Magny Cours.. Finally running

[Movieman]

The power usage HWmonitor displays is a lie imo. Says my 1090T pulls 110w when it pulls 320 from the wall.
Take it with a grain of salt.

ahhh

Hey Sam, BUY A KILL-A-WATT PLEASE!

[Sam_oslo]

ahhh

Hey Sam, BUY A KILL-A-WATT PLEASE!

I think the HM CPUID is reporting the right value for CPU, but the total system draw is of course more.
I'll try, gimme a few days to find one in stock first.

[Movieman]

I think the HM CPUID is reporting the right value for CPU, but the total system draw is of course more.
I'll try, gimme a few days to find one in stock first.

Make sure you get the "International" model that works with 220v..
App $35.00 here in the US..

[slaveondope]

I think the HM CPUID is reporting the right value for CPU, but the total system draw is of course more.
I'll try, gimme a few days to find one in stock first.

I didnt catch if you were overclocked or not but Im pretty sure there is no way in hell my 1090t at 4.1ghz with 1.52v is only using 110w from the board.
I think the value is preset based on stock consumption for the cpu.

[Sam_oslo]

I didnt catch if you were overclocked or not but Im pretty sure there is no way in hell my 1090t at 4.1ghz with 1.52v is only using 110w from the board.
I think the value is preset based on stock consumption for the cpu.

Yes, I'm OCing @4,18GHz 1.31v for 24/7.
The older HM CPUID had a general CPU-power report, which is still there at the bottom and is reporting 81.45W on both idle and load. It is wrong, but the new ver. 1.16 has added special readings for for GA MBs (it's called Gigabyte DES), at the top.
None of the power-readings is based on stock, because they changes with different vCore/multi/OC, and the GA DES readings seams to be right, but i don't know how accurate it is actually.
I guess you are looking at what it's reporting at bottom, but that's not accurate on my MB either.

[Sparky]

I always recommend taking any software reading with a grain of salt. If the software can't even get voltage readings from the power supply right most of the time, how can we trust it on the power reading? If anything, the power reading is harder to get, not easier, so I'd see even more room for error there.

JMO.

[jcool]

I would still guess that the X5650's on turbo at 2800MHz would do well over 65,000PPD in WCG

We don't have to guess, I've been running a stock Dual X5650 on a Tyan for a while now, remember
Turbo jumps around between 2800 and 2933Mhz, it does exactly 72k/d at 305W from the wall.

As for software power readings, they only work very very rarely. Some boards with digital VRM solutions have a readout of how much current passes through at a given time (DFI, for example). That way you can calculate with P=U*I (CPU Core voltage * current) to give you a pretty good idea on CPU power draw - however, that still doesn't factor in increased VTT and Vdimm voltages (which increase the CPU's power draw as well, especially increased VTT). In short: Software can't replace a good kill-a-watt.

[onex]

just a few points to add,
there's the 6128 opteron 8 cores 2.0 Ghz Magny Cours processors priced 264-300$ capable of a 4-way setup.
the motherboards are quite pricey at about 900$ range.

though 32 cores at 2.0Ghz might give ~72K PPD at WCG from a short calculation.

the price for the dual 1.9 12 core and the quad 2.0 including the motherboard is about the same.

there isn't really any sweet spot here,
intel's CPU's are highly priced, spending a 1000 dollars per CPU for a dual setup is no joke,
as for AMD, the processors are aimed at the hi-end server market, not overclockable (yet) and terribly low clocked.

it might be that 12 cores generate much more heat then 6+6 (with Hyper-Threading) and as well draw much more power, so AMD aiming at the server market had this in mind.
you can have a look at the 6176 specs, the TDP is growing to 137 already at 2.3Ghz, the TDP for the 6128 and the 6174 (12 cores) is the same, the 6174 is running 2.2 stock.

what is strange with these figures, is that the TDP for both the 8 core and 12 core chips is the same at 1.9-2.2Ghz,
AMD had to reduce the core clock to 1.7 for the HE 12 core 6164 edition.

[Chumbucket843]

I always recommend taking any software reading with a grain of salt. If the software can't even get voltage readings from the power supply right most of the time, how can we trust it on the power reading? If anything, the power reading is harder to get, not easier, so I'd see even more room for error there.

JMO.

this is true. computers will give you 10 sig figs of precision but they can be quite inaccurate.

on the other hand a cpu like itanium has quite amazing instruments for measuring power consumption and has 64 different clockspeeds. hopefully this will one day end up in x86 cpu's.
http://en.wikipedia.org/wiki/Foxton_Technology

[Enoc]

Ijust went and priced a top MC 6176HE cpu. $1628.76 from a quick look.

i was thinking that with 150$ more...you can buy 4 x 8-core 6128 2ghz MC (in the web around 265-280 shipped) and a quad socket G34 motherboard...you can find them around 700-800...just like Onex said it...

Movieman what would be the estimate of PPD production of a 32 core magnycours system and power consumption? would it be around the same as a stock SR-2 with x5680 or 5670 or 5660?

one advantage I see with the AMD G34 system is that next AMD architecture is a drop in...so that's a big plus...

[jcool]

The thing is that you won't _ever_ run an SR-2 at stock speeds. That would totally defeat its purpose. If you want to compare stock Dual LGA1366 and Dual/Quad Magny-Cours for price/performance, you should calculate with 250$ for a Tyan S7002 motherboard (or similar) which is less than half the price of the SR-2 (and also uses less power).

If prices are about the same, I'd say go with Quad socket (if you can find a case to fit it...) just for the heck of it

[lopium]

like this thread, really instructive...

[Movieman]

like this thread, really instructive...

That machine is just chuuging along every day doing it's 50K a day..
Went down stairs the other day, season's changed and it's colder here in NH
Room was 53F..
Coretemp said the machine was running at 33C at 100% load..

[OldChap]

53f....Aw heck, doncha know? that is just your house telling you that that room needs another rig