WCG-TPU Cruncher's Hardware/Tech Support Discussion Thread

[twilyth]

Thanks. They fit--so they're clearly physically compatible--but I'll try to rustle up some pinouts to make sure it won't short things out.

I got mine to fit too. They just didn't work.

 

[manofthem]

OK so let me open this up officially... and pre-pre-apology for long post

I have 3 i7s running pretty much 24/7:

• 4770k @4.3; RAC: 6,291
• 4790 @3.8; RAC: 4,479
• 2600k @3.5; RAC: 4,181

In addition to my system specs, some of their crunching info can be found here on my Free-DC page.


Problem: the 4790 and 2600k seem to be underperforming, and I'm having trouble figuring out how to bring them up. The 4770k seems to be pulling in decent numbers, yet it's the only rig that's gets paused temporarily when I do some gaming. All running Windows 10 (upgrade, not clean install), pretty sure all have power saving features turned off in bios and in Windows, all wired on same network. WCG reports all valid results, no errors and invalids.

The other night I compared the Boinc manager settings if the 4770k to the other to ensure they're the same. The 4790 rig basically sits and crunches and generally doesn't get used at all otherwise. The 2600k is the wife's pc, and she's on it here and there doing some typical Web stuff, little else.




So, what I'm asking for is perhaps a few possible suggestions as to improve their performance. If you have some ideas, please shout them out so I can try them.

Thanks!

 

[thebluebumblebee]

What are Whetstone/Dhrystone results? (in the event log)

 

[[Ion]]

OK so let me open this up officially... and pre-pre-apology for long post

I have 3 i7s running pretty much 24/7:

• 4770k @4.3; RAC: 6,291
• 4790 @3.8; RAC: 4,479
• 2600k @3.5; RAC: 4,181

In addition to my system specs, some of their crunching info can be found here on my Free-DC page.


Problem: the 4790 and 2600k seem to be underperforming, and I'm having trouble figuring out how to bring them up. The 4770k seems to be pulling in decent numbers, yet it's the only rig that's gets paused temporarily when I do some gaming. All running Windows 10 (upgrade, not clean install), pretty sure all have power saving features turned off in bios and in Windows, all wired on same network. WCG reports all valid results, no errors and invalids.

The other night I compared the Boinc manager settings if the 4770k to the other to ensure they're the same. The 4790 rig basically sits and crunches and generally doesn't get used at all otherwise. The 2600k is the wife's pc, and she's on it here and there doing some typical Web stuff, little else.




So, what I'm asking for is perhaps a few possible suggestions as to improve their performance. If you have some ideas, please shout them out so I can try them.

Thanks!

So the 4790 is the only one on that list where the PPD seems drastically off. The 4770k looks right on par. My 4790 (non-K) at work running Win8.1 x64 gets just under 5.8K PPD. So your 4790 should be similar.
The 2600k is low but not horribly. I'd expect that since it's in use and running Windows it should get somewhere around 4.5k or 4.8k PPD.

 

[manofthem]

What are Whetstone/Dhrystone results? (in the event log)

I don't know exactly what/where that is. Could you elaborate for me a bit on how to find/compare that please?



Just checked in on the 4790 and snapped this:

I never noticed that little url error before, but I set to no new tasks to re-add the project to see if it helps.

 

[[Ion]]

I don't know exactly what/where that is. Could you elaborate for me a bit on how to find/compare that please?



Just checked in on the 4790 and snapped this:

I never noticed that little url error before, but I set to no new tasks to re-add the project to see if it helps.

Go to Advanced -> Run CPU Benchmarks I believe and it'll show up in the log. That'll give you the scores

 

[thebluebumblebee]

Advanced-event log. Then just scroll through looking for it.

 

[manofthem]

Go to Advanced -> Run CPU Benchmarks I believe and it'll show up in the log. That'll give you the scores

Advanced-event log. Then just scroll through looking for it.

Got it, thanks.

Will run and update in a second



4770k: 4897 floating point MIPS (Whetstone) per cpu
18776 integer MIPS (Dhrystone) per cpu.

4790: 4295 floating point MIPS (Whetstone) per cpu
16637 integer MIPS (Dhrystone) per cpu.

