Help! Asus A8V deluxe rev2 + Crucial Ballistix problem...

[bo3bber]

More results! I was wrong...

Looks like I was wrong with respect to upping the RAM bus to more than 200Mhz. My theory was that the A8V was holding us back on the RAM bus. But, I am presently running at 230Mhz, for 12 hours in Prime95, and 4 hours in Jedi (max temp 52c). No errors.

CPU: 2300Mhz, HTT: 230Mhz, HTb: 800Mhz (4x), RAM: 230Mhz (1:1).


The difference was that I manually set the timings for the RAM, and didn't let Auto choose apparently bad numbers. I also specifically set them to the slower speeds specified in the SPD. My RAM is Corsair XMS (TwinX1024-4000PT). Not premium RAM, but not value RAM either. SPD timings specify 3-4-4-10 at 250Mhz. I'm running at those latencies, even though I'm only at 230Mhz.

I do pay a bandwidth penalty by going this route, instead of staying at 200Mhz, Sandra reports 5367MB/s instead of 5897MB/s. Of course, I'll be pushing it up to 250 if I can, which should recover the bandwidth.


I know that my CPU can go to 2400, but no higher, so I will try to get there by upping the 1:1 bus to 240.

Even given this result, it's not clear that the overall performance is better. There isn't much point in going to higher data rates, with longer latencies, the expected bandwidth is the same. And, there is no advantage to a high HTT, so going with my prior setting is essentially the same. The 1:1 bus setting is important on Intel, but irrelevant on Athlon64, so using RAM ratios is not a problem.

However, if I can get past 2400 on the CPU using this high latency approach, I will let you know.

-bo3bber

 

[adeathangel]

Hi all. A very interesting discussion indeed. I have an A8V w/ a 3200+ Winchester, and I've had exactly similar problems when it comes to ram. Although I made a really bad decision going with Kingmax PC3500 (4 256MB chips) I haven't gotten my ram stable past 200MHz. Currently I'm running CPU: 2375MHz (9.5x multiplier) , HTT: 250MHz, HTb: 800MHz (4x), RAM: 125MHz @ 2.5-3-3-6 1T (the DDR200 in the BIOS) ... I'm basically starting over with testing the ram.. I'll let you guys know what I can accomplish, although I doubt it'll be much with four sticks. Going to loosen the timings and see what divider will work.

 

[adeathangel]

Alright, the results are in. I couldn't get anything to run at a somewhat stable state until I upped the VCore to ~1.6V ... I thought that was a bit high for a Winchester, so I don't know how to explain it. A curious observation I made is this.. when I change the VCore in the BIOS to anywhere from 1.400V to 1.550V, the BIOS as well as CPU-Z report the VCore to be ~1.478V, no matter what setting I have on it. And when I set the VCore to 1.575V or higher, they report the VCore at ~1.64V.. So my question is this the BIOS that's screwing up or what? I can see why I couldn't OC to a higher freq when the VCore was at 1.478V.. at least, that seems like a plausible reason why.. but 1.6V+ seems kind of excessive. So there's the VCore situation, the vdimm is at 2.8V and the Hyper Transport I set to 3x (600MHz) with the HTT at 240MHz. The CPU multiplier is 10x giving me 2400MHz clock speed. I have my ram at the loose SPD timings, 3-5-5-10 1T, and it passes memtest86 no problem with the 5:3 divider setting it at exactly 200MHz. Unfortunately my idle CPU temperature is kind of high at 42C with the stock cooling, and I suspect it's because of the unusually high VCore of 1.648V that's currently being detected, check the screenshot link below. Prime95 ran for 20 minutes no problem before I stopped it, figuring I'll stress it for 12 hours later. Unfortunately, I don't think this is a stable set-up because the OS keeps crashing every 20-30 minutes or so. It also crashes when I reduce the HTT to 230MHz and leave everything else the same. So I want to know a few things,
1) What could be up with my VCore?
2) Aside from lowering and lowering the OC until it's stable.. is there anything else I can dabble with to attain stability? For now I don't want to buy a different cooling system, I just want to OC to 2.4GHz, maybe 2.5GHz if it's possible.
3) Do you have any suggestions for what else I can be doing, or am perhaps doing wrong, or could do differently?
4) The screenshot also has the memory bandwidth benchmark from Sandra, is that a decent score?

