PC Memory configuration question

I have a K8V motherboard with three mem slots. When I built the computer I put in 2 x 512MB PC3200/DDR400. I recently added a 1 x 1G PC3200/DDR400 into the mix. It’s the same brand and latency. The system recognizes that I have 2G of RAM now.

However, I don’t feel like I’m getting the performance I expected and in some cases the performance feels worse. For example, the Dark Messiah demo seems to run worse now than it did with only 1G of RAM.

When I installed the 1G chip I put it into the open third mem slot since the MB manual didn’t seem to indicated it really mattered. Today I put the 1G into the number one slot instead and the performance on the Dark Messiah demo is even worse.

Is there an easy way to check to see what speeds I’m getting on this configuration?


Some motherboards drop the memory speed to 333MHz with more RAM sticks.

I use CPU-Z to check this kind of stuff out.

If the RAM chips aren’t reporting their SPD properly to the motherboard it will often misdiagnose what speed to run at. I usually hard set my RAM configuration in the BIOS to avoid issues.

Sarkus - unless a utility can inherently use 2 Gig of RAM vs. 1 Gig, you will see slower performance with the extra RAM. Why? The RAM has to run slightly slower to sync with additional chips. This means your latency and responsiveness is slower. Some people claim it’s unnoticeable… but disagree. I felt it is noticebale in app responsiveness .

Games like Battlefield 2 will run a lot better with 2 Gig of RAM.

Steve and Knikos are correct. Check your mem paramters and hard set them. Just make sure after you manually configure them that your board is still stable. Run 4M SuperPi for a quick test, and Prime95 to check for 100% stability (if you feel you need super stability).

The K8V has issues with three DDR400 RAM chips. You may get lucky, you may not. Try cranking them down to 333 and see if it runs better.

Using the CPU-Z utility I’m getting some odd results. Even with only the original two 512’s in the slots 1,2 it’s reporting a 200mhz speed, not 400. Same with only the 1 GB chip in place. That was after going into the bios and changing the Memclock mode to manual and then changing the Memclock to CPU ratio setting to the 400mhz setting. The other memory settings are at auto, except CAS latency is set at 3.0 (all chips are 3’s).

At this point I’m pretty confused.

Don’t believe everything that CPU-Z reports. I have 3 identical 512 pcs of RAM, two of which report properly from slots 1 & 3. No matter how I swapped them, I cannot get an accurate reading from slot 2. This MB was designed to use all three slots exactly as I am so it’s not a MB problem, but a problem with the program.

That being said, here’s something to try out. Follow me:

Windows uses Virtual memory (paging file), right? Back when you had X amount of RAM, the page file was X amount big. Now that you increased your physical memory, Windows most probably increased the Page file too (My system did the same thing). You will see a DECREASE in performance if the page file is too large for the amount of physical RAM.

To correct this in Win XP, go to My Computer on the desktop, Right-Click, choose “Properties”, “Advanced” tab, The “Settings” button under performance, “Advanced” tab again, “Virtual Memory” choose the “Change” button, “Custom Size” Option. Reduce this to something that seems reasonable like 512-1G Max.

You will need to experiment with this setting to achieve the best performance for your setup. This will force your system to use the physical memory more and the virtual memory less. Virtual memory is slower because it reads\writes to your hard drive. By forcing Windows to keep things in RAM and not allowing it to write large amounts of data, it should help with any system that uses AGP as well. Since the K8V uses AGP x8 (make sure that’s enabled in the Bios with a 256 Aperture), you might see a notable difference if the page file is used less frequently and more data stays in your physical RAM.

If this doesn’t help, you might have to actually reduce your CAS latency in the Bios or slightly increase the voltage to your RAM to increase stability.

Let us know how you make out.

If you’re going to be manually setting your virtual memory swap file in Windows, always set it to 1.5 times your total physical RAM on your system. You want enough to be able to page out everything you have loaded in RAM plus a little more. It also helps to set the minimum and maximum to the same amount, and download the utility called “PageDefrag”. If you can get away with it, placing the page file on a drive that doesn’t have your games on it that is of a good speed can also give small speed increases.

Thanks for all the help but I’ve pretty much reached my limit. No matter what I try I only get 200mhz speed on these RAM modules, despite the fact that I should get 400mhz on any 2 module configuration according to the MB manual. A different utility is reporting the same results. With that apparently being the case, it appears I might as well use all three chips.

Sarkus: 200MHz is the actual clock speed of your DDR RAM; its specs get listed as “400 MHz” because it’s double data rate RAM (i.e., it has twice the bandwidth of SDR RAM at the same clock speed), but the physical clock speed is 200 MHz.

Then how do I judge what I’m getting using the various utilities? Regardless of configuration, I’m getting the same results on both CPU-Z and Everest. According to the MB manual, however, some configurations should give a max speed of DDR 400 and others DDR 200. Where is that reflected?

unbgonwoh beat me to it. I had that board as well, and it is very confusing when the memory apps & BIOS say 200MHz - when you’re supposed to get 400 MHz.

What’s going on is the bus clock speed is 200 MHz, and your mem IS running at 400 MHz - but apps report it acording to BUS speed. The best thing you can do if you’re concerned abot performance is download SuperPi and run the 4M test. Write down the time it takes, a the re-run with the new memory removed. You could also run a 3dMark test (before & after) to see if there’s a recordable difference. I still think what you’re witnessing is increased latency due to running all 3 chips. But you’ll never feel better about the situation if you don’t run some tests and see what benchmarks tell you. If your only consequence is non-gaming apps not running as fast - then all is OK.

If you can return the memory, maybe consider 2x1 Gig RAM chips. After all dual channel doesn’t really work with more than 2 chips.

First off, don’t confuse bandwidth with frequency: in layman’s terms, the former indicates how much data can be pumped out and is usually measured in bytes per second; while the latter indicates the clock speed at which something runs, measured in cycles per second (Hz). DDR SDRAM is capable of sending data on the rising and falling edges of the clock signal, effectively doubling its bandwidth over that of SDR SDRAM. Thus, DDR SDRAM which runs at a 200MHz clock speed is often refered to as “DDR-400” to indicate that it has twice the bandwidth of SDR SDRAM running at 200MHz. It does not indicate that the DDR SDRAM is running at twice the clock speed as the memory bus, as jpinard states.

Confusing? Yes, it is.

Also, don’t confuse double data rate with dual-channel. “Dual-channel” means that the memory controller - which is usually on the MB, but in the case of the Athlon 64 family, is built into the CPU - is able to access two RAM DIMMs simultaneously, rather than one at a time, effectively doubling memory bandwidth. In order to use dual-channel, you need (A) a MB and/or CPU which supports it and (B) a pair of identical RAM DIMMs installed into matching RAM slots. The K8V appears to be a Socket 754 MB, which is a single-channel CPU.

As to Sarkus’s original problem: try setting the RAM bus to 166/333MHz and see what happens. Some S754 MBs, including mine, won’t run 3 DIMMs at 200MHz stably.