Mission impossible : 290 1:1 2x512MB Hynix –D5 CMD 1T @ DFI

[Dagalidis]

When received my new Board (DFI LanParty UT NF3 250GB) and install it I was very impressed with the overall overclocking potential….
Was possible to run my ADATA VITESTA DDR500 with Hynix –D5 chips at 265 FSB @ 2x512MB 1:1 at default timing settings at CMD 1T.
When used 1x512MB was possible to run 290 1:1 again with CMD 1T..
But 290MHZ FSB 1:1 with ONLY 1X512MB was a big problem for gaming Performance and that was hurt a lot…

Last night I starting to spend a lot of time to playing with DDR timing settings included on BIOS….
After 5 Hours of searching and testing finally mission completed .
Now running a Perfect 285 FBS 1:1 @ CL 3.5 4-4-6 1T @ 3.08 VDDR rock stable error free on MEMTEST+ @ PRIME95 @ Super PI 32MB @ Games….

When Raising FSB up to 290 Getting 121 errors per loop on 5th test of MEMTEST+ and need to play a little more to get it error free….

Please note that I have an A64 3200+ Claw C0 1mb Cache 1 year old which means that is a BAD overclocker versus to new A64 stepping …

Need a little more tweak to raise bandwidth and FSB before I post all the DDR timings here…

Now enjoy with me some screens and just wait for the timings…..



Here is SANDRA MEM...



Just compare same FSB at 1x512MB @ auto DDR Settings....
As you see i dont loose much



And here is A64 Tweaker screen reporting CL4 but i choose CL3.5 on BIOS

[Ubermann]

What time do you get in Superpi 1M ?

[Dagalidis]

What time do you get in Superpi 1M ?

At 277 X 9 = 2493 i get 35sec

My last SPI 1MB at 278 X 9 = 2507MHZ @ 1x512MB @ auto timming settings was....



But i not completed all the tests yet....
I will compare also the 3D Performance to see +/- i get....

I'm at work now after a long long night and i don't have with me more Bench to show...

Stay tunned ... I'll BE BACK....

[Dagalidis]

Next step is to maximize FSB with 1X512MB and lets hope 310MHz - DDR620 @ CMD 1T can be stable..

[kostas7gtr]

Great Job Dagalidis ! Hope you get 300 1:1

Καλό το Dfi τελικά Μήτσο ε ?

[Dagalidis]

Great Job Dagalidis ! Hope you get 300 1:1

Καλό το Dfi τελικά Μήτσο ε ?

Thanks my friend....

Greeklish : Αστέρι φίλε - πραγαματικό διαμάντι. End Greeklish

[=[PULSAR]=]

Nice Dagalidis, now all you have to get is a nice clocking 3700+ and you can probuably get 31/32 sec on SPI maybe even 30 if you push it real hard.

[ahfung]

Congratulation on the oc! 290MHz is quite impossible for Hynix but you did it, though the timing is a little lousy...

What's the difference between the fastest at 2.5-3-3 1T vs 3.5-4-4?

[Dagalidis]

Congratulation on the oc! 290MHz is quite impossible for Hynix but you did it, though the timing is a little lousy...

What's the difference between the fastest at 2.5-3-3 1T vs 3.5-4-4?

Thanks....

I cant say for sure about the difference in Performance cause never used 2.5-3-3- 1T....
But i will drop the fsb to get stable 2.5-3-3 1T and compare 3D VS 3.5-4-4 1T at same FSB-CPU MHz and see...
I believe 1 - 2 % is the max drop in performance .... but on the other hand this drop is reversible due to fact of high FSB gain at same cpu mhz....

[ahfung]

Thanks....

I cant say for sure about the difference in Performance cause never used 2.5-3-3- 1T....
But i will drop the fsb to get stable 2.5-3-3 1T and compare 3D VS 3.5-4-4 1T at same FSB-CPU MHz and see...
I believe 1 - 2 % is the max drop in performance .... but on the other hand this drop is reversible due to fact of high FSB gain at same cpu mhz....

I for myself stick to 2x512MB Hynix D43 at 26XMhz 2.5-3-3 1T on my Lanparty. My 3200+ CG 1M max out at around 2600-2630MHz so I just let it run at 1:1 262x10. Wonder if it'd do 290 x 9 1:1 at 3.5-4-4 and any faster...

[=[PULSAR]=]

Dagalidis, is this true of what I am reading in this review this board can do 2.108vcore? (1.55 basevoltage x 136%) If thats the case Im getting this board and 3.1vdimm and 1.9cset voltage. Sounds like a really spectacular board.

[Dagalidis]

Dagalidis, is this true of what I am reading in this review this board can do 2.108vcore? (1.55 basevoltage x 136%) If thats the case Im getting this board and 3.1vdimm and 1.9cset voltage. Sounds like a really spectacular board.

Its true my friend both of your readings...
Also you can have up to 4.0 VMEM using modded BIOS @ PSU with 3.3Volts rail adjustable up to 4.2 Volts...