HERE IS MY SCREENSHOT

 

[Slug]

My vcore never seems to agree with the bios setting.

 

[lauy]

hi adeathangel

I think the "Vcore" problem that you are describing was fixed in a later version of the bios. The 1009 final bios is actually released now so you can grab it on the Asus site.

About the instability... Is it completely stable when NOT overclocked? because I had instability when running stock and it turned out to be my video card being incompatible with the motherboard. I actually fixed it by turning to AGP 4x and fast writes off and also increasing AGP voltage to 1.6 . But...if you don't have this problem on stock then just ignore this.

I am still unsure whether it is the actual memory that is limiting me or the board!! It seems that we all have a common denominator of the board....I was tempted to buy some new memory (OCZ pc3200 plat rev2) just to find out but have left it for now until I can be certain.

KheeWhee - if you are still reading discussion...what is highest overclock of your OCZ PC3200 plat rev2 ram at 1:1 ratio at 1T ? If you can achieve speeds of above 250 @ 1:1 ratio (don't use a half multiplier since this misleadingly lowers the memory speed!!) ...then it is probably safe to say that it is the Ballistix memory that is the problem. If not...then maybe it is the board?

One thing I can ask though is...mainly to Slug and others with the Crucial Ballistix memory. What timings is the memory set at and what speeds exactly. I see Slug has his CPU at 2430mhz so I'm guessing he has it at 10 x 243. But please give timings as it would really help! I can't get my ram past 230mhz still at 1T.

An important factor is whether it is at 1T or 2T. The performance drop from 1T to 2T is immense! After some more testing, I discovered that I could run my memory at 250mhz @ 2.5-3-3-8 but 2T rather than 1T. It turns out that I get a score of 6950MB/s on sandra with memory at 230mhz @ 2.5-2-2-5-1T and I get a much lower score of around 6000MB/s with my memory at 250 @ 2T. Whether or not Crucial's rated speeds are at 2T rather than 1T I don't know...but all other major brands I think are rated at 1T.


Finally...all that I'd like to find out as well as probably all the other ballistix owners out there that are having the same problem... IS IT THE RAM OR THE MOTHERBOARD?!?!?!

Thanks for keeping thread going!

 

[Slug]

Yes, I am at 10x243 HTT. But still running the divider on auto, giving me a mem speed of 202 @ 2-3-3-8-1T. I have been stable at 237 x10 at 1:1 for 2370. Timings were 2.5-4-4-10-1T. Sandra memscores about 600 points higher in the latter setup. I just like the magic number of 2400 MHz These settings are all prime stable overnight.

As for ram vs. mobo, I dunno. I tried some Geil ultra platinum PC3500 with markedly worse results. I don't have anything else to test with. I'd love to try some TCCD.

Is it possible this is a vdimm problem, ie- lack of?

Oh, and I've had vcore as high as 1.65 but it did't seem to help much. Currently at 1.525v. My cpu may just crap out at 2440.

 

[KheeWhee]

Hi Lauy,

 

[KheeWhee]

Well Lauy, with my board and 3500+ Winchester, my OCZ platinum Rev 2. i easely achieves 248-249 fsb at 1:1, but it seems to crash in 3d games for unknown reasons. In normal work or memory intensive programs, it works fine with absolutely never a hickup. This is at a timing of 2.5 3 3 10. There is no point to go at looser timings. I tried at 3 3 3 10 and did not achieved much better than 250-251 fsb and lost at sandra.
At 2 3 3 10 it achieves 243-244 fsb very stable.
The fastest stable oc that i got is 10.5*242 at 1:1 timing 2.5 3 3 10 with a reasonnable cpu voltage of 1.5 volts.

 

[adeathangel]

Damn, that's a sweet OC. I wish I could OC that high and be stable. Too bad I have all 4 dimm slots populated with value ram -_-...

 

[bo3bber]

adeathangel; as lauy noted, you want to be certain that your system is stable without any overclocking first. Once that is true, I would recommend going about it the other way, i.e. keep it stable, and keep pushing up parameters, rather than trying to find out what is making it unstable at a given set of settings.

