SR-2 Optimization Thread

[The-Real-Link]

Hey guys! Finally jumped in on this and I do have to say this is quite a crazy thread. Not only a lot of settings discussed but a lot of problem-solving too which may prove invaluable if I go any higher with my SR-2 as well.

Specs for my build are here:
http://www.overclock.net/intel-build-logs/864314-melchior-project-sr-2-build-oc.html

Sitting right now at 183 BCLK. Passed 20 Standard and 2 Maximum runs of IBT w/16 threads so happy so far . I just have to wonder if my voltages and timings are still ok or if I should lower them (or keep them the same if I OC higher)? Basically will Auto settings ever get to a point where it can fry something? Is there a "cut-off" where I should change to manual settings? I'm guessing though the more I read about these things that it sounds like that'd be past 200 BCLK.

First image just to show I'm at about 3.5 with turbo, 3.3ish without so far.

I can see the memory getting looser (I think) to compensate for the speed but I've just left it on auto. Actually everything's on auto except for Vcore of 1.3 and VTT of 1.35 I believe.

I know aside from temperatures that I can still push a little higher. It's just been so smooth though (one temp BSOD which a reboot fixed) that I want to make sure I'm fine before I get into "difficult" territory

 

[zero2dash]

AFAIK it seems like the consensus is 1.35 for either Vcc or Vtt is sure fire "safe", some run slightly higher with no problems. It also seems like IOH voltage is the key to SR-2 OC stability, most of us running 1.4 in that regard and no problems.
I achieved stable 200bclk on my L5640's with 1.20625 Vcc and 1.3 Vtt.

 

[musky]

Hey guys! Finally jumped in on this and I do have to say this is quite a crazy thread. Not only a lot of settings discussed but a lot of problem-solving too which may prove invaluable if I go any higher with my SR-2 as well.

<snip>

Sitting right now at 183 BCLK. Passed 20 Standard and 2 Maximum runs of IBT w/16 threads so happy so far . I just have to wonder if my voltages and timings are still ok or if I should lower them (or keep them the same if I OC higher)? Basically will Auto settings ever get to a point where it can fry something? Is there a "cut-off" where I should change to manual settings? I'm guessing though the more I read about these things that it sounds like that'd be past 200 BCLK.

<snip>

I can see the memory getting looser (I think) to compensate for the speed but I've just left it on auto. Actually everything's on auto except for Vcore of 1.3 and VTT of 1.35 I believe.

I know aside from temperatures that I can still push a little higher. It's just been so smooth though (one temp BSOD which a reboot fixed) that I want to make sure I'm fine before I get into "difficult" territory

A few thoughts:

Can your memory handle a higher bclk? My guess would be yes, but SR-2s are very picky about their memory. You can back your CPU multiplier down to 12x and see if you can run at 200+ bclk. 1.35 Vtt should be good for close to 200 bclk. You may need 1.375 to go over.

Leave IOH on Auto until you have a reason to not have it on Auto (i.e. stablity issue at a higher bclk.) One of the highest bclk setting we have seen was done by Tobit, and he hit 211 bclk with IOH on Auto. I have tried to duplicate that without luck so far on my other machines, so it is hit and miss.

Watch your actual voltages. I have 4 of these things, and none of them are dead on for any of their voltage settings. Memory is the one to really watch. All of mine show less voltage in the setting than actual. On one of them, the 1.65V setting for one of the memory channels was actually 1.69, which caused me some issues. The other voltages are all lower actual than the settings say, so they are not as bad. It is still a good idea to know exactly where you are, though.

How much Vcore is too much? My highest one is at 1.38V actual without Vdroop. I do not like having it that high, but those particular processors need a ton of voltage for some reason. How much is too little? My lowest at the moment is 1.28V actual on the E5640 quad machine @ 4.14 (197 x 21). That might be a sign that Westmere quads do not need as much Vcore, but I doubt it, It is really going to depend on the chips. zero2dash has two L5640s that hit 3.6 on 1.2V, and I have two that require almost 1.4 for the same clock.

I can tell you that ccityinstaller has a single E5620 at 4.16, so you definately should have some more room. The board and memory should not be your limiting factor right now, nor should temperatures with you HSFs. Keep on pushing it!

