Need help adjusting memory timing

[retsamradassaT]

Thanks for the download!

Currently, my ratio is 16:6 with my 1066 5-6-5-17 setup.

[genetix]

Yeah, well, testing other Ratios/speeds probably will increase that even further. It ain'ttide to any specific speed even while SPD profiles would be 1066Mhz it probably even POSTs on 1200Mhz on CL6. Although, I have no suspects that on those speeds it would generate errors, but the point is with correct settings even it could be used for test.

Every ratio has pretty much completely different style how they function in performance wise. 2:1 would be fast 1:1 would be fastest however your FSB200 doesn't allow that. Would make most sense to check some overclocking reviews for your board and see what they have used.

example this fellow who ever benchmarked this has 300MB/s slower speeds on their benchmark that we had speed on screen yesterday
and I didn't even finish proper changes yet to make it really fly like I would of wanted (even while we got a lot more out than where we started) :
http://benchmarkreviews.com/index.ph...1&limitstart=6

Edit #1:

btw, while you're at it and probably got the everest now good. mind dumping me with that 'motherboard' > 'motherboard' page Bus bandwidth speeds.. they should be there by now as last time they were [ TRIAL ONLY ].. ;P

[retsamradassaT]

Field Value
Motherboard Properties
Motherboard ID 05/12/2009-RD780-SB750-7A66AG08C-00
Motherboard Name Unknown

Front Side Bus Properties
Bus Type AMD K10
Real Clock 200 MHz
Effective Clock 200 MHz
HyperTransport Clock 2000 MHz
North Bridge Clock 2000 MHz

Memory Bus Properties
Bus Type Unganged Dual DDR2 SDRAM
Bus Width 128-bit
DRAM:FSB Ratio 16:6
Real Clock 533 MHz (DDR)
Effective Clock 1067 MHz
Bandwidth 17067 MB/s

[genetix]

Well, that at least looks fast as hell. 17GB/s, lol Best on any case on my intel would be 14,5GB/s..

[retsamradassaT]

Hmm, I just ran prime95 for roughly about 2 hours, and within the first 10-20 min, two cores have failed... Now I am wondering if it's still too soon to run the test.. I mean like I thought my current setup with the memory timing is pretty stable, but it appears not. and I ran memtest for 6 hours with no errors.


Edited: I have increased the NBVoltage to 1.4, and it seems that I am still running into the same problem.

Edited 2: Yeah, it seems that no matter what I do, my prime95 is failing after the first 10mins or so... maybe it's a problem with the memory timing?

[genetix]

Your problem is not stable Northbridge and cores, not the memory probably. Lowest voltage you can get to memory and memtest through = +0.06v to memory to be 'Windows Stable' for testing. (roughly)

Prime is incredibly power dedicated for example, if I run prime on condition NB = 1.488v and vCore = Stable, VTT = stable. Northbridge is too high for prime and it would fail not after 10 mins, but fail at 2 hours, if I lower my NB to 1.440v (it loops to 1.424v) It's rock solid prime for 48 hours. What I am trying to explain. This is very very small things we are talking about. My best advice would be check the lowest memory voltage for that setup. Apply the +0.06v minimum. and verify with LinPack that cores are good and rest is good. Then try Prime95.

Anyway, here at this point ends my help on this matter in general for 2 reasons I cannot help you to stabilize an entire 'fleet' of voltages and because it is impossibility to know how unique equipment works.

[GSKILL TECH]

Try raising tRC to 33 or 35. What happens on stock timings? You should be able to run these at 5-5-5-15, if not try 5-6-5-16 and see if that will stabilize for Prime95. Also, what is your tRFC set at?


GSKILL SUPPORT

[retsamradassaT]

Memtest fails at stock timing, so I had to made some minor adjustments for it to pass the memtest. However with the current adjustments, it passes memtest, but it seems to be failing prime95.

tRFC was set to 127.5, but I will take your suggestion and set it to 195.

Time to do some more testing!

[GSKILL TECH]

Try the other timing as well if tRFC- 195ns alone is no good.

GSKILL SUPPORT

[genetix]

Try the other timing as well if tRFC- 195ns alone is no good.

GSKILL SUPPORT

Are you man serious..

We've been tweaking this memory past week and ended up on 5-6-5-17. Read the damn topic through then drop advice. last one advice was 127.5ns(s) / 195ns(d) to balance the channels to be better tweak ability. benchmarked it on 800Mhz -> 1066Mhz on and trust me 5-5-5-15 ain't exactly flying here. 'Other timing', if you refer I dumped 3 times and he tested secondary timings which all were good in the first place, but simply for tweaking.

I mean not that I give a damn, but this is now like 6th topic you replicate exactly what I've written above. Which is kinda getting hilarious.

