OCZ 3200 EB Plat vs OCZ 3500 Plat LE

[Radelon]

Hardware used:
P4 3.0c
Asus P4C800-E Deluxe 1015 bios
Hercules 9800 Pro
WindowsXP Pro SP1

Ram:
1gb Kit OCZ PC3200 EB Platinum
1gb Kit OCZ PC3500 Platinum LE (BH5)

Introduction:
The goal of this test is to see if OCZ's "enhanced bandwidth" line can stack up to the ever so wonderful BH5 sticks that everyone wants or has. To test this I did a baseline benchmark of both at DDR400 and a highend of both at DDR490 using Sandra 2004, PCMark2002, 3DMark2001se, and Gunmetal Benchmark1. Side Note(in the baseline benches the 3200 EB Plat was set to 2.5-2-3-6 in bios but for some reason would run 2.5-3-3-6 instead) Lets see how it turned out:

First Lets get to the baseline:
3200 EB Plat Sandra Buffered (2.5-3-3-6)
http://home.mchsi.com/~radelon/newstuff/200ebbuff.gif

3500 Plat LE Sandra Buffered (2-2-2-5)
http://home.mchsi.com/~radelon/newstuff/200plebuff.gif

As you can see the 3500 Plat LE holds a slight advantage over the 3200 EB Plat and now to PCMark2002 Memory.

3200 EB Plat PCMark2002 (2.5-3-3-6)
http://home.mchsi.com/~radelon/newstuff/200ebpcmark.gif

3500 Plat LE PCMark2002 (2-2-2-5)
http://home.mchsi.com/~radelon/newstuff/200plepcmark.gif

It's beginning to look like at lower speeds the tight timings used with the 3500 Plat LE is really playing a role and staying ahead of the "enhanced bandwidth" loosened timings of the 3200 EB Plat.

3200 EB Plat 3DMark2001se (2.5-3-3-6)
http://home.mchsi.com/~radelon/newstuff/200eb3dmark.gif

3500 Plat LE 3DMark2001se (2-2-2-5)
http://home.mchsi.com/~radelon/newstuff/200ple3dmark.gif

The 3500 Plat LE is still in the lead with 500 more 3dmarks than the 3200 EB Plat, which really isn't that much considering the differences in timings.

3200 EB Plat GunMetal Bench1 (2.5-3-3-6)
Gunmetal Benchmark1 (http://home.mchsi.com/~radelon/newstuff/200ebgunmetal.htm)

3500 Plat LE GunMetal Bench1 (2-2-2-5)
GunMetal Benchmark1 (http://home.mchsi.com/~radelon/newstuff/200plegunmetal.htm)

Seems like the 3500 Plat LE is the winner of the baseline round with a 4 for 4 sweep. Will the 3200 EB Plat be able to make up some ground in the highend benches or will the 3500 Plat LE maintain the top spot? more in the next post.....

[Radelon]

Now the Highend tests:
3200 EB Plat Sandra Buffered (3-2-3-7)
http://home.mchsi.com/~radelon/newstuff/245ebbuff.gif

3500 Plat LE Sandra Buffered (2-2-3-6)
http://home.mchsi.com/~radelon/newstuff/245plebuff.gif

Although not quite there, the 3200 EB Plat is almost tied with the 3500 Plat LE.

3200 EB Plat PCMark2002 (3-2-3-7)
http://home.mchsi.com/~radelon/newstuff/245ebpcmark.gif

3500 Plat LE PCMark2002 (2-2-3-6)
http://home.mchsi.com/~radelon/newstuff/245plepcmark.gif

3500 Plat LE stays in the lead and it doesn't seem to be looking back, let's see if throwing 3d benches into the mix changes anything.

3200 EB Plat 3DMark2001se (3-2-3-7)
http://home.mchsi.com/~radelon/newstuff/245eb3dmark.gif

3500 Plat LE 3DMark2001se (2-2-3-6)
http://home.mchsi.com/~radelon/newstuff/245ple3dmark.gif

Did I just see that right? The 3200 EB just passed the 3500 Plat LE in 3dmark. Now we are talking. The "enhanced bandwidth" has been pretty much keeping pace throughout the test and now it has finally topped the 3500 Plat LE. Before I get too excited lets check out the final tests.

3200 EB Plat GunMetal Bench1 (3-2-3-7)
GunMetal Benchmark1 (http://home.mchsi.com/~radelon/newstuff/245ebgunmetal.htm)

3500 Plat LE GunMetal Bench1 (2-2-3-6)
Gunmetal Benchmark1 (http://home.mchsi.com/~radelon/newstuff/245plegunmetal.htm)

Looking at the average FPS you don't see much of a difference although the 3200 EB Plat still is on top, but when you look at the Maximum FPS you see the 3200 EB Plat really shining.

Conclusion:
I believe OCZ has really caught onto something here. I was a little skeptical at first looking at the baseline benches wondering if it could ever overcome the tight timings of the 3500 Plat LE. But in the highend benches, it definitely became a contender in memory benchmarks and took over the lead in 3d benchmarks. With this technology a whole new light is shining at the end of the tunnel, the question is, which of the 2 lights are you headed towards?

[Paragon]

Very nice review Radelon, I must say I'm a bit surprised :yikes: but I know if there's a way to make it faster and better OCZ will find it :thumbs: You guys rock :rock:

[eva2000]

lovely indeed.. what advance chipset setings were used ? i.e. PAM = Auto or Enabled, PM = Auto, Std or Turbo, burst length ? DDRAM Refresh etc ?

[Radelon]

ah..... glad you mentioned that.....performance mode standard, mam= auto, burst length 8 clocks, Idle Timer= Infinite, Refresh Rate = 15.6usec..... they were the same for both kits.

[Hawk]

Not bad. . . :thumbs: 245FSB. . . . . :fire:

[Radelon]

Well, 245 is the highest stable on my 3.0c, so it could possibly go higher, but I need a different cpu to test the full limits

[Tony]

Jeb

you may want to try with TRAS at 5..it shows even better 3d scores.

Lets not forget this is a CH5 alternative not a BH5 one...it is however VERY close to BH performance as we see here.

[Radelon]

i didn't have any ch5 so i figured bh5 would be sufficient, just did a run with 3-2-3-5 , Performance Mode Turbo, etc..... had a 3dmark2001se run of 20,390 marks, so about a 315 mark improvement over old settings.