 

[The-Real-Link]

A few thoughts:

Can your memory handle a higher bclk? My guess would be yes, but SR-2s are very picky about their memory. You can back your CPU multiplier down to 12x and see if you can run at 200+ bclk. 1.35 Vtt should be good for close to 200 bclk. You may need 1.375 to go over.

Leave IOH on Auto until you have a reason to not have it on Auto (i.e. stablity issue at a higher bclk.) One of the highest bclk setting we have seen was done by Tobit, and he hit 211 bclk with IOH on Auto. I have tried to duplicate that without luck so far on my other machines, so it is hit and miss.

Watch your actual voltages. I have 4 of these things, and none of them are dead on for any of their voltage settings. Memory is the one to really watch. All of mine show less voltage in the setting than actual. On one of them, the 1.65V setting for one of the memory channels was actually 1.69, which caused me some issues. The other voltages are all lower actual than the settings say, so they are not as bad. It is still a good idea to know exactly where you are, though.

How much Vcore is too much? My highest one is at 1.38V actual without Vdroop. I do not like having it that high, but those particular processors need a ton of voltage for some reason. How much is too little? My lowest at the moment is 1.28V actual on the E5640 quad machine @ 4.14 (197 x 21). That might be a sign that Westmere quads do not need as much Vcore, but I doubt it, It is really going to depend on the chips. zero2dash has two L5640s that hit 3.6 on 1.2V, and I have two that require almost 1.4 for the same clock.

I can tell you that ccityinstaller has a single E5620 at 4.16, so you definately should have some more room. The board and memory should not be your limiting factor right now, nor should temperatures with you HSFs. Keep on pushing it!

Hey Musky, thanks very much for that detailed reply. I was thinking I'd try to first get to say, 200 BCLK then try lowering my voltages and see if I can run less through them (looking for longevity here too)

The particular RAM I put in is Mushkin Redline DDR3 1600, 8-8-8-24. I don't know how to load my XMP#3 profile (it shows 800 mhz @ 8-8-8-24 in CPU-Z) but I guess I could always set the settings manually.

Man, you have 4 SR-2s, MIBW has 3... what do you guys do?! I'd love to build more of these rigs myself but gosh the $$$ lol.

 

[FLECOM]

my SR-2 learned a new trick... now it blue screens after about 2~3% of a 6900... tried a new hd and install of win7, still does the same thing... tried new vid card (since before it blue screens the video drivers crash) still does it

I think I give up

I am trying to get an invoice so EVGA will actually help me, but I really, REALLY hate manufacturers that do that... I will >NEVER< be purchasing an EVGA product ever again

 

[The-Real-Link]

Flecom, I can't blame you for being upset with EVGA for the SR-2 crashing but there can be many factors causing a BSOD. Are you overclocking? Did you install the latest BIOS? Did you up your voltages or change any BIOS settings? Have the latest drivers from EVGA and other manufacturers for your system components?

Sorry, not trying to downplay anything here, but I think it's safe to say we'd need more information to help you with your crashes before simply RMA'ing the board.

 

[may i be worthy]

my SR-2 learned a new trick... now it blue screens after about 2~3% of a 6900... tried a new hd and install of win7, still does the same thing... tried new vid card (since before it blue screens the video drivers crash) still does it

How long has it been folding successfully at your current settings?

What is the STOP error - is it #124 by any chance? If so I think you might need to drop uncore or increase vtt. I doubt it has much to do with HDs or Win7. I had issues with graphics drivers crashes on my main rig at 19x/20x uncore, moving to 18x cleared that up. Loosening memory settings can also help. I found a good way to see if you are out of the woods is to run 3D Furmark - I would get artifacts instantly.

6900s are like the 6701s of bigadv - they show up issues more than other work units.

This is assuming you want to persist with it - but it sounds like an overclock issue.

 

[musky]

Hey Musky, thanks very much for that detailed reply. I was thinking I'd try to first get to say, 200 BCLK then try lowering my voltages and see if I can run less through them (looking for longevity here too)

The particular RAM I put in is Mushkin Redline DDR3 1600, 8-8-8-24. I don't know how to load my XMP#3 profile (it shows 800 mhz @ 8-8-8-24 in CPU-Z) but I guess I could always set the settings manually.

There is no way to automatically load you profile. Set your DRAM Frequency to DDR-1066, and manually set your timings. Something like 8-8-8-24-8-72-4-4-4-20 with a command rate of 1T should work. You aren't anywhere near your rated speed at the moment. Once you get over 200 bclk, you may have to loosen them up a bit.

