Test Results

To test overclocked stability, we used the very demanding Gun Metal 2 - Benchmark 2, which pushes systems with its DX9 routines. To be considered stable for test purposes, Gun Metal, our Quake3 benchmark, UT2003 Demo, and Super PI had to complete without incident. Any of these 4, and in particular Super PI and Gun Metal, will crash a less-than stable memory configuration.

PMI4200 Gold - 2 x 512MB Double-Bank
Speed Memory Timings & Voltage Quake3 fps Sandra UNBuffered Sandra Standard Buffered Super PI 2M places
(time in sec)
400DDR
800FSB
2-3-3-6
2.75V
319.3 INT 2678
FLT 2742
INT 4463
FLT 4426
132
500DDR
1000FSB
2.5-3-4-7
2.75V
388.3 INT 3287
FLT 3292
INT 5543
FLT5467
107
535DDR
1066FSB
3-4-4-8
2.75V
410.6 INT 3413
FLT 3413
INT 5859
FLT 5841
101
543DDR
1086FSB
3-3-4-8
2.85V
416.2 INT 3716
FLT 3741
INT 5935
FLT 5933
99

The most important benchmark with any memory is that it is honestly rated and meets its specifications; PMI4200 Gold is certainly honestly rated at DDR533 at 3-4-4-8. However, for enthusiast memory, we usually find performance even better than the rated specification. Unlike most of the competition, PMI could not run any faster than its rated specs at DDR533. We were also able to reach DDR543 with a little more voltage, which shows some headroom with DDR533. However, this is, again, a small margin compared to comparable enthusiast memory from Corsair, OCZ, and Mushkin.

Performance Test Configuration Performance Comparisons
POST A COMMENT

13 Comments

View All Comments

  • KillaKilla - Saturday, March 27, 2004 - link

    [q]Why is my dinghy smaller after I go swimming? Why cant I buy pork boullion? Why cant I buy toast in the store?[/q]

    I too would like to know! Damnit, wes, Stop witholding information from the public!
    Reply
  • bldkc - Friday, March 26, 2004 - link

    color, color, color. Did I mention that I'm a sociopathic liar? Reply
  • bldkc - Friday, March 26, 2004 - link

    You may have told us before about the inability of the software to change the color, but I missed it. Thank you for doing so again. I am aware of the stratospheric bandwidth usage at Anand, and that is because you guys rock. Thanks for the excellent articles, and keep up the good work Wesley. I won't mention the colors again. Reply
  • Visual - Friday, March 26, 2004 - link

    The graphs really could use some colour coding though. Reply
  • Wesley Fink - Thursday, March 25, 2004 - link

    TrogdorJW -
    The flash charts are much smaller than JPEG would be but I don't have any numbers to share. Email Jason Clark, who does our IT reviews, and I'm sure he can give you some idea of the bandwidth savings.
    Reply
  • TrogdorJW - Thursday, March 25, 2004 - link

    Damn, Wesley... this is the first I've heard of this, and after looking back at the charts, I am only now discovering that you use Flash files for the graphs. Cool! I'm guessing that there's a Java application that is used to generate the graphs from some spreadsheet? Or are all the graphs generated in real time? Meaning, if there were an error in a graph, like a misplaced decimal or something, could you simply update the spreadsheet, or do you have to run the Java Graph utility as well? Either way, that's pretty friggin' cool. But then, I'm a programmer geek and like seeing such uses of technology.

    If you don't mind answering, how large are the generated graph files? I would guess that in JPEG format they would be on the order of 30-50K each. What sort of savings do you get out of this?

    Anyway, back to the memory being tested here. After the results not too long ago showing that low latency DDR400 timings with a 5:4 ratio on a P4 resulted in similar performance to 1:1 ration and higher timings, I really don't see much need for these "faster" DIMMs right now. My Mushkin can run 2-3-2-6 at PC3200, which is pretty close to the top. Seems like I can just run the system bus faster while keeping the memory bus at 400 MHz (200 MHz, actually) and still get similar performance. Or is that only the case with P4 systems?
    Reply
  • RuStYwAvE - Thursday, March 25, 2004 - link

    Probably PMI and OCZ go through the same manfacture like Dell or Alienware Laptops aren't really design by them. They just buy laptops from Clevo or Compro and retag their names on them. Creative Labs and PNY video card are retaged MSI cards. The only thing I see PMI copying is the packaging, if the price is lower then OCZ I would buy PMI. Reply
  • RuStYwAvE - Thursday, March 25, 2004 - link

    Reply
  • kamper - Thursday, March 25, 2004 - link

    They're never going to make it. Not with a name like Pmi and a logo with nice rounded letters and a halo(?) around the 'i'. With all the big companies as close as they are at the top why would you buy this when you could have cool looking names like "HyperX" or "OCZ" on your memory?

    Ok, I'm just kidding. It's weird how there's suddenly so much memory that is so close right around the top. I bet pricing plays the biggest factor in their success.
    Reply
  • Pumpkinierre - Thursday, March 25, 2004 - link

    Go on wesley, now that you're off the beaten track, get hold of some of that 2-3-3-6 Xtreme DDR PC3700+ (http://www.xtremeddr.com/products/x_pc3700+.shtml) and tell me if it is any good.
    Reply

Log in

Don't have an account? Sign up now