2600k: 3730 floating point MIPS (Whetstone) per cpu
11850 integer MIPS (Dhrystone) per cpu.

While I don't know exactly what these mean, I can already see a slight corresponding relationship between these numbers and what I've been seeing daily.



Edit: I've heard of these Whetstone and Dhrystone before, but I never knew what they were. Looking into it apparently it's an algorithm created years ago to evaluate cpus. Now I suppose the key is to figuring out how my numbers actually relate to others and/or where these numbers should be.

More tomorrow, off to bed now

 

[t_ski]

Matt, what projects are on each rig? Remember that some projects perform better than others. For example, it you are running a lot of OET, then your numbers will be lower.

 

[ChristTheGreat]

Hey guys, long time I didn't posted!

I had to clean my Watercooling has there was shit from old loop (stuck in rad I guess, I cleaned it well, I thought hehe) and I had a big bubble. But finally, I decided to sell all my wsatercooling parts. BOught back a NH-U12S and will change R9 290 for a GTX 970 (since it has a waterblock).

Going back to air! Less maintenance, less money, more new parts!

 

[manofthem]

Matt, what projects are on each rig? Remember that some projects perform better than others. For example, it you are running a lot of OET, then your numbers will be lower.

Thanks for the reminder there about different projects and points. Actually all rigs were running OET exclusively, but today I added in FAAH2. Also on the 4790 I updated Boinc version and reattached project due to the url error, so I'll give it a few days now and see how it goes.

 

[t_ski]

Looks like one of my rigs is down. I haven't had a chance to dig into it completely, but I noticed it was not running yesterday. All the lights seems normal on the motherboard, but unfortunately it won't power on. I was able to verify that it was not the switch on the case by swapping it out for the reset switch as well as just using a screwdriver to jump it. Lights continue to be at their normal state, but the rig never spins up.

One thing that comes to mind is the 8-pin EPS adapter I have worked up. If you have any other suggestions on what I can look for, please post them up. I probably won't be able to do anything with it until maybe Saturday at the earliest.

 

[[Ion]]

Well, without knowing anything about it, it's very hard to draw conclusions

 

[t_ski]

Well, without knowing anything about it, it's very hard to draw conclusions

What can I offer up?

Supermicro 2P board with two E5-2680 v1 CPUs, 8x8GB Samsung ECC ram, Thermaltake 600W (IIRC) PSU. Two AIO liquid coolers. Rig has been running fine for a week or so. There may have been a power outage, but none of the other rigs look like they restarted.

EDIT: Just looking for ideas to hopefully shorten my troubleshooting time.

 

[[Ion]]

What can I offer up?

Supermicro 2P board with two E5-2680 v1 CPUs, 8x8GB Samsung ECC ram, Thermaltake 600W (IIRC) PSU. Two AIO liquid coolers. Rig has been running fine for a week or so. There may have been a power outage, but none of the other rigs look like they restarted.

Nice CPUs

That it doesn't turn on at all is problematic. If it at least came on with no POST that would be better. I'd wonder about the PSU--Thermaltake has made a number of bad products before (hence, I think, the origin of the name Thermalcrap Pisswater to describe some of their LC kits) and it could well have failed. Do you have a spare you can test with?

Unless the EPS adapter you have shorted out, I suspect it isn't the issue. I'd imagine your board has a pair of 8pin plugs (correct, no?) so even if it loses power to one it should still power on (all of the SM boards I've had do that).

Feasible to test outside of the case? Seems unlikely that something would have shorted against the back of the board all of a sudden, but a solid troubleshooting strategy anyways.

I presume you've checked and neither of the AIO WC kits has leaked?

 

[stinger608]

Just looking for ideas to hopefully shorten my troubleshooting time.

First I would unplug any peripherals that are not needed. IE CD/DVD, Hard drive(s), card readers, etc... Then try it again. I have had hard drives fail and have the system not come to life. Pull the memory and test with a single different chip. Easy and quick test.
If that isn't it then I would test with another power supply as [Ion] suggested.

That would be the first three tests I would perform.

 

[t_ski]

Nice CPUs

That it doesn't turn on at all is problematic. If it at least came on with no POST that would be better. I'd wonder about the PSU--Thermaltake has made a number of bad products before (hence, I think, the origin of the name Thermalcrap Pisswater to describe some of their LC kits) and it could well have failed. Do you have a spare you can test with?

