Mushkin XP2-6400 (996580) 4x2 GB testing

[-n7-]

I previously had a 4x2 GB mix of 4 GB of this Mushkin & the 4 GB G.Skill DDR2-800 5-5-5 PQ kit.
The G.Skill didn't clock nearly as well though, so it got relegated to use in my HTPC & i bought a second kit of the Mushkin.

My P5B Deluxe finally got retired into use in my HTPC as well; it had a good run, but it was showing its age & wasn't quite doing what i wanted it to.

So now i have a DFI UT P35-T2R + Q6700 + 4x2 GB Mushkin in the main rig.
Still learning the ropes on this DFI, so results could be better, voltages lower, & timings tighter, etc.
But for now i'm just getting the hang of tweaking things with it.

I will say this DFI is far better already for working with OCing, though it requires a lot more fine tuning.

Overnight i ran:
8 GB DDR2-1050 5-5-5-15 (7 hours P95 custom)

Now i tightened up perf. level to reasonable levels & am trying for something i can run for 24/7 use i hope.
TRFC is deliberately very loose; will work on that later.

8 GB DDR2-1080 5-5-5-15 in testing


NB & PWM get absolutely scorching, as can be seen in the first screenie, so i'm using the house fan to help a little for now :p, & will likely pick up the side panel for my case that has the 250mm fan on it; hopefully will help a bit.

So far i can't seem to use the 1333 MHz strap with this RAM; it seems to prefer the 1066 MHz strap, as it won't even boot at the speeds i can do with the 1066 MHz one, which seems kinda backwards to me.

But it's a fun learning process, & damn, this mobo's voltage & settings control blows away my old P5B-D.

[Spyrus]

I have gone up to 530 with these so you are good there.
Sure 4 sticks is another deal for you.

[Mango31]

I got a question - currently i run an E6600 at 3.6 GHz, FSB 400 and the Mushkins 2x2 GB at 1:1, 4-3-4-10. Would it benefit in game performance if i try to get around 500 MHz with another divider and 5-5-5-15 timings ? Or is the lower latency more important ?
Sorry if this sounds dump, i never experimented in this direction.

Thanks !

[Spyrus]

If you can pass 1000+@5-5-5-15 then it is better than 800@4-4-4-12

[Mango31]

Thanks for the info ! Looks like 1:1 safe and stable @ CL4 is still the better long term choice .

[-n7-]

So i wanted to update this tonight, since i finally got around to testing the dimms individually.

I did testing relatively briefly to save some time, so don't take these as absolutely 24/7 stable clocks, only more thorough testing would show that for sure.

I ran Memtest86+ initially, & then HCI to about 20+% to get screenies.
Though not always, if i can get 20% or higher with HCI, it's usually very close to stable if not stable.

Mushkin 996580 XP2-6400 - DDR2-800 4-4-4-12 (two 2x2 GB kits) tested one dimm at a time.
vdimm = 2.1V set in bios, which is read as 2.02V actual via software

dimm 1 = DDR2-1124 5-5-5-15
dimm 2 = DDR2-1080 5-5-5-15
dimm 3 = DDR2-1100 5-5-5-15
dimm 4 = DDR2-1090 5-5-5-15

I'm very happy with the results.

Worst of the four dimms does 540 MHz @ less than warrantied vdimm.

[Cronos]

Man, these 1:2 and 2:3 dividers together with high PL levels just killing your RAM performance. What the point in having high memory OC if the actual performance is low?

4:5 is the highest sensible divider for DDR2.

[-n7-]

Man, these 1:2 and 2:3 dividers together with high PL levels just killing your RAM performance. What the point in having high memory OC if the actual performance is low?

4:5 is the highest sensible divider for DDR2.

I'm not running those same subtimings for actual use.

I deliberately loosened TRFC & TRD to make sure they were outta the way & wouldn't prevent higher OCs.

I can run TRD @ 5 or sometimes 4 if i'd like, & significantly tighten TRFC as well (not that it makes much of a difference).

I obviously don't run 1:2 for normal use.