The reason is that your instability could be caused by multiple things. Hypothetically let's say both the CPU and RAM are pushed too far, if you only back off on one, you still have instability, and hence won't think that parameter mattered. If you go about it the other way, starting from stability, if you push up a single parameter and it falls over, you know that was it.

This is especially true, because the parameters interact, making the problem much, much more complex.

-bo3bber

 

[bo3bber]

In answer to Lauy's question, of whether it's the board or the RAM, I still believe it's the board.

I had previously thought that I could not exceed 200Mhz, but that was an error. I can exceed 200Mhz, as long as I accept the slow timings of my SPD of 3 4 4 10. Since the RAM is rated at 250Mhz, if it were running that fast, that would not be a problem. I would have the same Sandra score as 200Mhz with tighter timings.

When I run loose, I can bump up the RAM bus speed to a maximum of 230Mhz. Anything higher gives me errors when playing 3d games. From my tests, it appears that Prime95 stresses the CPU mostly, and the 3d games stress the memory subsystem mostly. Running stable in Prime95 is not sufficient to demonstrate stability, as I had numerous times where I could run Prime95 for 12 hours, but crash within 30 minutes in Jedi Academy. This is similar to KheeWhee's experience too.

Even at 2.8v, I cannot push the memory bus faster than 230Mhz though, which is a drag, because I'd bought my RAM hoping to be able to run at 250 with a 1:1 ratio. That isn't going to happen.

The best I can do, is to run my RAM at 200Mhz, at 3-3-3-8-1T, which produces a Sandra of 5987. Not killer, but respectable. That's with the CPU at it's maximum OC of 2400Mhz too. Anything higher than that falls over, even with CPU at 1.7v.

So, my maximum overclock appears to be:
CPU: 2400
HTT: 240 x10
HTb: 800 (4x)
RAM: 200Mhz from 5:3 divider, at 3-3-3-8-1T.

My CPU is limited to 2400, so I'm good example of not being able to go higher, regardless of the RAM settings. No complaints.


I still believe the board is limited in the RAM speeds that it can support, specifically it appears to run into trouble at about 230-240Mhz. This shouldn't be terribly surprising, that's still 20% more than the rated speed of 200Mhz. I looked on other forums as well, and haven't found anyone that can successfully reach 250Mhz for their RAM clock, regardless of RAM quality.

Also, in order to reach that 230Mhz it appears that you need to either have premium RAM, or loosen the timings substantially. Premium RAM will provide you with a nice boost here, as Lauy, and KheeWhee have seen. If I were doing it over, I'd buy premium RAM, not higher clocked RAM like I did. (I was looking to the future, but you can also do that with better RAM).

Loosening the timings is pointless, except for experiments. In terms of throughput, you aren't gaining anything.


Thanks to everyone for your experiences! This has been very helpful to understand the limitations, and see what reasonable expectations might be.

-bo3bber

 

[adeathangel]

Yes, I know. My system does indeed run stable stock =P... I've been too busy playing games to really start pushing the parameters from stock by 5MHz and testing completely etc.. Maybe this'll be a good project when I'm on Spring Break.. In the meantime, gotta take pics of my Asus V9999GT in preparation for selling it! Looking at eVGA 6800GT, haven't heard any complaints from that card.

 

[CrucialLabs]

This thread should answer most or all of your questions.

http://forums.anandtech.com/messageview.aspx?catid=29&threadid=1500007

 

[Waylay00]

Sorry if this is a little off topic but, if I get some PC4000 Ballistix (2x512) will I be able to run them at DDR400 2-2-2-5?

 

[lauy]

I think I might have to agree with bo3bber that it is the board that is the issue.

To KheeWhee, if you are using a half multiplier like 10.5, your memory will be running at a slower speed than you believe even though it says that it is running at the right speed in programs such as CPU-Z. So your 242mhz will not actually be 242 but something lower. Try running at a 1:1 ratio with a full multiplier like 10 or 11 and make sure 2T is disabled so that it is running in 1T and see how far you can reach. Test it with Memtest86. For my Ballistix, it is always test #6 that it fails on.

