 3-5% at the most on NF2

mrspec3 said:

What are you talking about? 48%? You get 3-5% at the most on NF2

Click to expand...

Well, if you trust MemTest86 as a memory testin program and you look at the information on the screen, other than the pass counter, test number and error column, you will see the memory data rate in the upper lefthand area. It's a great starting point for seeing the positive or negative impact that different settings have upon your memory's performance. Here's what I mean:



With one 512Mb stick of Buffalo BH-5 in single channel mode running 10x220 2-2-2-10 Synched the data rate is 1118MB/s

With two 512Mb sticks dual channel and everything else set the same, the data rate is 1656MB/s

Do the math...

Hoot

P.S. The picture is from my current setup running 11x220 2-3-3-11. Look at the hit the performance takes from retarding the timings. At 11x215 2-2-2-11, its 1714MB/s That's why I'm not fond of trading FSB for retarded timings, but that's just my perspective.

Run an unbuffered sandra test and post back

btw here is how to run an unbuffered test if you were wondering



I await the 48% increase

Ok wait are we talking about the same thing here? Your saying that memtest shows higher data rate? That may be true but there is no real performance gain...more than a few %

I think apples and oranges, though they are both fruit. I hate retyping so look at the latter part of this post from yesterday.

Hoot

lol, this is funny kind of got a little off topic. btw, I forgot that PC3200 memory actually runs at a 200 Mhz frequency, but sends data on the up, and down side of the sine wave, thus twicing the speed and then dual channel doubles that again for a total of 800 Mhz worth of bandwidth. Is that right?

Not exactly. Intel manages to accomplish that 4x but not AMD chipsets. As best as I can understand it, the term DDR is what characterizes being able to strobe data on both the positive and negative clock transitions. Dual Channel, again, as best I understand it, is more akin to the memory equivalent of running two disk drives in Raid0 (striped), well kinda...

Motherboards that support dual channel have two memory channels instead of one. So, instead of making the bytes of data wait in line, single file and sending that data to your RAM one byte at a time on the clock upstroke and another on the downstroke, the data waits in two lines, one bound for one stick and one bound for the other, sending them off in pairs, again one on the upstroke and one on the downstroke.

Now, you'd think that would equate to 4x, but (here's where my understanding falters) there are other mitigating factors, possibly between the Northbridge and CPU, that prevent taking full advantage of the data being sent off or retrieved in pairs, from resulting in full realization of the increase. I'm sure someone else has a link with a better explanation or perhaps in their own words.

Hoot

The older version of CPU-Z (under the memory tab) used to show DDR speed. The newer version (1.20 I think) just shows the frequency, so all you have to do is double it for DDR speed.

For example: A CPU/mobo running at 200 FSB with PC3200 RAM running at it's rated speed using the 1:1 memory ratio is 200 memory frequency. Since DDR is doubled, that gives you DDR400. The dual channel just adds another channel for data bandwidth to travel. The speed is the same. It's like adding an extra lane to a highway. The speed limit is the same but now twice the traffic can travel on the same road.

Hoot said:

Now, you'd think that would equate to 4x, but (here's where my understanding falters) there are other mitigating factors, possibly between the Northbridge and CPU, that prevent taking full advantage of the data being sent off or retrieved in pairs, from resulting in full realization of the increase. I'm sure someone else has a link with a better explanation or perhaps in their own words.

Hoot

Click to expand...

no, I see, and it makes sense. Everything passes through the chipset, and a system is only as fast as it's slowest point (sort of) Anyway, I fully understand now. Thanks for the help everyone!