7970 undervolting - page 2. | Bitcointalksearch.org

ddd1

full member

Activity: 154

Merit: 100

Quote from: lightlord on May 13, 2012, 11:11:12 PM

at 1230, it crashes within 5 seconds
at 1225, it crashes within 45 seconds
at 1220, it crashes within 2 mins with strange artifacts
at 1215, Currently running haven't crashed yet (Updated Just crashed)
at 1210, Trying now (crashed 10 mins)
at 1205, (crashed 10 mins)
at 1200 (20mins)

Threw the clock to 1190, hopefully it will be stable 24/7 at this when I go to sleep.

So you are testing stability with mining?

Which drivers cause PC to just crash and not shut off?

lightlord

donator

Activity: 3228

Merit: 1226

★Bitvest.io★ Play Plinko or Invest!

at 1230, it crashes within 5 seconds
at 1225, it crashes within 45 seconds
at 1220, it crashes within 2 mins with strange artifacts
at 1215, Currently running haven't crashed yet (Updated Just crashed)
at 1210, Trying now (crashed 10 mins)
at 1205, (crashed 10 mins)
at 1200 (20mins)

Threw the clock to 1190, hopefully it will be stable 24/7 at this when I go to sleep.

lightlord

donator

Activity: 3228

Merit: 1226

★Bitvest.io★ Play Plinko or Invest!

Now going to push my card as high as it goes in core clock before crash. Lucky the new drivers just make it not respond and I don't have to restart computer really nice for over clocking.

Btw over clocking isn't what is dangerous, its overvolting that is.
So If I yeilded 1154 MV from 1174 MV at 1180, at stock 1174 MV I should get around 1200 +

lightlord

donator

Activity: 3228

Merit: 1226

★Bitvest.io★ Play Plinko or Invest!

Okay I have tempered, and lowered the clocks very slowly, Btw the longer you run the hard ware the lower the voltage can go. After one year my Cpu was able to undervolt from 1.45>1.1875 Stable.
Where it used to not be stable at 1.35 or so. Its interesting.

7970 Stable Clocks
---------------------
1180 Core Clock, 700 Mhash, Msi after burner: 1174 -> 1156 MV -> Gpu-Z sensor: VDDC 1.04 Volts (Any lower Crashes)
1100 Core Clock, 650 Mhash, Msi after burner: 1174 -> 1068 MV -> Gpu-Z sensor: VDDC 0.988 Volts (Any lower Crashes)
1025 Core Clock, 600 Mhash Msi after burner: 1174 -> 1024 MV -> Gpu-Z Sensor: VDDC 0.949 Volts (Any lower Crashes)

Seems like a diminishing return curve, around Core 1100, I yield the most mhash to least power/voltage.
1180 being least efficient, lower then 1100 being less efficient.

jjiimm_64

legendary

Activity: 1876

Merit: 1000

Quote from: lightlord on May 13, 2012, 08:15:43 PM

I have a wattage meter device, and dropping 1174>1156 drops watts, but actually putting in 0.86 in Cgminer does completely nothing.
Wattage does stay same. Stayed at 430>430, no drop.

I really think you are miss looking over the numbers. Stock at 0.936 Volts is possible, but 1180/1100 at 0.86 is not possible.

i would have to agree. i am doing 930 at 0.937v. at 940 I was dropping cards. at 930 you can see from my last post the rig has been up for days now.

lightlord

donator

Activity: 3228

Merit: 1226

★Bitvest.io★ Play Plinko or Invest!

I have a wattage meter device, and dropping 1174>1156 drops watts, but actually putting in 0.86 in Cgminer does completely nothing.
Wattage does stay same. Stayed at 430>430, no drop.

I really think you are miss looking over the numbers. Stock at 0.936 Volts is possible, but 1180/1100 at 0.86 is not possible.

ddd1

full member

Activity: 154

Merit: 100

Do you guys use gpu-z/hwinfo64 to determiner the "real" VDDC?
Because It seems if you put VDDC in MSI AB it shows different in GPU-z.

Also Ive read there are different voltage leak on 7970, if memory serves me correct there is 1.17v, 1.111v and 1.05v.
So if you have the 1.111v card then putting 1.117v in Afterburner in reality gives 1.111v.

So if you put +0.020v in MSI AB so it is 1.190v in truth if you have a 1.111v card the "real" VDDC is 1.131v.

I think the above is true?