I think that maybe a good bios update might be able to fix these nightmare problems. Hopefully anyway.

Right now, I'm quite content anyway with 11 x 230 = 2530mhz and sandra score is at 7000MB/s, but it has hit a barrier that isn't necessarily the limit which is what is so annoying!

In response to Waylay00, Yes. My PC4000 ballistix modules do run at DDR400 @ 2-2-2-5-1T.

 

[Waylay00]

Is is stable?

 

[Slug]

Yes, mine are stable at these settings as well.

 

[Slug]

A minor breakthrough today. After some reading, I changed the mem setting in BIOS on the second-to-last menu called READ PREAMBLE SETTING. It's set at 9.5ns (haven't tried anything lower yet). It has allowed me to be prime stable (for a short time anyway) at 280 HTT. Still can't run 1:1 with RAM at 250, but it's a start. All other mem settings are at auto, even the divider. Seems any time I manually set the mem divider, I get lockups. I'm currently prime stable for about an hour (also with Doom III running).

Settings are CPU 2412 MHz, 9x CPU multi, 268 HTT (at 3x--haven't tried 4x yet), mem at 219 MHz (auto divider) @ 2-3-3-8-1T.
CPU 1.55v, mem at 2.8v, AGP at 1.6v, Vlink at 2.6v.

I also haven't done anything yet with the ASYNCLAT setting.

 

[DPaulson]

I just ordered the A8N-SLI and a bunch of other goodies. Was looking into some specs and performance tests when this forum came up. I want to overclock everything i can for i got the case and cooling to do so, but don't want all of these problems (of course). I will test many things and keep up with this topic. I'm not using ballistix memory, instead i'm using a pair of corsair 512mb 3200XLPRO. I see that the ultimate question (board or memory) hasn't been answered so maybe i can help by determining if it is the board or not since i'm not using the same memory. Will do some OC tests when i get everything together and running.

 

[Slug]

I'm now at 240 HTT and mem for 1:1 and 2400MHz (about the limit of my CPU). Used the READ PREAMBLE SETTING as in my above post. 2.5-4-4-10 timings and DDR400 divider set in BIOS 1010.003 beta. Enable 2T command at auto for 1T. Testing Prime and F@H stability right now. So far stable with those running as well as multiple DIII timedemo loops at the same time. This has never been stable with this hardware before. Sandra memscores increased an additional 600 pts over baseline to 6588MB/s.

EDIT: Changed ASYNCLAT to 8 and Trc to 12(auto was 11) after random reboot. Now Prime and F@H stable for 2.5 hrs so far at 1.575v.

 

[lauy]

Yeh. Mine is fully stable at those settings I said in my previous post.

 

[KheeWhee]

Lauy, i can hit either 10*247 stable even in games (1.5volts on cpu)(SuperPi 35 sec.).Heat is only at 39-43deg.
But at multiplier 11, it is not as great 11*230 only (1.6 volts on cpu and quite a bit more heat, anyway more than i like 50to55deg.)(superPI 34 sec.).At 1.5 volts, it just does not want to go higher than 222*11.
The memory is at 2.5 3 3 10 1T (1:1)1.7 volts in both cases.(for some reason at 1.8 volts it just does not boot at all with my memory.)
This seems to confirms that our problems are board/cpu connected, not memory.
So, you should be happy with your 230*11 at 2.5 2 2 5 1T timing.
With the MSI, people are easely going to 300 fsb.
At what speed is your fan working at, the thermaltake i use is at 2500 turns and is not too noisy. I tried a 1400 turns very quiet fan before but the thermaltake took 2 to 3 degrees out besides looking great through the side windows being with 3 blue leds.

 

[lauy]

It probably is the board then.

Is your 10*247 @ 1:1 at 1T ?? Because I can't seem to get that even on 10 multi.

My fan is working at 1600rpm because it is an ultra quiet one but the airflow it creates is more than the standard akasa fans with leds. I did have a blue led one.

Right now at 11*230 my temp is quite high :| idle around 40-45 and under load is up to 55c. It is alright but a tad high considering I have an XP120. But it is fully stable and i've been using this setup for quite a while now without problems, so I'll keep it at this.

