How am I doing so far? (ballistix 2x1gb)

K

[k]ar|

Gawd
Joined
Jun 17, 2002
Messages
561
I decided to splash out on the Ballistix BL2KIT12864Z503 (2 x 1Gb), and ATM I'm seeing how high it will go before I start on the FX (I'm following the [H] A64 OC'ing guide for n00bs :rolleyes: ).

I just finished my first runs, using MemTest to test for initial stability. I cranked my CPU multi down to 10, set HTT at 4x and locked the PCI at 33 & PCI-express at 100. I left the memclock at 400, dropped to 1t command rate, but left the latencies at stock 3-4-4-8 for now.

I just ran 2 passes of MemTest without errors at 255 (@ 2.8v) which looks OK for a start - I tried for 260 (@ 2.9v) but MemTest threw up a lot of red :eek: .

I'm still a bit new to this whole A64 OC'ing thing - is that an OK result? I was hoping to get a bit higher, but in a brilliant move :rolleyes: I went for the Asus mobo rather than the DFI. Any tips on how I can squeeze a bit more out of my ram? I rigged up a temporary RAM fan but that didn't seem to do a whole lot.

Thanks in advance!

[k]
 
thats a decent result for the 2x 1gb crucial ballistix. i have seen some people who got 20-30mhz more out of em.
 
Change the command rate from 1T to 2T and it should overclock much higher! I also have an ASUS A8N-SLI Deluxe mobo and Ballistix (2x512MB) ram. This mobo has a well known bug with 1T not working at clocks higher than 240-250. ASUS has not fixed this yet.
 
Thanks for the comments!

Looks like 250 @ 2.8v might be the maximum for me as well. 255 wasn't prime95 stable, even though MemTest ran through twice without any issues. Upping the voltage to 2.9 didn't help either - I backed off to 250 and running the Prime95 torture test and SuperPi 32m simultaneously for 10 mins, things initially look to be OK.

I must say I'm a bit disappointed - I've seen some people getting up to 300 out of these sticks, albeit on DFI mobos. Aside from dropping to 2T is there anything else in the BIOS I can try? I don't want to sacrifice too much overall system performance just to increase the RAM's e-penis :D .

If the Vapo & FX-57 can cope with it, 14x250 = 3.5Ghz isn't too bad I suppose. Although the chipset was getting mighty warm (too hot to touch) despite the fancy heatpipe cooler. Might have to do something about that next!

[k]

edited to add - ran with Prime95 torture test and SuperPi 32m calc simultaneously for about 2 hours. No errors from Prime, but SuperPi threw up a couple. Not too sure what this means? I always thought Prime was *the* thing for testing stability?
 
SuperPi is very sensitive to memory timings. I would suggest you play around with the timings.

I have 2x512MB Ballistix and my most stable max on them (passing SuperPi and Prime95) is 250MHz also with the following timings: 2.5-2-2-8-12-14-2T

I also am considering getting a DFI Lan Party SLI-DR mobo if ASUS doesn't fix this 1T bug.
 
you would be surprised, but the Ballistix actually OC better with lower voltage. I am stable at 270 w/ 2.7vdimm, but not at 2.8vdimm. When reading on other OC databases, you will find that alot of them are getting good results by lower volts. Try it! Take it down to 2.7vdimm and give it a go...

http://www.xtremesystems.org/forums/showthread.php?t=67762
 
yeah - I was going to say, that seems like a little too much juice... seems like a lot of these 1gb stix like lower volts when OCing... and I am completely ok with that. :)
 
Thanks for the suggestions - I'll drop the voltage and try PRCop's timings. Unfortunately, I have a huge pile of work to do this evening, so it might have to wait until weekend, but I'll try a few more runs.

I'll report back on my progress!

Thanks again everyone,

[k]
 
the 1gb and 512mb sticks of ballistix use different chips.. usually. micron 5b D and 5b G rspectively.

the 5b D that's in the single sided 512mb and 1gb crucial sticks typically doesn't scale much over 2.8-2.9v, and it puts out a decently large amount of heat. putting a fan on them to keep it cool could probably be the best thing you can do for that stuff

also, it likes somewhat looser timings than the 512mb sticks, like 3-3-3
 
OK - I tried dropping the voltage to 2.7, but MemTest started throwing up errors, even at 250MHz. I've therefore gone back up to 2.75v and this seems to be all right with no MemTest errors after several passes. I've previously ghetto-mounted a fan for the RAM, but unfortunately that didn't seem to help my OC.

So it looks like I will have to delve into what is for me the murky world of latencies, which TBH I don't really understand, despite reading the FAQs :rolleyes: . Here are my current settings from teh BIOS:

[HTT frequency = 250; DDR voltage = 2,75v; CPU multi = 10x]

Memclock index value = 400MHz
CAS# latency (tcl) = 3
Min RAS# active time (Tras) = 8T
RAS# to CAS# delay (Trcd) = 3T
Row precharge time (Trp) = 3T
Row cycle time (Trc) = 11T
Row refresh cycle time (Trfc) = 14T
Read-to-write time (Trwt) = 5T
Write recovert time (Twr) 3T
1T/2T memory timing = 1T
>4Gb remapping S/W & H/W = Enabled

Any ideas what I should be looking at in terms of improved settings? From what people have suggested earlier in the thread, I'm already running at CAS-3, Trcd-3, Trp-3. Any other setting changes that might assist?