Graet

vip

Activity: 980

Merit: 1001

Quote from: lightlord on May 13, 2012, 03:56:48 AM

Quote from: Graet on April 06, 2012, 01:45:43 AM

cgminer.exe -o http://us.ozco.in:8332 -u graet.20 -p p -o http://eu.ozco.in:8332 -u graet.20 -p p -I 9 -k poclbm --auto-fan --gpu-engine 1140-1180 --auto-gpu --gpu-memdiff -150 --gpu-vddc 0.88 --temp-overheat 95

gives me 704Mhash @ 74 deg on one card, 680mh @ 83deg on other.

Actually found out to be false. I got the exact same temperatures.
But I checked gpuz sensors and found out cgminer under-volt does nothing.
How ever I tried msi after burner lastest version undervolting and going down any where near 0.88 just crashes
and ati driver stops responding. I did mange to go from 1174>1156
A read out drop on the sensor from 1.07>1.04 VDDC

And any lower it crashes at 1180 OC

The above numbers are impossible, just fallacy.
I thought it was working too as well, and I realized it wasn't
Download GPU-Z

You aren't going mange a 20% under volt with a 930>1180!
Dream on, getting a 15% higher clock and 20% under volt is everyone's dream.
That simply can't be done.

I CANNOT run the cards at stock clocks with those settings, I start gpuz hashrate drops and cards crash

BUT the wattmeter plugged into the wall reads substantially lower, even though cgminer reports no change, proof is in the draw from the wall. fallacy all you like

jjiimm_64

legendary

Activity: 1876

Merit: 1000

I am running them at 930/800/0.937v

downvolt with msiafterburner, close AB start cgminer with conf file.

pulling 830watts at the wall. 3.33 Mh/watt

Code:

cgminer version 2.3.6 - Started: [May 10, 2012, 8:52 am] Rig:miner19
(5s):2772.32 (avg): 2759.81 Mh/s | H: 86.8 Q:168959 A:173544 R:927 HW:113 E:?% U:38.07/m
TQ:? ST:143 SS:? DW:5646 NB:424 LW:80863 GF:222 RF:169
Connected to http://gpumax.com:8332 with LP as user ?
Value:
GPU 0: 72.0C 2504RPM 42% 114 | 554.3/552.1Mh/s | 99% | 930Mhz 800Mhz 1.17V A:34667 R:190 HW:0 U:7.60/m I: 9
GPU 1: 66.0C 1052RPM 0% 66 | 554.8/552.1Mh/s | 99% | 930Mhz 800Mhz 1.17V A:34859 R:198 HW:0 U:7.65/m I: 9
GPU 2: 72.0C 1184RPM 33% 105 | 554.7/552.1Mh/s | 99% | 930Mhz 800Mhz 1.17V A:34519 R:196 HW:88 U:7.57/m I: 9
GPU 3: 67.0C 1042RPM 0% 67 | 554.0/551.6Mh/s | 99% | 930Mhz 800Mhz 1.17V A:34762 R:171 HW:25 U:7.63/m I: 9
GPU 4: 72.0C 1092RPM 10% 82 | 554.6/552.0Mh/s | 98% | 930Mhz 800Mhz 1.17V A:34737 R:172 HW:0 U:7.62/m I: 9

lightlord

donator

Activity: 3228

Merit: 1226

★Bitvest.io★ Play Plinko or Invest!

Quote from: Graet on April 06, 2012, 01:45:43 AM

cgminer.exe -o http://us.ozco.in:8332 -u graet.20 -p p -o http://eu.ozco.in:8332 -u graet.20 -p p -I 9 -k poclbm --auto-fan --gpu-engine 1140-1180 --auto-gpu --gpu-memdiff -150 --gpu-vddc 0.88 --temp-overheat 95

gives me 704Mhash @ 74 deg on one card, 680mh @ 83deg on other.

Actually found out to be false. I got the exact same temperatures.
But I checked gpuz sensors and found out cgminer under-volt does nothing.
How ever I tried msi after burner lastest version undervolting and going down any where near 0.88 just crashes
and ati driver stops responding. I did mange to go from 1174>1156
A read out drop on the sensor from 1.07>1.04 VDDC

And any lower it crashes at 1180 OC

The above numbers are impossible, just fallacy.
I thought it was working too as well, and I realized it wasn't
Download GPU-Z

You aren't going mange a 20% under volt with a 930>1180!
Dream on, getting a 15% higher clock and 20% under volt is everyone's dream.
That simply can't be done.