I was running 2:3 for a while, but i cannot seem to figure out what magic is required on this board to allow true stability with four dimms.

Angering me quite a bit actually.
Seems that even with the RAM clocked down to default speed & the NB voltage cranked it loves to fail HCI

__________________________________________________ __________________________________

Here are some results i did on another forum just comparing 1:1 tighter timings vs. 2:3 looser & then 2:3 with TRFC & TRD tightened up a bit.

Q6700 @ 10x350 = 3500 MHz 2:3 DDR2-1050 5-5-5-15 TRCD=55 - TRD=5
vs.
Q6700 @ 10x350 = 3500 MHz 2:3 DDR2-1050 5-5-5-15 TRCD=60 - TRD=8
vs.
Q6700 @ 9x389 = 3501 MHz 1:1 DDR2-778 4-4-4-12 TRCD=60 - TRD=8

Some of the the subtimings that were set to auto did tighten slightly @ 1:1, so there's a bit of an advantage to the 1:1 benching, but performance level aka TRD remained the same, at 8.

All benches were run at least three times then averaged.


Everest Memory Benchmark
Read
2:3 TRD 5 - 8410.67 MB/s
2:3 TRD 8 - 7575.33 MB/s
1:1 TRD 8 - 7314.67 MB/s
Write
2:3 TRD 5 - 6379 MB/s
2:3 TRD 8 - 6378.33 MB/s
1:1 TRD 8 - 7100 MB/s
Copy
2:3 TRD 5 - 7066 MB/s
2:3 TRD 8 - 6901 MB/s
1:1 TRD 8 - 7139 MB/s
Latency
2:3 TRD 5 - 59.67 ns
2:3 TRD 8 - 68.37 ns
1:1 TRD 8 - 70.87 ns

WinRAR bench
2:3 TRD 5 - 1583.25 KB/s
2:3 TRD 8 - 1521.75 KB/s
1:1 TRD 8 - 1429.5 KB/s

3DMark01 defaults
2:3 TRD 5 - 45823
2:3 TRD 8 - 44806
1:1 TRD 8 - 45344.33

Cinebench R10 64-bit
2:3 TRD 5 - 14327
2:3 TRD 8 - 14147.33
1:1 TRD 8 - 14245.33


Yes, TRD makes a massive difference

[Cronos]

On my ASUS P5K-PRo, setting tRFC to 55 solved my stability problems with G.Skill DDR2-1000
I have it running 4x2Gb 1000MHz 1.94V PL6 4:5.

On the other side, on Gigabyte DS3P i could not make it stable no matter what.

[Spyrus]

It is the chip-pcb that does not like it?
At 540+ with Evo i am still with tRFC=50 with no problem.
They also boot fine with tRFC=42 at 1020MHz.

[Cronos]

It is the chip-pcb that does not like it?
At 540+ with Evo i am still with tRFC=50 with no problem.
They also boot fine with tRFC=42 at 1020MHz.

Try to test in Linpack
Also, If you have only 4GB, probably tRFC 50 will suffice. I have not tested 4GB extensively.

[Spyrus]

No Linpack yet, but OCCT and HCI Memtest are ok.

[-n7-]

It is the chip-pcb that does not like it?
At 540+ with Evo i am still with tRFC=50 with no problem.
They also boot fine with tRFC=42 at 1020MHz.

Sure, you're only running 2x2 GB.

I can do DDR2-1080 as well with only 4 GB.

But i have 8 GB.

Try getting 8 GB stable over DDR2-1000...it seems to be basically impossible.

Last night i tried loosening every single subtiming to absolutely stupidly loose levels to see what happened.

And i mean stupid, like 5-6-6-20 & then a slew of retardedly loose subtimings.

Still can't do it.

It's not the RAM, it's a damn chipset limitation i'm seeing.

If anyone is getting 4x2 GB @ DDR2-1000+ speed stable in HCI Memtest, i'd love to learn how!

[Spyrus]

I think Tony from ocz have managed to run four modules of PC8500