Unless the EPS adapter you have shorted out, I suspect it isn't the issue. I'd imagine your board has a pair of 8pin plugs (correct, no?) so even if it loses power to one it should still power on (all of the SM boards I've had do that).

Feasible to test outside of the case? Seems unlikely that something would have shorted against the back of the board all of a sudden, but a solid troubleshooting strategy anyways.

I presume you've checked and neither of the AIO WC kits has leaked?

1. Thanks

2. I'd have to pull a PSU out of a different rig to test, so while it's possible, it might be something I do later as stinger suggested.

3. EPS "adapter" is something I've hacked together from an EPS extension and two 4-pin Molex connection using wire nuts. Not pretty, but it had worked for at least a week on this rig and far longer on a previous rig. However, as anything made with wire nuts goes, especially considering any cable movement that occurred while swapping parts around a week or so ago, it could have come loose. BTW, this is only for one of the dual 8-pin headers, as the second one has a genuine EPS plug connected to it.

4. Feasible, yes, but a pain to do. As I said before, these are v1 chips, so they run a lot hotter than v2 chips do. I had the setup in a different case to start with, but I ended up buying two Noctua 92mm coolers and a second AIO liquid cooler. When I got all the parts I figured that I'd move these v1 chips & mobo to the second case, due to it being better suited to house two 120mm AIO's. The first case was less suited for the AIO's so the mobo & CPUs that were moved over to one this one was a 2P v2 setup. My plan worked fine in that the v1 chips run about as hot with the AIO's as the v2 chips do with the air coolers.

Anyway, long story shot, the v1 rig has two AIO's that would need to come out with the mobo if I run it outside the case, which presents a bigger challenge.

5. I did think of that, especially since one of mine is the Cool-it version you said you had issues with before, but I see no signs of leaking.

First I would unplug any peripherals that are not needed. IE CD/DVD, Hard drive(s), card readers, etc... Then try it again. I have had hard drives fail and have the system not come to life. Pull the memory and test with a single different chip. Easy and quick test.
If that isn't it then I would test with another power supply as [Ion] suggested.

That would be the first three tests I would perform.

Yessir, good ol' basic troubleshooting steps that I should remember to follow, even when I often give that same advice to others. It's been a while, but I worked on a family member's PC once and found a NIC was keeping it from booting by doing these same steps.

I think my first step will be to check the home-made EPS adapter, since it's easiest to verify. From there I will start removing parts to get it down to basics, although I might need to stick with two DIMMs due to dual sockets. I can always swap sticks around to make sure they're eliminated from suspicion. After that I may see if I can pull cables over from the second rig's PSU to hook up to this rig. Unless I have any other tips from you guys I might just open up an RMA with Asus on this board to see what steps they have me jump through. My biggest fear is that they end up telling me that the pins in the sockets are messed up and I'm screwed. I might have a way to prove that they're not if it comes to that, so I plan to take precautions

 

[FireFox]

Does someone here own an E5-2695 v2?

 

[t_ski]

1. Thanks

2. I'd have to pull a PSU out of a different rig to test, so while it's possible, it might be something I do later as stinger suggested.

3. EPS "adapter" is something I've hacked together from an EPS extension and two 4-pin Molex connection using wire nuts. Not pretty, but it had worked for at least a week on this rig and far longer on a previous rig. However, as anything made with wire nuts goes, especially considering any cable movement that occurred while swapping parts around a week or so ago, it could have come loose. BTW, this is only for one of the dual 8-pin headers, as the second one has a genuine EPS plug connected to it.

4. Feasible, yes, but a pain to do. As I said before, these are v1 chips, so they run a lot hotter than v2 chips do. I had the setup in a different case to start with, but I ended up buying two Noctua 92mm coolers and a second AIO liquid cooler. When I got all the parts I figured that I'd move these v1 chips & mobo to the second case, due to it being better suited to house two 120mm AIO's. The first case was less suited for the AIO's so the mobo & CPUs that were moved over to one this one was a 2P v2 setup. My plan worked fine in that the v1 chips run about as hot with the AIO's as the v2 chips do with the air coolers.