Roadhog2k5

full member

Activity: 131

Merit: 100

Quote from: cmg5461 on April 05, 2012, 05:46:49 PM

I found HWinfo was pretty much in line with my calculated figures.

https://bitcointalksearch.org/topic/3x7970-mining-results-57410

Graet

vip

Activity: 980

Merit: 1001

cgminer.exe -o http://us.ozco.in:8332 -u graet.20 -p p -o http://eu.ozco.in:8332 -u graet.20 -p p -I 9 -k poclbm --auto-fan --gpu-engine 1140-1180 --auto-gpu --gpu-memdiff -150 --gpu-vddc 0.88 --temp-overheat 95

gives me 704Mhash @ 74 deg on one card, 680mh @ 83deg on other.

cmg5461

sr. member

Activity: 369

Merit: 250

bulanula

hero member

Activity: 518

Merit: 500

Quote from: Buckwheet on April 05, 2012, 02:11:34 PM

Quote from: bulanula on April 05, 2012, 11:39:01 AM

Quote

xfx 7970 black edition. 76.7% ASIC quality

How do you get ASIC quality in the 7xxx cards ?

Thanks !

Its built into GPU-Z as far as I am aware.

I meant with Linux.

Just like you cannot use GPU-Z to see VRM temperatures.

Has anyone tried with WINE and GPU-Z Huh

DeathAndTaxes

donator

Activity: 1218

Merit: 1079

Gerald Davis

Quote from: cmg5461 on April 05, 2012, 03:18:12 PM

Quote from: Gomeler on April 05, 2012, 11:23:41 AM

Quote from: DeathAndTaxes on April 05, 2012, 09:20:03 AM

85% efficiency? Get a better PSU.

Also would be nice if you have/get a Kill-A-Watt so you can compare calculated load @ wall w/ actual load at wall.

Very interesting. Also interested in the actual load as I don't believe the input power for the memory, while tiny, is included. Might be time to pick up a used 7970 myself.

Does not matter. Voltage is read going into the card as well as the amperage. Memory voltage is drawn off that main voltage.

Of course it matters. The recorded CURRENT is VDDC which does not power the memory (or I/O). A separate VRM provides CURRENT for memory, a third provides voltage for IO functions. While the voltage remains the same without knowning the VDDCI & MVDDC current you don't know entire card consumption.

Total Power Consumption = Voltage * (VDDC current + VDDCI current + MVDDC current)

cmg5461

sr. member

Activity: 369

Merit: 250

Quote from: Gomeler on April 05, 2012, 11:23:41 AM

Quote from: DeathAndTaxes on April 05, 2012, 09:20:03 AM

85% efficiency? Get a better PSU.

Also would be nice if you have/get a Kill-A-Watt so you can compare calculated load @ wall w/ actual load at wall.

Very interesting. Also interested in the actual load as I don't believe the input power for the memory, while tiny, is included. Might be time to pick up a used 7970 myself.

Does not matter. Voltage is read going into the card as well as the amperage. Memory voltage is drawn off that main voltage.

scifimike12

sr. member

Activity: 303

Merit: 250

Quote from: DeathAndTaxes on April 05, 2012, 09:20:03 AM

85% efficiency? Get a ~~better~~ more efficient PSU.

Fixed that for you. Remember, just because a PSU is rated at 80+ Bronze, Silver, Gold, etc. doesn't mean it is "high quality". I'd rather take an Antec TruePower New over a ABS Majesty or Mushkin Enhanced any day of the week.

Quote from: bulanula on April 05, 2012, 11:39:01 AM

How do you get ASIC quality in the 7xxx cards ?

Thanks !

Right-click the toolbar and select Read ASIC quality...

Buckwheet

full member

Activity: 169

Merit: 100

Green Mining 85% Cheaper

Quote from: bulanula on April 05, 2012, 11:39:01 AM

Quote

xfx 7970 black edition. 76.7% ASIC quality

How do you get ASIC quality in the 7xxx cards ?

Thanks !

Its built into GPU-Z as far as I am aware.

bulanula

hero member

Activity: 518

Merit: 500

Quote

xfx 7970 black edition. 76.7% ASIC quality

How do you get ASIC quality in the 7xxx cards ?

Thanks !

Gomeler

hero member

Activity: 697

Merit: 500

Quote from: DeathAndTaxes on April 05, 2012, 09:20:03 AM

85% efficiency? Get a better PSU.

Also would be nice if you have/get a Kill-A-Watt so you can compare calculated load @ wall w/ actual load at wall.

Very interesting. Also interested in the actual load as I don't believe the input power for the memory, while tiny, is included. Might be time to pick up a used 7970 myself.

Topic: 7970 undervolting - page 2. (Read 17563 times)