Thanks again!

[k]
 
how much voltage can you give them? my 5b D based crucial value peaks out with 2.9v
 
I took them as high as 2.9v trying to stabilise at 260, with active cooling from a CoolerMaster 80mm Aero blower, but neither helped and I had to back off to 255 (not 100% stable) then to 250 (appears stable at 2.75v). I've run about 5-6 passes of MemTest without errors, but I've not yet tried my previously used test of simultaneously running a 32m SuperPi calc and Prime95 torture test run. I'll set that going now.

I didn't really want to blast the Ballistix with 3v in case something caught fire :eek: :D. The NB is already getting much hotter than I'd like, to the extent that I'm seriously considering ripping off the heatpipe thingy and replacing it with a DangerDen chipset WC block.


[k]
 
changing ram voltage should only effect the ram. if it makes the chipset hotter.. well your mobo has issues :p

also.. try changing Trcd (ras to cas delay) to 4, and seeing where that takes you
 
Lol - just commenting generally on the quite alarming amount of heat this mobo is pumping out. Whilst the heatpipe is all very clever, the NB is literally too hot to touch, and therefore makes me concerned that spontaneous combustion is imminent :eek: :D

I've just run my stability test - 32m SuperPi and Prime95 torture test simultaneously. The Pi calc went through in 34mins 49, and I interrupted the Prime95 torture test at 2 hours 16 mins with no errors or warnings. Looks like 250 at 2.75v is OK.

I'll have another go tomorrow with the changed Trcd setting. Thanks for the tip :D

[k]
 
is your case W-atx? or is your mobo upside down like on of those Lian-Li cases? that would be causing your heatpipe cooler to not work. and im guessing you might be able to eek a bit more outta that ram, i think it has it in it.
 
CCUABIDExORxDIE said:
is your case W-atx? or is your mobo upside down like on of those Lian-Li cases? that would be causing your heatpipe cooler to not work. and im guessing you might be able to eek a bit more outta that ram, i think it has it in it.
Click to expand...
My rig is built in a Lian-Li PC-75, which I think puts it "right way up" for heatpipe operation. However, I remain unconvinced of it's effectiveness and I am considering whacking a DangerDen waterblock on there. Might be a bit of a squeeze once the tubing for the 7800s gets in there along with those for the NB block, but it might just make it.

Back OT, I'm about to do another run with the DRAM setting changed so Trcd=4. Hopefully this will make the difference. :eek: :D

[k]
 
Well, that didn't work :(

I tried changing the Trcd to 4, then HTT to 255. Memtest threw a fit, so I tried bumping the volts back up to 2.8. Still unstable in MemTest. So I went back to my "last known good" settings (as above), dropped to 2T and gunned it using ClockGen whilst in Windows, to see if that was the issue. The whole thing locked up at about 260.

For whatever reason, this mobo *will not* go any higher than a HTT of 250 :confused: :(.

[k]
 
[k]ar| said:
Looks like 250 @ 2.8v might be the maximum for me as well. 255 wasn't prime95 stable, even though MemTest ran through twice without any issues. Upping the voltage to 2.9 didn't help either - I backed off to 250 and running the Prime95 torture test and SuperPi 32m simultaneously for 10 mins, things initially look to be OK.
Click to expand...
Remember that the memory controller is electronically separate from the rest of the chip, to the point where I believe they run off of two separate voltages. If MemTest passes at 255, it means that both your RAM and your memory controller were fine at this speed. If Prime failed, it means that the processor itself (or chipset) is unstable. Try feeding a little bit more voltage through the CPU, leaving your RAM at 2.8V. You might also want to try dropping the HTT multiplier to 3x instead of 4x, since 4x255 is still technically over 1GHz.
 
You night also want to use a divider for your RAM because HIGH cpu clock is what really matters with A64. ;)
 
I can't get my HTT over 250 either... I have the same ballistix but I have a DFI nf3 250gb...

I have tried everything but have no clue what is wrong. Even using a divider doesn't help.
 
Yashu said:
I can't get my HTT over 250 either... I have the same ballistix but I have a DFI nf3 250gb...

I have tried everything but have no clue what is wrong. Even using a divider doesn't help.
Click to expand...
i had that problem once. it's your memory controller.. as my clawhammer hates going over ~250mhz, but with my newark, i can hit >280mhz without problems.
 
Well, seems like I'm not having any luck at all with this OC'ing lark :rolleyes: .

I tried dropping the CPU multi to 9 and the HTT multi to 3. Still coming up with errors in MemTest at anything over 250.

