Observations with an FX-8350

[Idontcare]

There we have confirmation from guskline running the same linpack binaries. It has to be a some voltage setting problem in the BIOS(being much over recommended value). Or a dud CPU...

What is the recommend value for the voltage?

 

[Idontcare]

Just bare in mind TDP is about cooling not CPU power consumption. Its perfectly possible for a CPU to draw my power than its TDP up until its raised its temperature sufficiently that it needs to slow down. The cooling system must be able to at least remove tdp watts to be sufficient but a CPU can and will exceed that for periods of time. What a processor must do however is ensure that over seconds and minutes that it doesn't exceed TDP on average.

This is a sustained power-draw for more than an hour. 5 cycles of LinX with problem size 43122 at 4GHz. We aren't talking brief moments above 125W power use.

 

[guskline]

Interesting. Just to ensure we are talking apples-to-apples, your power numbers are with a kill-a-watt or software measure program utility? And did you run with problem size 43122 using 8 threads? What GFlops did you get? And lastly, what is your CPU voltage at load?

With DDR3-1866 10-10-10-28-T2, I'm currently peaking at ~85 GFlops w/8 threads. Kill-a-watt reports pulling ~287W, and my CPU voltage is ~1.377V. Loaded CPU temps are ~58C with 19C ambient.

Sorry Idontcare I'm now at work and not near the computer. I ran a short 5 run of the LinX program with the AMD changes per the link. I also have a Kill-O-Meter and did not measure cpu voltages during the run.

I don't have all the sofisticated equipment you do but I will try to replicate your run tonight. BTW, my ram was set at 1600 eventhough rated at 1866.

I'll try to be back tonight. You might want to try running a short sequence with the ram at 1600 to see if that makes a difference.

 

[guskline]

There we have confirmation from guskline running the same linpack binaries. It has to be a some voltage setting problem in the BIOS(being much over recommended value). Or a dud CPU...

I doubt a dud cpu. I'm thinking that Idontcare's Asus mb (top end) lives, breathes and sleeps overclocking SO I wonder if even at stock the BIOS is upping voltages?

I assume Idontcare has the latest BIOS.

I saved my OC'd settings as a profile in the BIOS and when I returned to STD I engaged the "Optimized Default" settings.

Idontcare, does your BIOS have an Optimized Default setting? You may want to try that.

 

[Idontcare]

Sorry Idontcare I'm now at work and not near the computer. I ran a short 5 run of the LinX program with the AMD changes per the link. I also have a Kill-O-Meter and did not measure cpu voltages during the run.

I don't have all the sofisticated equipment you do but I will try to replicate your run tonight. BTW, my ram was set at 1600 eventhough rated at 1866.

I'll try to be back tonight. You might want to try running a short sequence with the ram at 1600 to see if that makes a difference.

Any idea roughly on what ram setting you used? Was it the default 1GB?

If you can't measure the CPU voltage that is alright, I'd be happy just to hear your CPUz reported value when loaded. If yours is ~1.35V then that is one thing (being comparable to mine), if it is ~1.25 or even lower then that is a whole other thing and that would explain the power-usage differences.

Idontcare, does your BIOS have an Optimized Default setting? You may want to try that.

It is the latest BIOS, and I thought I told the BIOS to load optimized defaults (F5), but I will do it again just in case I failed at that before.

 

[guskline]

I'm trying to remember about the ram setting, I think 3 Gb but not sure. I'm going home for lunch so I can do a run of @1/2 hr. How do I measure the peak wattage on my Kill-O Watt without continually looking at the Kill-O Watt digital screen? I have the smaller Kill-O-Watt. Is there a software program from Kill-O-Watt?

As to the cpu voltage, under AIDA64 sensor readings, my cpu voltage with the Default optimized settings drop as low as .9 v. In the OC state (4.53Ghz ---21 x 215 with vcore at 1.4437) the v core stays at the 1.44 state.

I'll try to run both an Optimized Default state and an OC state and monitor both Max wattage and max vcore.

 

[jvroig]

I'm trying to remember about the ram setting, I think 3 Gb but not sure. I'm going home for lunch so I can do a run of @1/2 hr. How do I measure the peak wattage on my Kill-O Watt without continually looking at the Kill-O Watt digital screen? I have the smaller Kill-O-Watt. Is there a software program from Kill-O-Watt?

You should use all RAM possible because the load increases as you use more RAM, and so as to mirror IDC's setting. Don't worry about finishing the tests (using 14GB takes a long time), it just has to run for long enough so that the computation actually starts (as opposed to just allocating the RAM, which can take a minute or two).

For example, when I use ~14.6GB RAM (all RAM available option), looking at task manager shows the RAM usage climbing steadily (and rather slowly) until it reaches around 15GB+ and then it peaks there. After that, the power consumption shoots up by as much as 40W. That's when the real data crunching happens, and that pretty much characterizes your power draw accurately.

