Mushkin PC4000 Memory Review
 

I'l waiting for the cheaper twinmoss pc4xxx

I need ram that can do 240 at cas 2 2-5-2
now I have ram that does 240 cas 3 (!) 3-6-3

vddr + 0.2 :)

it's at 2.85 now...
I've put it at 3.0 and ran into severe stability problems...

Remind me, wich settings did I have to try out for you doc? Raw membandwith or real time performance?

250 4/5 5-2-2 CL2 DC
vs
250 1/1 5-2-2 CL2 SC
vs
240 1/1 5-2-2 CL2 DC
vs
240 1/1 6-3-3 CL3 DC

?

no raw bandwith. 3dmark or quake or what is easiest for you

250 (pc4000) 1:1 @ cas 3 4-8-4
versus
250 1:1 @ cas2 2-5-2 (if possible, or highest possible)
versus
250 5:4 @ cas 2 2-5-2


and

233 (pc3700) 1:1 @ cas 3 4-8-4
versus
233 1:1@ cas 2 2-5-2
vs
233 5:4 @ cas 2 2-5-2

that would be ultimately interesting to see how these compare at default ram speeds (pc3700 and pc4000)

that'd be 6 runs or sth, if you could do that, it'd be great!

it would give lots of answers as to the bandwith hungriness the p4 dual channel has lost or still retains...

1. the advantage at equal speed of fast timings... is it still there?
2. the difference between fast timings @ default pc3700/pc4000
or divided at 5:4, coz if there ain't much difference between 1:1 at slow timings and 5:4 at fast timings as we have seen, then it's of no use buying hi-speed ram... then pc 3200 that runs 200 2 2-5-2 would be just as good as pc4000 3 4-7-4 or even 3-6-3... however if the difference between 1:1 fast timings and 1:1 slow timings is high, then we need uberfast timings pc4200, however if the difference wouldn't be that big between 1:1 fast timings and 1:1 slow timings and 5:4 fast timings, we can all buy hiquality pc3200 and get the same results...


@ piotke : even at vddr 3.3 in windows, I still can't run cas 2.5 at 233 (default speed)

ok, i'm going to run those benches after i did my shopping.

250fsb test will be done single channel (only got one stick twinmoss), the others with 2*256mb mushkin level2.

thanx!
I'm back 2morrow, i look forward to that, if you have the time, stick in one run of 250/233 cas 2.5 1:1 3-6-3 too, as intermediate

Just to see how that stacks up

thanx richyB!

We need Information :)

did you do a vmod on the P4C800 ?

it's an abit ic7max 3
i can go all the way to 3.2

the p4c800 had a socket problem, remember :)

Hmmm, i'm having some serious difficulties with my 2.8C setup. I get BSOD's all over the place when running everything at stock. I tried running UT2K3 bench 10~15 times, i managed 3 successful runs. The funny thing is: it's always on a different moment, and nor memtest nor prime95 dedect errors. : /

I'm testing piece by piece in another mobo right now, but i don't think it's a HW problem. I must have hosed my WinXP setup somehow, I can't imagine how.

The P4C system is fixed. Made the benches you asked for:

Conclusion:

Get some cheap PC3200 that can do 5-2-2 cl2 at 200mhz, instead of expensive PC4200 that does 8-4-4 CL2.5 at 250mhz!

Sorry doc, my system won't boot at CL3, so the difference will be even larger.

I told you guys :)

read the above posts, throw away that pc3700 and higher, it serves nuttin' :)
probably however dualchanel 250 @ 2 2-5-2 could gain a bit more, coz the jump from sc to dc is quite high in botmatch, but overall it's preferred to use 5:4 div if you can't run 1:1 at fast timings

thanx richyB!

differences are marginal anyway, so spending that amount of money on the ram is a lost cause

my ram doesn't even boot @ anything lower then cas 3 even with divider on and even with 3.2 volts, quite some pc3700 by corsair :)

that jump from SC -> DC is indeed pretty impressive!

that is the only confounding factor, if dualchannel cas 2.5 4-8-4 would jump the same amount, it would bring it a little bit above the dualchannel 5:4 @ cas 2 2-5-2, but it'd be a little bit :)
it's quite stunning to see that @ pc 3700 speeds, the difference between cas 2 2-5-2 @ 233 mhz and cas 2 2-5-2 @ 186 mhz (divider 5:4) has a differnce of about 1 frame / second????

Spending over 100 euro's more for 500 mb tight pc3700 compared to readily available tight pc3200 is scandalous...

SC was 256mb ram, DC 512mb. The SC <> DC results should not be compared to each other, since there was some light swapping during the benchmark which slowed down the performance.