I decided to try to concentrate just on breaking that 250 barrier so I tried again with [email protected] with HTT back at 4x and the proc at 10x multi. I must have gotten lucky last time, as MemTest threw up errors this time round. I tried again at 2.75 to see if that will help, but that didn't work either. Just to be on the safe side, I'm trying 2.7 but I'm not hopeful.

Guess I'll have to try a divider next to see if that will work. I've just had a quick look at the guide to this on www.eclipseoc.com showing how to work everything out. Any sugestions as to any particular dividers I should try first?

Thanks again to everyone for all the suggestions and info so far :D

[k]
 
I get my second stick tomorrow, Im excited. First stick did 260 with no problems, didnt try any higher. Hope the second does the same.
 
hehehe I have a newark...

it is because I am using two dimms...

I can get over 250 HTT using 1 dimm but not two...
 
Yashu said:
hehehe I have a newark...

it is because I am using two dimms...

I can get over 250 HTT using 1 dimm but not two...
Click to expand...
same exact problem here actually. maybe it's just a board quirk? :eek:

(and sorry, you'd think i'd remember you having a newark since you're one of like.. 3 people with a working one these days)
 
Got my second stick in today as I said. Stuck it in, left the settings as they were (250) and it booted right up. Ran memtest in the bios for about an hour, no errors. Im at 2.8v, 10x250, 3-4-4-8-1T for now. Plan to try higher in a while. I was at 10x260 with my Redlines, I would like to at least get back to that.
 
Good? Its what they should be at.. they're rated for DDR 500 :confused:

Ill try higher later.. after the kids go do bed.. if the game doesnt distract me too much.
 
Had a little extra time, so I pushed the ram some. So far, Im at HTT 4x, 10x265@1:1 3-4-4-8-1T, 2.8v.

Ballistix-265fsb.jpg


1920x1200 500k pic; http://home.comcast.net/~denverspictures/7800GTX/Ballistix-265.JPG

Its a lot slower in these synthetic benches. Everest I was at 36ns I think, about 10ns lower than these Ballistix, with the Redline at just 260. Sisoft is about 1000 points slower with these Ballistix, again with the Redlines at just 260. I dont care too much about the tests though, as they dont often reflect game performance, at least for me.

For hardware, Im on a DFI NF4-D/623-2 bios/yellow slots, 3700+ SD, 2x1gig Ballistix PC4000, PC P&C 510 SLI.

I ran a Farcry benchmark looped a few times, seems stable to me. Havent run memtest, guess Ill do that later. Ive got great active cooling on the ram, maybe thats helping me. I may try for more later, but Im pretty happy where I am at. I wanted to be at FX-55 speed, and so far Im above that a little.
 
fallguy said:
Good? Its what they should be at.. they're rated for DDR 500 :confused:
Click to expand...
my bad, i was retarted and thinking in the crucial value mindset... and kinda forgot you have ballistix :p

try 3-3-3 timings ;)
 
fallguy said:
So far these sticks are going beyond my expectations.
Click to expand...
not past mine :D
i would expect 275-285mhz, tweaked to the max. i've seen em do 300mhz at 3-3-3 :cool:
 
Oh well, it doesnt really matter if you're happy with them or not. :p They're rated for 250, they're at 265, at lower than rated timings. I havent even played with them much yet, they've done everything Ive thrown at them so far.

I may try more later, but I doubt Id get much more, if any better performance in frames. I dont have to run everything on the edge.. its plenty fast enough.
 
maybe I should invest in a dual channel X2 system soon... but it is hard to justify the "upgrade" when I have one of the fastest single core systems you can get (socket 754 or otherwise)

Seems the 250HTT limit is a single channel thing with these sticks... (woot s754)...

oh well...
 
Yashu said:
Seems the 250HTT limit is a single channel thing with these sticks... (woot s754)...

oh well...
Click to expand...
nah, dfi thing, as i never had the issue on my k8n neo
 
A quick update with no pics. I dropped the multi down to 9, and left everything else the same. Tried 270, worked fine. Tried 275, still worked fine. Timings are still at 3-3-3-8-1T, 2.8v. Ill try to go a littler higher later.

edit, so I had some more time... Same settings as before, except I upped it to 280, and its still stable.

Ballistix-280fsb.jpg


1920x1200; http://home.comcast.net/~denverspictures/7800GTX/Ballistix-280.JPG

Ill try for more later...
 
OMG - definitely a mobo issue for me then. Then again, I never one to take the easy option :rolleyes:.

I'm having to work backwards from 255, which is where it craps out for me at 3-3-3-8 & 2.75v. ATM, I'm running MemTest on what I hope is going to be a stable run at 253 :( If I go any higher, I get errors in MemTest, or Windows crashes with a page fault in non-paged area.

Time to strap the Vapo onto my FX57 - see if some additional voltage and cooling will help the mem controller :cool:

[k]
 
[k]ar| said:
see if some additional voltage and cooling will help the mem controller :cool:
Click to expand...
you probably don't want to hear it, but i've seen a lot of cases where cold = bad for the memory controller.. and some people suspect that the memory controller is what causes the cold bug.
 
You must log in or register to reply here.
Back
Top