So just use all RAM, let it run for fifteen minutes while you have coffee or do something else, then just come back to it and then visually inspect what peak wattage is shown by the kill-a-watt. It naturally jumps around a bit, so just make a note of the values you see. You don't need to finish the test, you just need to make sure it is actually ON the test phase already, and not the "pre-test" / warm-up phase where it's just allocating RAM.

 

[guskline]

Thanks jvroig: Will do that. Had enough coffee this morning. There's some Chili in the refrig that will have to be heated up.

That should give it enough time to reach peak wattage. I'll report my findings. That's how I monitor IBT. When it first starts low wattages and then jumps just before a run is completed. i'll let it run @15 minutes stock and @ 15 minutes at my OC stage.

 

[grimpr]

Another great mem and stability tester is y-crunchers stress test mode, it catches memory controller instabilities that prime95 cant.

 

[Rvenger]

Any idea roughly on what ram setting you used? Was it the default 1GB?

If you can't measure the CPU voltage that is alright, I'd be happy just to hear your CPUz reported value when loaded. If yours is ~1.35V then that is one thing (being comparable to mine), if it is ~1.25 or even lower then that is a whole other thing and that would explain the power-usage differences.


It is the latest BIOS, and I thought I told the BIOS to load optimized defaults (F5), but I will do it again just in case I failed at that before.

Also, did you check to see if turbo core was functioning properly? I know Asus uses a multicore tweak to push all the cores to the maximum turbo setting.

 

[Magic Carpet]

My undervolted Thuban box consumes ~90W from the wall under heavy load. That includes 1 x 2.5" HDD, on-board GPU and a couple of memory sticks on a ~82% 330W PSU.

In my experience, the top ASUS boards have never been power efficient. They are geared towards the best overclock and stability, not the best power savings (you can't have three). Also, the "top-end" AMD chipset draws quite a bit more power than it's Intel counterpart. No surprises here.

 

[Idontcare]

I'm trying to remember about the ram setting, I think 3 Gb but not sure. I'm going home for lunch so I can do a run of @1/2 hr. How do I measure the peak wattage on my Kill-O Watt without continually looking at the Kill-O Watt digital screen? I have the smaller Kill-O-Watt. Is there a software program from Kill-O-Watt?

As to the cpu voltage, under AIDA64 sensor readings, my cpu voltage with the Default optimized settings drop as low as .9 v. In the OC state (4.53Ghz ---21 x 215 with vcore at 1.4437) the v core stays at the 1.44 state.

I'll try to run both an Optimized Default state and an OC state and monitor both Max wattage and max vcore.

Yeah you have to watch the screen by eye but you get a quick feel for it one the LinX program finishes filling up the ram (watch task manager). You'll have a decent idea of what the peak wattage is within 10-15 minutes or so.

I go to a little more trouble and I setup a webcam that captures the kill-a-watt output as a video feed, which is then shown on the desktop in the webcam window.

Then I have an automatic screenshot program running in the background that captures a screenshot every 15 seconds.

That way I have a record of the power-usage, cpu voltage, and ambient temperature every 15 seconds during the LinX stress testing. It is easy to flip through them like a slideshow afterwards.

Also, did you check to see if turbo core was functioning properly? I know Asus uses a multicore tweak to push all the cores to the maximum turbo setting.

CPU-z and ASUS AI Suite confirm turbo-core is functioning correctly.

 

[guskline]

OK. On my lunchbreak and ran the LinX program filling up ram and same problem size.
Stock:
Max watts - 241
vcore - 1.296
CPU Temp Overall 41C cores 27

OC to 4.53 Ghz with vcore set to 1.4437 for max stability

Max Watts - 344
vcore - 1.44
CPU Temp Overall - 51C core 41 C

Hope this info helps, BTW I ran each test @20 minutes.

 

[MentalIlness]

guskline...

Would it be worth while for me to upgrade to a FX 8350 from a FX 8120 ? Right now it is running stock speeds. But it can go as far as 4.5 GHz stable. Ran it there for a few weeks stable.

 

[Idontcare]

OK. On my lunchbreak and ran the LinX program filling up ram and same problem size.
Stock:
Max watts - 241
vcore - 1.296
CPU Temp Overall 41C cores 27

OC to 4.53 Ghz with vcore set to 1.4437 for max stability

Max Watts - 344
vcore - 1.44
CPU Temp Overall - 51C core 41 C

Hope this info helps, BTW I ran each test @20 minutes.

Your stock temps and Vcore are astonishingly lower than mine.

According to Coretemp, the VID for my FX8350 is 1.3875V and it peaks around 60C under IBT.

Did you happen to notice the GFlops for your stock run? I'd be curious to know if you are seeing ~85GFlops or if perhaps you end up with lower GFlops because your mobo engages a current restrictor mechanism of some kind to the CPU.

 

[Ed1]

be interesting to see his VID at stock and OC speed .