Man, you have 4 SR-2s, MIBW has 3... what do you guys do?! I'd love to build more of these rigs myself but gosh the $$$ lol.

We clearly spend entirely too much money on folding parts... Keep in mind, mine are all screwed down to pieces of plywood, have 6 x 1Gb memory, and run antique 80Gb HDDs and Radeon X300 video cards. You have enough "extras" with yours to make a whole new SR-2 in my world...8)

 

[The-Real-Link]

There is no way to automatically load you profile. Set your DRAM Frequency to DDR-1066, and manually set your timings. Something like 8-8-8-24-8-72-4-4-4-20 with a command rate of 1T should work. You aren't anywhere near your rated speed at the moment. Once you get over 200 bclk, you may have to loosen them up a bit.



We clearly spend entirely too much money on folding parts... Keep in mind, mine are all screwed down to pieces of plywood, have 6 x 1Gb memory, and run antique 80Gb HDDs and Radeon X300 video cards. You have enough "extras" with yours to make a whole new SR-2 in my world...8)

Haha that's good to know XD. Woot finished my first BigWU - stable all the way through in one go. Thanks for the timings suggestion - I'll experiment more once I push onward.

Hmm next project up for my SMP is a 2684 - I think that's another big one Awesome.

 

[may i be worthy]

AFAIK it seems like the consensus is 1.35 for either Vcc or Vtt is sure fire "safe", some run slightly higher with no problems. It also seems like IOH voltage is the key to SR-2 OC stability, most of us running 1.4 in that regard and no problems.
I achieved stable 200bclk on my L5640's with 1.20625 Vcc and 1.3 Vtt.

I second all that. From my research vtt is the more dangerous of the two. I run 1.35 vtt on two machines, 1,375 on a third with 12 DIMMs.

Vcore I am happy to go up to 1.4 something. I am right at the limit of what I am comfortable doing - 1.44 on one core of one machine, 1.425 on another. This is reported as 1.43 and 1.40v in eleet, for what that is worth.

The rest of the machines are at 1.40625 to 1.43v. The first rig has been at ~1.40v for 5 and a half months, with no noticeable degradation.

Man, you have 4 SR-2s, MIBW has 3... what do you guys do?! I'd love to build more of these rigs myself but gosh the $$$ lol.

I am self employed and use these for 3D rendering, so I am a bit unusual as my business pays for my addiction.

Musky is just certifiably [H]ard / Insane depending on your frame of reference. And like he said, you can save a lot my making your own case, using old HDDs etc - my case for the second two cost $5. The "trays" are just surplus floor tile samples. It is true, the fastest SR2s in the [H]orde run on the finest marble.

Don't ask me what I spent on CPUs. Even with the x5680s second hand, it adds up. If I had my time over again I would have got the x5680s second hand for fun/overclocking/e-peen/fastest rig in the [H]orde fun, then 3 extra SR2s with fleabay L5640s for less money.

 

[zero2dash]

Leave IOH on Auto until you have a reason to not have it on Auto (i.e. stablity issue at a higher bclk.) One of the highest bclk setting we have seen was done by Tobit, and he hit 211 bclk with IOH on Auto. I have tried to duplicate that without luck so far on my other machines, so it is hit and miss.

Once I take mine down, I'm going to try that (IOH on Auto). Maybe I can hit higher.
Seems like Tobit's cpu's take less juice than everyone elses and he hit the 211 stable (think he got 213 unstable)....I'm wondering about mine now, seems like mine hit higher on lower juice also. I didn't go much higher than 200bclk but I didn't leave anything on Auto (out of habit....always punch in a setting as opposed to leaving things on Auto, figuring that Auto usually overestimates). Hmmmmmm.........

 

[The-Real-Link]

MIBW - Cool stuff. Yeah I'm starting to contemplate the same thing - ie. use the SR2 system to start a company and have the business fund itself. Still working out the details though.

Wow at that bench. For some reason the track "Devil's Lab" from FFVI / FFIII comes to mind looking at that I'll admit, I'm running a far more tame installation of my system.

 

[musky]

Once I take mine down, I'm going to try that (IOH on Auto). Maybe I can hit higher.
Seems like Tobit's cpu's take less juice than everyone elses and he hit the 211 stable (think he got 213 unstable)....I'm wondering about mine now, seems like mine hit higher on lower juice also. I didn't go much higher than 200bclk but I didn't leave anything on Auto (out of habit....always punch in a setting as opposed to leaving things on Auto, figuring that Auto usually overestimates). Hmmmmmm.........