Here's where we are now:
http://i26.tinypic.com/2464akg.png

Full timing List exception of below:
http://i29.tinypic.com/2lw7x5c.png
* with atm, change of tRFC to above & tRC is 31T.

----

To get back to topic. Try adding tRRD to 4T and if there is 'Read PRE Time' somewhere add that as 3T. Drop tWTR(write to read delay) to 6T, if you can and if there is an tRTW(read to write delay) try that at 7T (if there is separated like tWTR/tRTW different/same row/bank let me know before doing this).

[GSKILL TECH]

You said you were no longer helping him, so I decided to join in and figure out what's going on. Although he did try close settings, I don't think he set everything correct, which is why I am simply stating what I think his computer is asking for.

I know that 5-6-5-17 tRC 31T setting is the best as of now... but his tRFC was off. You did tell him to change it, but did he?

"tRFC was set to 127.5, but I will take your suggestion and set it to 195."

Obviously not. So where am I wrong by repeating?

I may repeat things or ask the same things, but it's my way of making sure and figuring out what the situation is. If you are not satisfied, that's too bad, because you are not the one in need of help. I will get things done my way, and input advice my way. You can help him your way. I don't see him complaining about me repeating.

Plus, we should be repeating the same things if we are giving out the correct information. That's why sometimes I don't respond because you do a good job giving out the right information.

GSKILL SUPPORT

[genetix]

I was not trying to be offensive.

Just though it was pointless as you know the actual specs of this memory and best advice was to "Try the other timing as well if tRFC- 195ns alone is no good." which we "all" see from specs. You could of simply dump him the values accordingly to timing 1.90ns to this board & rest of timings as I do not have these specific blocks on hands and cannot check SPD data.

Repeating information is useless with this kinda of fellow who understands 4 pages of testing before hand I'd say. Was pointing out with screenshot we can see and check what is wrong you can clearly see what has been try'd out.

[retsamradassaT]

"All tRFC is minimum of 127.5ns (Best would be, if you use DIMM 1 & 3 as 127.5ns and DIMM 2 & 4 as 195ns.)"

I did had my DIMM 1 and 3 set to 127.5, but my 2 and 4 were at 195, so that's why I said set it to 195, because the other two weren't set to 195.

Anyways, here's the result: two core just failed after 40 minutes with the 5-5-5-15 timing.

Now, I am gonna try where I left off last time and re-adjust my settings to the genetix recommended, and then report back!



btw,
Thanks for the help guys.


Edited: there is no tRTW(read to write delay)

Edited: Okay, 3 cores have failed within the first 10 minutes of prime testing.

[genetix]

"All tRFC is minimum of 127.5ns (Best would be, if you use DIMM 1 & 3 as 127.5ns and DIMM 2 & 4 as 195ns.)"

I did had my DIMM 1 and 3 set to 127.5, but my 2 and 4 were at 195, so that's why I said set it to 195, because the other two weren't set to 195.

Anyways, here's the result: two core just failed after 40 minutes with the 5-5-5-15 timing.

Now, I am gonna try where I left off last time and re-adjust my settings to the genetix recommended, and then report back!



btw,
Thanks for the help guys.


Edited: there is no tRTW(read to write delay)

Edited: Okay, 3 cores have failed within the first 10 minutes of prime testing.

This most be something else. While you check memory test through mnimum voltage you get it straight there plus +0.06 = Windows Stable memory.

The Prime95 as well as LinPack and PI depends on that also all depends how Windows API allows them to use resources so make sure after changes you made the 'Windows Performance'-update and, if you are sure that memory is MemTest Stable then start looking cause from NB/CPU voltages and the HT. I don't think it is really the memory.

Well, this is very complicated issue since, if it's something like too high 'NB' or 'HT' this would require me to know how your PSU Rails are hanging on as said before really cosmetic stuff. Anyway, for now make sure memory is absolute stable and add it some more voltage (stay on between 2.2-2.3v it won't need more) and check CPU voltages "a hinch" higher higher the core voltage is (better stick below 1.5v limit shouldn't need this much almost in anycase) basically that could balance some.

Perhaps google some voltages etc etc what other overclockers use on similar CPU or motherboard in general would help a lot.
http://www.extremeoverclocking.com/r..._X4_955_3.html <- 3,2Ghz on 1.350v for example.

anyway, as AMD is so tight together it could literally be anything unbalance on my eyes.

[retsamradassaT]

NB/HT are on auto.
I have just increased my CPU voltage to 1.4 and my memory to 2.3, hopefully I will get better results.

Edited: seems like increasing the voltage didn't helped at all.

My computer could pass prime95 test if I adjust my memory to run @ 400MHz instead of 533, so it seems like it's a memory problem. Just don't know what is wrong with it.