Anyway, long story shot, the v1 rig has two AIO's that would need to come out with the mobo if I run it outside the case, which presents a bigger challenge.

5. I did think of that, especially since one of mine is the Cool-it version you said you had issues with before, but I see no signs of leaking.



Yessir, good ol' basic troubleshooting steps that I should remember to follow, even when I often give that same advice to others. It's been a while, but I worked on a family member's PC once and found a NIC was keeping it from booting by doing these same steps.

I think my first step will be to check the home-made EPS adapter, since it's easiest to verify. From there I will start removing parts to get it down to basics, although I might need to stick with two DIMMs due to dual sockets. I can always swap sticks around to make sure they're eliminated from suspicion. After that I may see if I can pull cables over from the second rig's PSU to hook up to this rig. Unless I have any other tips from you guys I might just open up an RMA with Asus on this board to see what steps they have me jump through. My biggest fear is that they end up telling me that the pins in the sockets are messed up and I'm screwed. I might have a way to prove that they're not if it comes to that, so I plan to take precautions

Well, I finally got some spare time to mess around with this. It wasn't the EPS adapter, it wasn't any of the devices, but it appears to be the PSU. I remembered I had a PSU tester, and when I plugged that it only the 5V standby light comes on. Usually the whole PSU powers up with this tester as soon as I flip the switch.

Good ol' Thermalcrap

 

[[Ion]]

Well, I finally got some spare time to mess around with this. It wasn't the EPS adapter, it wasn't any of the devices, but it appears to be the PSU. I remembered I had a PSU tester, and when I plugged that it only the 5V standby light comes on. Usually the whole PSU powers up with this tester as soon as I flip the switch.

Good ol' Thermalcrap

Well, I have some pretty old (Truepower Trio maybe?) ~500w Antec PSUs if you want one for the cost of shipping. Nothing more modern though, sorry.

 

[t_ski]

Well, I have some pretty old (Truepower Trio maybe?) ~500w Antec PSUs if you want one for the cost of shipping. Nothing more modern though, sorry.

Thanks, but since this is going in the dual 2680v1 rig, it needs to be more than 500W with some good amperage on the 12v rail(s).

 

[[Ion]]

Thanks, but since this is going in the dual 2680v1 rig, it needs to be more than 500W with some good amperage on the 12v rail(s).

Just checked--it's only 36A combined on the 12V rails.
If you don't mind something with bizarre cabling/sleeving, I recycled a Dell XPS at work today with a 750W PSU. It claims it'll do 732W IIRC between the two 12V rails. I suspect the PSU will still be in the scrap bin when I'm back on Saturday, so I could fish it out and test it. Not very new (came out of a C2Q Q9450 system) but should still work OK for your uses. IIRC plugs are normal (2x6pin, EPS, 24pin, a bunch of SATA).

 

[xvi]

For anyone familiar with used memory prices, how much should I be paying for DDR3 w/EEC? Where can I get my grubby little hands on about 6x2GB of something on the cheap?

Also, regarding the same system, what's with HP's weird heatsink mounting system? Any suggestions for modding my way around that?

 

[[Ion]]

For anyone familiar with used memory prices, how much should I be paying for DDR3 w/EEC? Where can I get my grubby little hands on about 6x2GB of something on the cheap?

Also, regarding the same system, what's with HP's weird heatsink mounting system? Any suggestions for modding my way around that?

I pay $10 per stick for 2GB non-ECC stuff. Sometimes a bit less, but not much. If you can deal with 1GB sticks I can send you a load of them for basically the cost of shipping.

What HP is this? That influences the "weird heatsink mounting system"

 

[xvi]

What HP is this? That influences the "weird heatsink mounting system"

HP Z600 workstation. Has the stupid $60 heatsinks. It's the one that I've been having overheating issues with on top of that. The CPU fan won't ramp up when the proc gets hot. Tempted to just get some eBay CPU waterblocks for it and find some screws that thread in to the HP mounts.

If you can deal with 1GB sticks I can send you a load of them for basically the cost of shipping.

6GB is definitely better than the, uh, 2-3GB it came with? Mostly just want to populate all 6 slots for memory bandwidth and to get sticks in there for the second proc.