Crunchers Helping Crunchers

[twilyth]

Please, I am really concerned about power consumption.

Thanks - please see my edit.

Are you interested in a similar chip and concerned about power use?

 

[FireFox]

Thanks - please see my edit.

Are you interested in a similar chip and concerned about power use?

I just want to compare your Machine's power Comsuption vs my Machine' power consumption

 

[twilyth]

OK. I'll dig out one of the meters and try to do a proper test but it will be off of the ups. I have another rig plugged into that which I think has caused the problem in the past.

Believe me, power usage is an issue for me as well. That's one of the reasons it's been like 3 years since my last build. Also a reason why I was holding out for a pair of 14c chips. I don't think there will be any surprises but I'll check.

 

[twilyth]

OK, I did it but the results don't make any sense. According to one watt meter, the 2695v3 is drawing 390w. That's 160w over the tdp for the chips. That means that video card, ram and hdd are drawing 160w. Video is only displaying windows so it's basically idle and I don't see how the memory (32g total on 4 ECC 2133 chips - which don't even have spreaders).

I know that the dual 2687w would draw an extra 200w on a full load on 2 HD7850(?). But idle I don't think it drew much more than the total TDP.

Turns out that the other watt meter is connected to the dual hex core but that looks like it's reading voltage not watts. Will have to check that next time I'm in there.

 

[theonedub]

Are you accounting for the inefficiency of the PSU in that?

 

[twilyth]

I'll have to check. I could be wrong but I think it's a gold Antec 1kw. I'll check.

 

[twilyth]

OK, dual hex is running at about 290w. The 14c is still almost 390w. But the psu is an Antec High Current Pro 1200w which seems to be only 80% efficient (80+ gold).

 

[Heaven7]

Looks like your new build is really doing great, @twilyth . Congratulations!

 

[twilyth]

It's drawing a lot more power than I had expected - by about 100w, but that's fine. It should produce almost as much output as all of the other rigs combined - well, 3/4s at least. So I should be able to retire one of the other servers, probably the dual quad core. However that's really just speculation right now. My original estimate for output was is the low to mid 20k ppd range. Now it looks like it might be closer to 30k. So I'm really anxious to see what the output looks like after it's stabilized between pendings and valids.

I'm still in the market for another set of 14c chips but there seems to be a temporary lull. Right now I'm only seeing 3 listings on ebay and 2 of those are at more than $1100 each. The only one that's reasonable is at $775 but there's only one chip. I paid less than $600 each for my set. There are some 18c chips but both are still very expensive. The cheapest is about $1200 @2.3ghz.

 

[thebluebumblebee]

It's drawing a lot more power than I had expected - by about 100w

Could that be because of the ES chips? Does that pose a danger to the motherboard?

 

[twilyth]

Could that be because of the ES chips? Does that pose a danger to the motherboard?

It could be because they are ES. I've had that issue at least once before. But these don't seem to be first gen ES. From the listing -

I'm not really worried about the m/b but I guess we'll see. Supermicro has always been very solid in the past and I do have a small fan pointed at it, targeting the heatsinks. Plus chip temps are stable at between 55 and 65C. The wide difference is because one fan is getting the exhaust from the other. I did have a reason for doing that but at somepoint I'll probably change that around.

 

[Toothless]

The wide difference is because one fan is getting the exhaust from the other. I did have a reason for doing that but at somepoint I'll probably change that around.

Cardboard separator?

 

[Norton]

It could be because they are ES. I've had that issue at least once before. But these don't seem to be first gen ES. From the listing -

I'm not really worried about the m/b but I guess we'll see. Supermicro has always been very solid in the past and I do have a small fan pointed at it, targeting the heatsinks. Plus chip temps are stable at between 55 and 65C. The wide difference is because one fan is getting the exhaust from the other. I did have a reason for doing that but at somepoint I'll probably change that around.

Any settings available in the bios to reduce voltage on the cpu at full load? A little bump down may decrease heat and power usage by a decent amount...

 

[twilyth]

Cardboard separator?

Any settings available in the bios to reduce voltage on the cpu at full load? A little bump down may decrease heat and power usage by a decent amount...

Great ideas guys. @Toothless - There's only about 5mm of clearance between the rads but that probably doesn't matter since the air would spill out the sides from the front rad. The only concern would be if the intake on the second one might be impeded. A very thin piece of non-corrugated cardboard might work. Another thought is whether or not the cardboard might get sucked up against the second fan. It's worth checking out.

The best solution would be to turn the f/hs 90 or 180 degrees. The problem with 90 degrees is that then the base of the hs will be going across the narrow part of the chip and I'm afraid it won't cool as well since less of the ihs will be covered. The problem with 180 degrees is that then the rad fans are blowing into each other, but that might not be a big deal. I just have to get motivated enough to try it. Motivation is a real issue for me. So I'm probably going to wait until I get some stable production numbers before I do anything. The downstream chip isn't hot, it's just a little warmer than I would like. Low to mid 60's isn't bad.

@Norton - That's an idea. I haven't done anything at all in the bios so I don't know what my options are. I'm hoping that I might be able to bump the multiplier one notch and get the chips closer to 3ghz w/o having too much affect on the temps. But I'll also see what I can do about voltage. I like the idea of cutting back on the electricity usage. I know that sounds contradictory but I'm open to both.

What sort of incremental voltage change would you recommend assuming I can do anything about that?

 

[Toothless]

I wonder if you can somehow make the exhaust fan somewhat sideways away from the 2nd cooler. I assume you have dual-fan setups per heatsink.

Tilt the exhaust fan away from the 2nd heatsink intake and put a wall/barrier so the heated air won't flow into the 2nd heatsink's intake fan. Just an idea.