 

[KheeWhee]

Yes it is at 1:1 at 247*10 and very stable.

 

[bo3bber]

:disgust:

The question has been, is it the Board or the RAM? I think I have a pretty solid answer: neither. Allow me to explain.


I wanted to see if anyone was having luck with the A8V, so I did some highly targeted google searches to see if anyone has much success. The answer is yes:

http://www.abxzone.com/forums/showthrea...370&page=7&pp=15&highlight=a8v+corsair

That's a mighty big thread, but there are multiple people there who have success getting their RAM up to 245Mhz or so. It's also worth noting that KheeWhee has a solid configuration at 247Mhz.

So, that's an existence proof that the board can overclock the RAM, and run solidly. So, it's not the board.


Secondly, I wanted to see if my RAM in particular wasn't good enough to reach these numbers.

My Corsair XMS4000 contains Hynix BT-43 RAM chips, which successfully overclock to 530Mhz on other boards. These are the same chips used in some very successful Geil and OCZ RAM sticks, so this is not sleazy RAM.
See Anand's review at:
http://www.anandtech.com/memory/showdoc.aspx?i=1849&p=8

For Lauy's question about Ballistix RAM specifically, Ballistix is also high quality RAM, there is no question it will run at the rated speed on other boards. Ballistix PC4000 contain Micron 46V32M8 -5B G chips.
See Anand's review at:
http://anandtech.com/memory/showdoc.aspx?i=2145&p=18

That review compares quite a number of RAM, including Ballistix 3200, which reportedly use the same Micron chips as the 4000. Also of note on that graph is that my Corsair XMS went to 530Mhz at the max. Crucial 3200 went to 514Mhz.

The RAM we are all using is unimpeachable. So, it's not the RAM.


Hence, my conclusion of neither. With a lean toward the board being at 'fault' because most of us are overclocking the board, its spec is only DDR400.

When reading that first thread, it's clear that there are specific RAM/Board combinations that are more successful than ours. For example, the OCZ PC3200 that KheeWhee uses is successful more often than not. The Ballistix 4000 version is not successful very often. Same with my Corsair XMS.

Here is a thread talking specifically about which RAM chips work best for A8V:
http://www.ocforums.com/showthread.php?t=332591

Thus, the limiting factor is really that some RAM just have minor incompatabilities with the board, and thus cannot overclock well. They will run fine at the rated speed of the board, but have no headroom. This is the fundamental premise of overclocking of course, no guarantees, just lots of guesses.


So, to get those higher numbers, you either have to change RAM, or change your board.

One last possiblity is that we aren't doing it right. It's clear that the Auto settings aren't particularly smart when it comes to the RAM. Slug has demonstrated that he can get to 240 with Ballistix, by tweaking the Read Preamble and AsyncLat settings. This didn't work for XMS RAM though. Slug, how confident are you in your current settings?


I cannot get more than 10% out of the board, with my RAM (220 Mhz maxes out). But, my CPU goes to 2400 from 2000, which is my main concern. Still, KheeWhee and others can get about 25% more out of the board, by using the OCZ RAM. From this board's perspective, that RAM is superior.

From my perspective, the added headroom isn't enough to justify the conversion just yet. I'm probably going to move to a PCI express SLI board anyway, and will try to use my XMS Ram there. But, I may just get a new AGP video card when the prices drop instead.

The ASUS is a good choice from the perspective of being reliable and well supported. But, I'd say it's clearly not the preferred overclockers board, regardless of the reviews. If you have exactly the right RAM, it's fine, but that wasn't at all clear when I bought the board... To be blunt, we chose poorly.


There is a chance that a BIOS update will fix the problems, because it's possible that the OCZ work well simply because they match the BIOS settings better. I'm not holding my breath.

If you want to swap RAM, the ones that seemed to work best were the OCZ 3700EB, and OCZ Platinum 3200 Rev 2. There was a lot of propaganda regarding other RAM, these were two that stood out as working for people who clearly cared about 100% stability.

I will play with the RAM settings a bit more, but can't really justify much more time for a 10% improvement. Now that my RAM/overclocking education is almost complete....


Please let me know if you think my conclusions are wrong. And hey, hope this helps.

Thanks,
bo3bber