[saaya]

wow..... wow wow wow....

2 512mb sticks at almost 300 on a 754 is impressive! and at 1t?

[krampak]

Here is my OCZ PC4000 Gold rev2 (Hynix D5):

FSB300 Cas 3-3-3 2T (1 stick 512 with DDR booster):



FSB280 Cas 2'5-3-3 1T (2 sticks 512 without DDR Booster):



I got it up to 305 2'5-3-3 2T but didn't finalyze the test :P And my max a 1T was 285 2'5-3-3 with 3'2V. Don't tried more Voltage yet (the mobo is stoping me, with a divider 200:180 it freezes at more than 310Mhz).

PD: DFI lan party everything air cooled. (yea the second one has a good cpu temp, it was outside of my house @ 7ºC ambient )

[Rabbi]

wow..... wow wow wow....

2 512mb sticks at almost 300 on a 754 is impressive! and at 1t?

2 x 512mb isn't realy holding me back here:



This with 1T, 2 x 512 Ocz pc3200 rev2 and a Lanparty Ut.

[=[PULSAR]=]

Rabbi is that real??? or are you using a half multi? Can you do some benches for us.

[Rabbi]

Off course those settings are not stable, but it was only to show that 2x512mb isn't a bottleneck.


Here a bandwith shot:

Cpu-Z shows only 512, but this was with 2x512. I can run it again if needed.

This is quite stable:



This is with mp 7, for some reason my motherboard won't post with mp 8 htt 300.

[Ubermann]

Run SuperPI and try =)

[Rabbi]

Ok, will try that tomorow.

[Dagalidis]

Ok my friends...

Here is some specs of my system

CPU = 3200+ Claw C0 1mb Cache
DDR = Adata Vitesta DDR500 (Hynix -D5) 2x512MB
Mobo = DFI LanParty UT NF3 250GB
PSU = Antec True Power 480W
Cpu Watercooled by EXOS

LTD = x2.5 For FSB > 300Mhz
x3 For FSB < 300

Those are timmings used for 295 2x512MB @ 1T...

Dram Frequency Set(Mhz) = 200(Mhz)(1/01)
Command Per Clock(CPC) = Enable(1T)
Cas Latency Control(tCL) = 3.5
RAS# to CAS# delay(tRCD) = 4
Min RAS# active timing(tRAS) = 06
Row Precharge timing(tRP) = 4
Row Cycle Time(tRC) = 11
Row Refresh cyc time(tRFC) = 17
Row to Row Delay(also called Ras to Ras delay)(tRRD) = 0
Write Recovery Time(tWR) = 2
Write to read Delay(tWTR) = 1
Read to Write delay(tRTW) = 5
Refresh Period(tREF) = 3632
DQS skew Control = Increase Skew
DQS Skew Value = 16 or 32 (i use 32)
DRAM Drive Strength = Level 1
Max Asynce lantency = 8
Read Preamble time = 5.5nano sec
Idle Cycle Limit = 0 or 16 (i use 0)
Dynamic Counter = Enable
R/W Queue Bypass = 8x
Bypass Max = 5x
32 byte Granulation= Disable (8burst)

Until now i manage to have 295 1:1 @ 2x512MB 1T error free on MEMTEST but 287 stable on Windows

Also using same timings using 1x512MB @1T i have 315 error free on MEMTEST+ but again ONLY 290 stable on windows....

I know its not a RAM issue but probably CPU limit for my old Claw C0 3200+ 1MB cache..
My cpu was of one the first step and is 1 year old..
So is a bad overclocker and suffer from Mem Controller instability...
I cant explain for sure why i can pass 2,5 hours wall time of 5th MEMTEST using 312 Mhz - DDR624 and when i boot into windows and try to bench using Super PI - FAST PI or a heavy 3D Bench, i get random black screens and Restart...

I will try finding out why get those problems and will inform you...

[Ubermann]

Thanks for posting settings, always good with new ideas!
And might solve some clocking for other that have CO cpu.

[ugp]

@Dagalids

I see you have a C0 ClawHammer. What is the highest stable overclock you have reached with that CPU? I can't seem to get mine high with any memory settings. I think mine is just a crappy CPU

[Rabbi]

SuperPi @ 2100mhz:

[Zeus]

Dagalidis, great job as always my friend!

C0 ClawHammers seem to have bad memory controllers.
I get errors in test 6 above 290 which i think are CPU related, test 5 passes np even at DDR600 2,5-3-3-8.
I hope to test a new CPU soon.

Rabbi, nice to see you here (OC4ever @ GoT right?) you have some very nice RAM and CPU!!
Great OC.

[ugp]

Dagalidis, great job as always my friend!

C0 ClawHammers seem to have bad memory controllers.
I get errors in test 6 above 290 which i think are CPU related, test 5 passes np even at DDR600 2,5-3-3-8.
I hope to test a new CPU soon.

Rabbi, nice to see you here (OC4ever @ GoT right?) you have some very nice RAM and CPU!!
Great OC.

at 290 you get errors but at 300 you do not?