Yours has a chance, if that is really the trick. I have Tobit's rig now, and it is legit. I'm at 205 bclk easily, with everything on Auto. I'll bump it on up to 210/211 this weekend. My other two L540 rigs were a complete flop without IOH at 1.4V, so it is definately hit and miss. The other variable is the memory, so I may play musical memory just to take that out of the equation. I still think it is the processors, though.

 

[jebo_4jc]

I'm dying to know if the memory has anything to do with it.

 

[nitrobass24]

I think its the mobo.

I can take my E5530s to 220blck on my Sabertooth with everything on auto.
On my SR-2 195blck is hard using any of memory kits.

 

[The-Real-Link]

Up to 3.75 now! Tests repeated as successfully and flawlessly as before as if this is barely breaking a sweat. I know I'm not doing say, 20+ runs of IBT on Max but at least this is a quick way for me to test stability and allows me to safely know I can test out programs for my graph stats without crashes or stuff. 197x18 seems all good here and with the 19 turbo, still holding in the mid-60s at worst. Still left all things on auto. For some reason, manually putting in my RAM timings (tighter) and DDR to 1066 as was suggested above dropped my IBT performance by over 10 GFlops... So I put the ratio off 1066 and put it back to Auto, yet in Eleet, that's what it's picking anyway. I don't know how the discrepency is coming about though. Oh well.

Oh and if the numbers are indeed accurate, my DIMM voltage is indeed 0.03 too high in Eleet. It's still at 1.65 in BIOS.

 

[The-Real-Link]

One more question please - I am wondering how far I can push with just the 8-pin power connectors on the board. While at stock clocks, I'm idling at 300w and with CPUs loaded, I'm at 380 so 40 watts each. Of course the thing I'm wondering though is how much additional power are they using even at "idle"? If power connectors are indeed the same as per GPUs, that'd mean an 8-pin would give each CPU socket about 150w to use.

On CPU load just now, I notice the wattage goes up to 535 watts with the same test as before. This would equate to an additional boost of 140w so 70 more watts each. HWMonitor originally was showing 80 watts per, and now generally spikes at 90.75 watts per, but it goes in giant jumps and I would think, is probably not quite accurate.

Again I can't confirm for sure that I can't boot with both 8 and 6 pins plugged in, though the 6 pin on CPU 1 is so tight next to the fan that it could barely go in straight when I tried originally and in fact, it kept me from my first boot. CPU 0's 6-pin I think is still plugged in though.

Is there a point where I'll be over-drawing the power that is provided / required to the CPU sockets?

 

[may i be worthy]

One more question please - I am wondering how far I can push with just the 8-pin power connectors on the board. While at stock clocks, I'm idling at 300w and with CPUs loaded, I'm at 380 so 40 watts each. Of course the thing I'm wondering though is how much additional power are they using even at "idle"? If power connectors are indeed the same as per GPUs, that'd mean an 8-pin would give each CPU socket about 150w to use.

On CPU load just now, I notice the wattage goes up to 535 watts with the same test as before. This would equate to an additional boost of 140w so 70 more watts each. HWMonitor originally was showing 80 watts per, and now generally spikes at 90.75 watts per, but it goes in giant jumps and I would think, is probably not quite accurate.

Again I can't confirm for sure that I can't boot with both 8 and 6 pins plugged in, though the 6 pin on CPU 1 is so tight next to the fan that it could barely go in straight when I tried originally and in fact, it kept me from my first boot. CPU 0's 6-pin I think is still plugged in though.

Is there a point where I'll be over-drawing the power that is provided / required to the CPU sockets?

Yes, I hit this when trying my how-far-can-you-go tests at 1.400v vtt and 1.43v vcore. I would reboot when Intelburntest peaked at 685 watts (idle about 300w). I would explode on the second pass each time. Plugged in optional plugs and I could pass 25 mins+, peaking at 693 watts. (I did not test further as that was well above my comfort level with voltage for 24/7)

Dropping to 1.375v vtt actually without plugs also got a bit further as the load was less. So the power was the problem for sure.

Having said that, I am also a big fan of not dying wondering if leaving plugs out was the limiting factor for your overclock, so if you can do it, it might be a good idea.