Maybe AMD has big delta between chips, some need higher voltages , or at least AMD thinks so .

 

[guskline]

guskline...

Would it be worth while for me to upgrade to a FX 8350 from a FX 8120 ? Right now it is running stock speeds. But it can go as far as 4.5 GHz stable. Ran it there for a few weeks stable.

That's a tough one to answer. If you can run your 8120 stable at 4.5 Ghz upgrading will not result in a much higher OC, even with a 8350 unless you have a custom water cooling system. I use a Corsair H100 and I notice that you use a H80. My 8350 is solid at 4.53Ghz and uses less power than my 8150 did. Where you will notice the difference most is at stock. The 8350 really runs very nice at stock and "feels" faster than the 8150 did at stock.

It's really a question of money.

 

[guskline]

Your stock temps and Vcore are astonishingly lower than mine.

According to Coretemp, the VID for my FX8350 is 1.3875V and it peaks around 60C under IBT.

Did you happen to notice the GFlops for your stock run? I'd be curious to know if you are seeing ~85GFlops or if perhaps you end up with lower GFlops because your mobo engages a current restrictor mechanism of some kind to the CPU.

I'll check that tonight. BTW, what cpu cooler are you using? I have a Corsair H100 with 4 fans with cool air intake at the top of the CM HAF 932 Advanced. From what I read you need a decent water cooling system to keep the heat from spiking when you crank up these 8 core/ 4 module systems.

 

[MentalIlness]

That's a tough one to answer. If you can run your 8120 stable at 4.5 Ghz upgrading will not result in a much higher OC, even with a 8350 unless you have a custom water cooling system. I use a Corsair H100 and I notice that you use a H80. My 8350 is solid at 4.53Ghz and uses less power than my 8150 did. Where you will notice the difference most is at stock. The 8350 really runs very nice at stock and "feels" faster than the 8150 did at stock.

It's really a question of money.

Well it isn't like the 8120 doesn't do what I need it to. I guess I am just trying to get the most out of the setup as possible. I have the Sabertooth 990FX as well. If I needed a motherboard, I would probably be going with Intel. But since the board is fairly new then no sense to replace it. Hence why I am considering the 8350.

And yes, I use a H80. Keeps the 8120 nice n' cool at stock and load.

I guess I just have the upgrade bug. I should probably upgrade my graphics card first Id think.

And no, money isn't an issue. Just doing some debating here.

 

[Idontcare]

I'll check that tonight. BTW, what cpu cooler are you using? I have a Corsair H100 with 4 fans with cool air intake at the top of the CM HAF 932 Advanced. From what I read you need a decent water cooling system to keep the heat from spiking when you crank up these 8 core/ 4 module systems.

Right now I am just using the stock HSF that came with the FX8350.

Once I know for certain that the numbers are not artifacts of some mobo voodoo, that they are bonafide power numbers for the cpu, then I'll worry about doing non-stock things with it (including OC'ing).

 

[sb.0326]

Figured I'd go ahead and post my results with my 8350. Now I do have a bunch of items(monitors and printers etc.) plugged in to the powerstrip that I had plugged in to Kill-A-Watt but they were all running at the same time so the results are what they are.

I use three different power profiles depending on my usage..

Normal everyday use is an underclocked undervolted 3.5ghz with my ram running at 1333. Idle power consumption was 240 and Load was 309.

For encoding purposes I have another profile set at 4.5ghz and RAM speed at 1600. Idle power consumption was 242 and Load at 442.

Now for gaming I have it set at 5.1ghz with NB and Hypertransport both at 2600. And RAM at 1866. Now this isn't "technically" stable. It won't pass IBT. The computer will just turn off after awhile. But all I care about is can I game on it without it bluescreening on me which it doesn't so to me that is stable "enough". Idle power consumption is 294 and load is 495.

My kid brother has a nice LGA 2011 with a 3930k. Gonna have him check in a sec and I'll post his results.

 

[Exophase]

A few of the watts in the idle to load delta will be used by the RAM, instead of the CPU or VRMs.

 

[guskline]

Your stock temps and Vcore are astonishingly lower than mine.

According to Coretemp, the VID for my FX8350 is 1.3875V and it peaks around 60C under IBT.

Did you happen to notice the GFlops for your stock run? I'd be curious to know if you are seeing ~85GFlops or if perhaps you end up with lower GFlops because your mobo engages a current restrictor mechanism of some kind to the CPU.

Having trouble with the Gflop reading but at stock with your settings watts were 251

 

[grimpr]

Enabling HPC mode in am3+ motherboards where supported disables CPU throttling due to vrm protection.

 

[Martimus]

I also built a Beowolf cluster in 99, although it was to do an image processing calculation for lane finding at real time. I only had moderate success, as my filter calculations weren't the best.

EDIT: More accurately, I helped my roommate build a Beowulf cluster out of spare computers from the computer lab. I also think it was in 2000, not 1999 now that I think about it.