Giving up, need another geek

_k

I just had to change out some gear that died so I decided to go cheap because it worked for my last folding rig. I put a Biostart A760G M2 board in my old rig with a 7750 Black and 2GB(2x1GB) with a GX2. PSU is one of the older 550 NeoPowers. The system keeps restarting, its headless so I am not 100% sure what happens when it restarts. I have pulled all the fans and HDD off the PSU so its just running the board, GPU and CPU cooler(second PSU jumped on if you are wondering how I did it).

Everything was running fine with a 9600GSO in it folding and SMP client then I installed this GX2 and issues have plagued me.

mirage

Appears to me that either PSU is not sufficient or the GPU is faulty. I would try the card with a higher wattage (on the 12V rail), good quality PSU, if you can just borrow it. If the card checks okay, you can buy a better PSU. If the problem insists with a better PSU, GX2 is most probably faulty.

_k

Its not the card because I just switched my GX2s around in my computers and it is hammering away while the problem system quickly crashed after approx. 20 min. I would guess its not the PSU because that neo is a triple 12 rail with 18A and the other system with the GX2 is lower eff. and single rail at 38A. I have one other thing I am going to try but all I am down to is mobo sucking a hard one and the express bus failing under full load.?.

Mt_Goat

_k_ wrote:

Its not the card because I just switched my GX2s around in my computers and it is hammering away while the problem system quickly crashed after approx. 20 min. I would guess its not the PSU because that neo is a triple 12 rail with 18A and the other system with the GX2 is lower eff. and single rail at 38A. I have one other thing I am going to try but all I am down to is mobo sucking a hard one and the express bus failing under full load.?.

Test the line load on the PSU. If it is up to snuff then the board is as you say "sucking a hard one".

mirage

_k_ wrote:

Its not the card because I just switched my GX2s around in my computers and it is hammering away while the problem system quickly crashed after approx. 20 min. I would guess its not the PSU because that neo is a triple 12 rail with 18A and the other system with the GX2 is lower eff. and single rail at 38A. I have one other thing I am going to try but all I am down to is mobo sucking a hard one and the express bus failing under full load.?.

Triple 18A rails correspond to 648W which is higher than the 550W rating of Neo 550 (12v x (3x18A) = 648W). 18A is the maximum of each 12v rail. The three 12v rails can supply maximum 504W in total (from the specs of Neo 550) which corresponds to maximum 42A (504/12) total on three 12v rails (i.e. less than 3x18A=54A). On the other hand, GX2 requires a 580W PSU. So, I think Neo 550 might not be sufficient for your card. I would check another more powerful PSU before replacing the motherboard.

_k

Yes you never get all the amps on a split rail system. However, not to sound like a complete punk but check out the first three sites and their testing of the actually consumption of the card. My other rig with a GX2 is powered by this 650W Ultra, which only runs a single rail with 38A. I have the 550 Neo hooked up to the other machine to see if it will fail, if it doesn't I am just going to go return and replace.

The other reason I have been leaning to the board was it was replace a 4200+ and there is a difference of +500MHz for this new proc and it folds roughly 200-300 points per day less, which to mean already says the board transit is junk.

mirage

Sure, I just tried to eliminate other possibilities before a MB replacement

_k

Yup that was the trick, cost me another 34$ but it works. Gigabyte GA-MA74GM-S2H is a cheap good board if you need something that will run anything stock.

Gldm

Hate to say it but I'm betting the PSU killed the former board and will kill this one within 2 months. How old is that 550? Have you checked what its voltages actually run at now?

_k

The only thing I can say off the top of my head until I get home is +12 rail sits at 11.84v.

Gldm

That sounds ok, but which rail of the 3? Also the rail sum of that PSU is 705W and it's rated for 550. So if you've got the 12v rails all loaded it may run dry on the 3.3v or 5v. I've had issues with this specific PSU (several of them actually) killing quite a few MBs in the past (including an engineering sample that almost got me in trouble). Plus as a PSU ages its max output decreases (the capacitors don't age gracefully especially on cheaper models hence the whole "solid cap" fad on motherboards now). If that PSU is over 2.5 years old I'd suspect it's the real problem and it will be causing more shortly. The change from the GSO to the GX2 probably caused enough vdroop from the additional load to be a problem for the older board, which likely had its own cap aging issues in addition to the power supply. The thing is bad power will cause problems in new parts more quickly, brownouts are worse than blackouts.

System stability is like a pyramid. The analog input voltage is at the bottom. If it's got issues, everything above suffers. Modern systems have tons of ways to compenstate with things like ECC and busses that can tolerate errors and retry packets and drivers that do exception checking. So things actually can go pretty wrong before any of it really manifests as a critical error. But a lot of times the reason an older machine just "feels slow" despite a reformat and looking fine on benchmarks is because everything's in this constant state of marginal compensation due to bad power. All those little bus retries and resets manifest as perceptible issues but since they're outside the framework they don't show on diagnostics or benchmarks, hence "it just feels slow I don't get it" is a common complaint. 90% of the time this is a power (aging PSU), heat (aging thermal paste), or signal (aging data cables) issue.

Ryder

The 12V all comes from the same place.. all 3 are going to read the same voltage when tested with a DMM.

Gldm

RyderOCZ wrote:

The 12V all comes from the same place.. all 3 are going to read the same voltage when tested with a DMM.

Depends, they all come from the same place, but should have independent current limit circuits. If there's an issue with one of the limiters it may be why changing to the GX2 suddenly caused constant reboots despite the rail being rated for 18A.

_k

^ that really doesn't explain why a mobo and proc change would fix the issue when a new set were already in there. That Biostar is based off the 6100 chipset which is roughly 3 years old which I am attributing a lot too, especially since with a slower processor folding run faster.


v outputs from HMonitor
+12 11.97-12.03
-12 12.54-12.54
+3.3 3.25-3.26
+5 4.73-4.76
+5 VCCH 5.46-5.48

This is just a short run for a few minutes because a storm just reboot the rig after monitor had been running for about an hour.

Gldm

12v looks fine but 5v seems a bit low. Could just be the sensors, might want to check with a multimeter if you have one. Since 5v isn't heavily used anymore (back in the old days the CPU used it) a lot of modern motherboards use the 5v rail to drive the power phases for the RAM. The Neo 550's rated for 3%, it's almost twice that far from spec if those numbers are accurate. Some 5v logic starts to have issues around 4.75v, some tolerates as low as 4.5.

As for the MB... the Biostar A760G M2+ was based on the 760G chipset, not the NF6100. The Gigabyte GA-MA74GM-S2H is based on the 740G. The main difference that I can see (without actually having the boards to look at or the design schematics) is the GB board seems to use higher quality chokes for the memory power circuits, and an additional phase for the CPU (4 vs 3 on the Biostar). Might be wrong on that since the pics on the sites are low res.

My guess is the Gigabyte board either uses 3.3v or 12v for the memory power and/or has tighter tolerances for the power circuitry. Hard to know for sure though.

_k

My bad I forgot what board it was I was thinking of the MCP6P when I said that. And not to sure how accurate HMonitor is on the 5v since on the PCPnC reads 1.XX and the Ultra LSP its 5.00V +-.03(pretty steady 5.0v which should not happen). Sitting here watching line variance makes shows its not trust worthy enough to worry.