 

[jebo_4jc]

The plugs helped my stability as well. I recommend them
Posted via [H] Mobile Device

 

[Kendrak]

Mord power to the board is never a bad thing.

Plug it in, plug it in.
/jingle

 

[zero2dash]

Well, pulled a 2684 after finishing the last 6900 so...here goes nothin'. [prepares to reboot to fart around in bios]
Wish me luck [H]

 

[DeFex]

OK, I am having a little trouble. I installed some E5645 ES CPUs (not from ebay, dont get excited!) and everything seems to be detected fine, however there is no CPUID on the advanced CPU config page in the BIOS, also when i change Baseclock, the bios "remembers" it changed, but the CPUs are still booting at stock. (A50 BIOS) do i need to install Locolaptop A49 BIOS?

BTW I am waiting for a new windows license, so i am just booting to a Linux or DOS USB stick.

 

[FLECOM]

it might be lying to you... does it show them running @ stock in windows??? my CPU's post "stock" but ARE running at the desired speed... just one of the more fun features of this board...

I seem to have gotten my SR-2 stable again by (get this) using the other onboard NIC.... mmkay

 

[The-Real-Link]

Got a few things to take care of over the weekend so will probably leave the system @ 3.75 for a few days to ensure all is well. Had another question or two though...

1. I know I've gone over this with other forums before and it seems so simple yet I'm still confused and want to be sure. Basically once my other GTX 580 comes back, I feel I'm going to be UPS-bound in terms of power draw. Right now the OC I have + one 580 says it's pulling right about 700 watts from the wall, however I'm also pretty sure the Cyberpower PP1500SWT2 unit I have has a PFC of about 87%. So let's say I go on and plug that second bad boy in. Am I limited when the wall output hits 1K or simply once the PFC on the UPS is at 1K? I'd imagine the latter, but even then it's within the SR-2 PSU's continuous limit of 1.2KW. I still would imagine it's not ideal to load my UPS to 1,300 watts if it converts 1,000 properly. That's still too much on the borderline I think,

2. Ok guess I was slightly mistaken on the labels of my RAM. While I said 8-8-8-24 in my video tutorials and what things are set to, the XMP profile is indeed 6-8-6-24 according to CPU-Z. The system seems to be very stable as-is right now. Would I really gain any seeable benefit from simply tightening the timings? Might put an extra fan around the RAM banks of CPU1 though as that is definitely a fair bit warmer than CPU0. Too bad there are no sensors to measure how hot the RAM is .

Thanks.

 

[musky]

OK, I am having a little trouble. I installed some E5645 ES CPUs (not from ebay, dont get excited!) and everything seems to be detected fine, however there is no CPUID on the advanced CPU config page in the BIOS, also when i change Baseclock, the bios "remembers" it changed, but the CPUs are still booting at stock. (A50 BIOS) do i need to install Locolaptop A49 BIOS?

BTW I am waiting for a new windows license, so i am just booting to a Linux or DOS USB stick.

Every "extra special" chip I have had in an SR-2, and there have been a lot, show stock speed on boot regardless of what speed the chips are actually running.

As long as you can boot to DOS or Linux on the USB stick, you don't need LOCO's bios. If you get stuck on "Loading NVRAM", you do need his bios. Sounds like you are fine there also.

 

[musky]

Got a few things to take care of over the weekend so will probably leave the system @ 3.75 for a few days to ensure all is well. Had another question or two though...

<snip>

2. Ok guess I was slightly mistaken on the labels of my RAM. While I said 8-8-8-24 in my video tutorials and what things are set to, the XMP profile is indeed 6-8-6-24 according to CPU-Z. The system seems to be very stable as-is right now. Would I really gain any seeable benefit from simply tightening the timings? Might put an extra fan around the RAM banks of CPU1 though as that is definitely a fair bit warmer than CPU0. Too bad there are no sensors to measure how hot the RAM is .

Thanks.

You may gain some folding speed with tighter timing. Nothing huge, but a few seconds/frame. Is it worth pulling your hair out to get stable at tighter timings? Not even close. Since it doesn't look like you are running your memory at its rated 1600 MHz, I'd definately give it a try. Command rate makes a much bigger difference, so as long as you are running 1T, I would worry that much about memory timings.

I can't help you on the UPS question - your one machine is going to pull close to what three of mine pull, and i don't use UPS's.