 

[Norton]

What sort of incremental voltage change would you recommend assuming I can do anything about that?

Those chips specify a voltage range of 0.65-1.3 between its various states- if it were my chip I would try increments of 0.03-0.05 volts and check for errors/BSOD'S, BOINC errors, and temp/power drops. Would probably let it run a day or two at each step to make sure it's stable.

SuperMicro may do what some Gigabyte boards do- run the chip at a slightly higher voltage in auto to guarantee stability at the cost of heat and power. My GB AM3+ boards always seemed to run the chips hot in auto

 

[twilyth]

I wonder if you can somehow make the exhaust fan somewhat sideways away from the 2nd cooler. I assume you have dual-fan setups per heatsink.

Tilt the exhaust fan away from the 2nd heatsink intake and put a wall/barrier so the heated air won't flow into the 2nd heatsink's intake fan. Just an idea.

Thanks. This particular f/hs has a narrow channel on one side of the heat fins that the fan clips into so there's really no way to do that. But I'm still liking the cardboard idea.

Those chips specify a voltage range of 0.65-1.3 between its various states- if it were my chip I would try increments of 0.03-0.05 volts and check for errors/BSOD'S, BOINC errors, and temp/power drops. Would probably let it run a day or two at each step to make sure it's stable.

SuperMicro may do what some Gigabyte boards do- run the chip at a slightly higher voltage in auto to guarantee stability at the cost of heat and power. My GB AM3+ boards always seemed to run the chips hot in auto

Thanks for that info. I'll look to see if the voltage is dynamically adjusted. I dl'ed the manual.

I hate to turn the rig off for even a few minutes right now until I get some ppd stats. Also I think about all of those threads that could be running. Plus there's high probability of my screwing something up. that's pretty much std operating procedure for me. LOL.

 

[Toothless]

Thanks. This particular f/hs has a narrow channel on one side of the heat fins that the fan clips into so there's really no way to do that. But I'm still liking the cardboard idea.

Thanks for that info. I'll look to see if the voltage is dynamically adjusted. I dl'ed the manual.

I hate to turn the rig off for even a few minutes right now until I get some ppd stats. Also I think about all of those threads that could be running. Plus there's high probability of my screwing something up. that's pretty much std operating procedure for me. LOL.

Plastic wrap the side? Tin foil?

 

[xvi]

Plastic wrap the side? Tin foil?

I vote for liquid.

 

[twilyth]

Plastic wrap the side? Tin foil?

It looks like this isn't going to be an option. I turns out that the fins are shaped so that they have small ledges on each side. That effectively closes the sides so air can only pass straight through. Actually I knew this already but I hadn't made the association. So my only option, at least with these particular f/hs's, is going to be to change the orientation.

I did move the clip-on fan so that most of the air is forced between the coolers. Hard to explain but following the direction of the air flow, one socket and therefore the f/hs is slightly offset from the other. So the overlap between them isn't complete, more like 3/4's.

I vote for liquid.

I would consider that if I could get a sealed all-in-one type unit that could handle 2 cpu's on the loop. But if I build one myself, I assume there's a certain amount of maintenance involved.

I am probably going to look for another cooler than has more contact surface area but I don't ever see that specified anywhere. If anyone has any suggestions though, I'd be very interested. Bear in mind that the chips are about 2"x1.5". So we're talking about a fair amount of area. Also it has to be lga2011 compatible of course.

 

[Norton]

I am probably going to look for another cooler than has more contact surface area but I don't ever see that specified anywhere. If anyone has any suggestions though, I'd be very interested. Bear in mind that the chips are about 2"x1.5". So we're talking about a fair amount of area. Also it has to be lga2011 compatible of course.

I would likely look for something with a solid base rather than HDT. With that said, this one has a really slim profile and is rated for 180w (HDT though):

Thermaltake NiC L32
http://www.newegg.com/Product/Product.aspx?Item=N82E16835106235&cm_re=nic_l32-_-35-106-235-_-Product

I have one brand new (review sample) that you're welcome to have and try out.

 

[Toothless]

Two DeepCool Lucifers should be big enough.

 

[twilyth]

I would likely look for something with a solid base rather than HDT. With that said, this one has a really slim profile and is rated for 180w (HDT though):

Thermaltake NiC L32
http://www.newegg.com/Product/Product.aspx?Item=N82E16835106235&cm_re=nic_l32-_-35-106-235-_-Product

I have one brand new (review sample) that you're welcome to have and try out.

The Thermaltake looked really good until I checked the measurements. The fan is 40mm deep and so is the rad for a total of 80. The current combo is 84mm. I do like the bigger base though. I also like that there is some open space on the sides. I'm going to have to think about those. It's rated for 30 more watts than the Arctic Freezer i11's too so I assume it compares favorably. Hmmm.

Two DeepCool Lucifers should be big enough.

Yeah, those look massive but they'd never fit - which I assume you already guessed.

 

[thebluebumblebee]

The fan is 40mm deep

No, it's 25mm. http://www.thermaltakeusa.com/products-model_Specification.aspx?id=C_00002247

 

[t_ski]

I did have a reason for doing that but at somepoint I'll probably change that around.

Cardboard separator?

I wonder if you can somehow make the exhaust fan somewhat sideways away from the 2nd cooler. I assume you have dual-fan setups per heatsink.

Tilt the exhaust fan away from the 2nd heatsink intake and put a wall/barrier so the heated air won't flow into the 2nd heatsink's intake fan. Just an idea.

Building on these ideas, I give you this:

The black are the heatsinks and the red are the fans (assuming one fan per heatsink). Motherboard is at the bottom. Reverse the fan on one of them, and insert the baffle between them to deflect the air down & away from the first one. That should let the second one pull in cooler air from above.