 

[jebo_4jc]

All I know is, if you're ups limited, you'll know it immediately.

It will probably beep and shut down if you push it too far. Several of my rigs are unfortunately too big for my UPSs
Posted via [H] Mobile Device

 

[DeFex]

Every "extra special" chip I have had in an SR-2, and there have been a lot, show stock speed on boot regardless of what speed the chips are actually running.

As long as you can boot to DOS or Linux on the USB stick, you don't need LOCO's bios. If you get stuck on "Loading NVRAM", you do need his bios. Sounds like you are fine there also.

I just installed windows, and you are right thanks! my cpus are "000 @ 2.40GHZ (ES)"s

 

[Kendrak]

I just installed windows, and you are right thanks! my cpus are "000 @ 2.40GHZ (ES)"s

Glad to see the system is happy.

 

[DeFex]

well not 100% for some reason a GTX460 which worked fine on another PC, I can no longer control its fan speed, and it is stuck on full blast.

edit: oh, that noise was the little chipset fan, i managed to tame it in the BIOS, I still cant get the GPU fan to change manually, and all monitoring programs say it is a "40%" all the time, however the speed does increase slightly, as the GPU makes it up to a whopping 55c in furmark

I have the 5645s at 3.42 (180x19) and i will finish my current WU at that speed just to make sure it is good, maybe over the weekend ill start trying to get up around 3.6-3.8 with them, I dont want to go crazy voltage/heat with it. right now its actually quieter than my 970 was with 2 fans on a H50.

cool to see over 100 GFLOPS on intel burn test

 

[SazanEyes]

1. I know I've gone over this with other forums before and it seems so simple yet I'm still confused and want to be sure. Basically once my other GTX 580 comes back, I feel I'm going to be UPS-bound in terms of power draw. Right now the OC I have + one 580 says it's pulling right about 700 watts from the wall, however I'm also pretty sure the Cyberpower PP1500SWT2 unit I have has a PFC of about 87%. So let's say I go on and plug that second bad boy in. Am I limited when the wall output hits 1K or simply once the PFC on the UPS is at 1K? I'd imagine the latter, but even then it's within the SR-2 PSU's continuous limit of 1.2KW. I still would imagine it's not ideal to load my UPS to 1,300 watts if it converts 1,000 properly. That's still too much on the borderline I think,

This is why I went with dual PSUs, so I can plug each one into a different UPS. Currently my X5650 box with dual GTX 295s runs off of one SmartUPS 1500, but it is limited by the UPS if I fire up the GPUs. As jebo said, if you're limited, you'll know it immediately. First sign is usually the UPS internal fan becoming extremely loud (if there is a fan) and/or lights flashing. Next sign is a loud alarm and shutdown. If you look up the specs for your UPS, there should be a Watt rating in addition to a VA rating. My SmartUPS with VA of 1500 only provides around 1060W, and the cheaper BackUPS 1500s are more like 900W.

Other than dual PSUs and UPSes, you have a couple options. You can get a dedicated 20A circuit and buy a bigger UPS, or you can live with the risk of power failure and just use a surge protector. I ran my dedicated GPU box with 1200W PSU with only a Tripp Lite surge protector for a while since no UPS could handle the load. BTW, the Tripp Lite warranty service on their Isobar line is fantastic, so I'm a customer for life.

 

[The-Real-Link]

Hey Sazan, thanks for the input on that. Yeah I'll just have to load it up and see - the auto-trip sensor should beep and shutdown at any rate (of course since it's there in the first place). I was just trying to figure if I'd hit that limit at 1000w from wall or more like 1,300. Guses I can just try and see.

Good news tonight though is while I might try out 200 BCLK tonight, I took others' suggestions and sure enough, I had wired the system wrong. My CPU0 actually had the 3 pin combo meant for CPU1 haha! A few zip-tie cuts and some rerouting (and some red hands after an hour or two) and we're all wired up. Sure enough thankfully compared to my first power-up, all the 8 / 6 pin connectors were *securely* connected and I booted up first time now without any issue. Running IBT again just to make sure it's all good. At least now I hope I don't have to factor a lack of CPU power into my crunching. Onward!

 

[DeFex]

note about the 5645 ES, im not sure if it is because i happened to put the right chip in the right socket, but the temps are very close to equal at the same voltage. no 10 degree diff.

 

[The-Real-Link]

(Crossposted from EVGA's forums):

OC'ing is going pretty well and I've been stable without a hitch up to 197 BCLK! I tried moving to 211 (for a turbo of 4ghz even) and then really hit a wall all at once.

I had it boot into Windows once, sound played but was distorted. I restarted but then it gave me a few different codes like d0, F4, and FF. I know on Shamino's template that the CPU signals are very low but I've mostly kept things on auto or close to it. Just printed out the boot codes but I don't see any of those listed. Any ideas please? Maybe I should go up more slowly than 10 BCLK at a time now, haha.

 

[zero2dash]

sound played but was distorted

http://hardforum.com/showpost.php?p=1036590900&postcount=823
Put these signal tweaks in, they fix the "sound echoing like a tin can" problem. I had the same problem myself and tried several drivers with no change.....punched in those tweaks, sound is perfect again.

My >200bclk on the Even Steven L5640's is at a point where....honestly I just don't care anymore. I was having LinX crashes with 206 and occasional ram disappearing. I bumped Vcc, Vtt, Vdimm, IOH Vcore....basically still kept having the crash problem. (Fail after just over 1 min. @ 12x206)

200, 203, 206...if I get up there, 211+....etc....meh. An extra hundred mhz for much higher voltage....just not really worth it. More importantly I just want to be back "up" instead of endlessly farting around in ELEET or the bios.

 

[The-Real-Link]

http://hardforum.com/showpost.php?p=1036590900&postcount=823
Put these signal tweaks in, they fix the "sound echoing like a tin can" problem. I had the same problem myself and tried several drivers with no change.....punched in those tweaks, sound is perfect again.

My >200bclk on the Even Steven L5640's is at a point where....honestly I just don't care anymore. I was having LinX crashes with 206 and occasional ram disappearing. I bumped Vcc, Vtt, Vdimm, IOH Vcore....basically still kept having the crash problem. (Fail after just over 1 min. @ 12x206)

200, 203, 206...if I get up there, 211+....etc....meh. An extra hundred mhz for much higher voltage....just not really worth it. More importantly I just want to be back "up" instead of endlessly farting around in ELEET or the bios.

I'll take a look at that link, thanks Zero. Yeah I'm feeling the same way in a sense - I don't mind trying to tweak the system but if it gets to where push is more akin to shove and nothing remains stable, then I'll gladly revert to a great usable setting. This 197 BCLK is working just fine

 

[Kendrak]

Z2D,

I feel the same way. Once I hit an OC I'm happy with I let it go. The ppd lost durring days of crashes or tweeking is hard to make back up with and extra 70mhz OC.

 

[Aix.]

I've read a couple of the threads discussing high-end BigAdv and how GPU3 isn't worth it, but how does GPU2 compare with todays WUs? I snagged a GTX 280 for $20 yesterday that runs hot and apparently artifacts (I didn't see any during my initial testing), but I figure I'll take a shot at baking it and possibly fold with it depending on the results.

Would my BigAdv take a hit running a single GPU2 client on my SR-2? I haven't run a GPU client in more than a year (silly AMD cards).

 

[Tigerbiten]

Would my BigAdv take a hit running a single GPU2 client on my SR-2? I haven't run a GPU client in more than a year (silly AMD cards).

Yes.

On an SR-2 system it not really worth running a GPU client unless you have a top of the range card.
The points lost from the bigadv client are just about matched by the points gained by the GPU client on a lesser card, less that +/-1,000 PpD gain is normal.
Plus you then need to cope with the extra power draw and heat from the card.

I run the GPU2 clients on my i7 920 boxen becuase it does not make that much difference to the PpD of the box.

Remember the bonus scales up with speed so ...........
A 5 sec difference per frame on my SR-2 box is worth around 1,000 PpD
A 5 sec difference per frame on my i7 box is worth only around 100 PpD.
You see the difference.

Luck ...........
[H]

 

[zero2dash]

On an SR-2 system it not really worth running a GPU client unless you have a top of the range card.

IME gpu2 doesn't really cause much of a 'cpu cycle fight' between the smp2 client running bigadv vs the gpu2 client(s).

I think when I averaged it out I gain 10k PPD for running my 260s even after losing 1-2k to the smp2 client itself. Maybe things have changed since I checked this; admittedly, it's been awhile and there's been new WU's released for gpu2.....guess I'll have to investigate this myself.