• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

7970 overclocking woes

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
moriz said:
That's the one I'm on right now actually.

And it made no difference. The card simply do not go to 1150mhz, even with 1.275v. I'm really hesitant to try 1.3v, and I'm not sure if it can make a difference anyhow.
Click to expand...

I had the same issue where adding more volts didn't help stability. My core temps were pretty decent with the fan cranked so I'm pretty sure it was the VRM's getting too hot. After I threw a waterblock on it the overclocking headroom got much better.

On the reference cooler it maxed at 1225Mhz@1.175V.
 
IIRC, Furmark is flagged by the drivers as a "power virus" (that's why you had to change the EXE's name for the 5800 series) and the cards behave differently in order to run it.
so no, furmark is not a good indication.
 
I'll ask the obvious question, you are using manual fan right? You CANNOT use auto fan for overclocking, chances are you need manual fan 50% or possibily higher. HardOCP required 70% fan to get 1250 stable.

Try that before posting here, a really high manual fan. Also, furmark is worthless - every driver from both nv and AMD tag it as a program to throttle. Do not use it.
 
I managed to get up to 1575mhz for memory.

And yes, I have a custom fan profile going. It was spinning well past 50% on load. Temps never got beyond 80C. I'll try a manual fan speed later today.

The vrms might be the problem here, since I get no artifacts, just straight driver crash. But then again, I get the same results even if I use stock voltages.
 
Furmark is good for thermal limit testing, but that doesn't mean it does a good job of testing as many critical paths on the die as possible. As others have pointed out: modern games are much better at that, having DX11 etc. which, AFAIK, Furmark doesn't exercise at all.

Also: there's a reason AMD set the standard clocks at 950Mhz. There is a lot of variability in the 28nm process. Very fast dies, not so fast dies. I'm sure AMD set the limit such that everyone will see some overclockability (as you do). I'm also sure that samples sent to review website are mysteriously not the slowest parts of the batch.
 
Another odd thing that I've noticed, is that VDDC in gpuz goes DOWN when a load is applied. What exactly is it measuring?
 
Last edited:
VDDC is voltage to the core. They follow a table in the bios for a certain state at a certain power level, pre-defined according to TDP to match a clock state. This power level is called a VID, and is set from the factory in the bios according to chip binning (aka ASIC level measured in gpuz). VDDC going down just means that gpuz is measuring the actual voltage output from the VRMs to the core. What you set in Afterburner or Trixx is a VID. The loss you experience is "Vdroop".

VCore: The core supply voltage of an 'important' chip like your CPU or GPU, usually not the Northbridge. Most frequently used to indicate CPU voltage.
VDD: The supply voltage to your Northbridge chip or the supply voltage for the input buffers and core logic of your memory chips (mostly on graphic cards).
VDDQ: The supply voltage to the output buffers of a memory chip.
VTT: Tracking Termination Voltage. Compared to VREF to determine Hi/Lo
VMem: Supply voltage to a memory chip.
VDDR, VDimm: Supply voltage to the memory on your motherboard.
VRef: Reference voltage for the input lines of a chip that determines the voltage level at which the threshold between a logical 1 and a logical 0 occurs. Usually 1/2 VDDQ.
VGPU: The supply voltage to your graphic card's processor.

Terms used by ATI internally:

VDDC: GPU Voltage
MVDDC: Memory Core Logic Voltage
MVDDQ: Memory Voltage supplied to the output buffers of the video card memory.
VTT: Termination Tracking voltage for video card memory.

Electrical Engineering Information

Positive voltages:
Vcc- Positive supply voltage of a Bipolar Junction Transistor.
Vdd- Positive supply voltage of A Field Effect Transistor

Negative voltages/ground:
Vee- Negative supply voltage of a Bipolar Junction Transistor.
Vss- Negative supply voltage of A Field Effect Transistor.

The letters c,d,e and s originated from the name of the legs of the transistors Collector, Drain, Emitter and Source.

The absolute distinctions between these common supply terms has since been blurred by the interchangeable application of TTL and CMOS logic families. Most CMOS (74HC / AC, etc.) IC data sheets now use Vcc and Gnd to designate the positive and negative supply pins.

The doubled suffix indicates that the voltage is "common", i.e. it is the supply voltage to one or more collectors (in the case of cc) and not just the voltage at a specific collector. Similarily, Vee is a common voltage for all emitters etc.
 
Yeah droop is normal.

overclocking a video card is like overclocking a motherboard + cpu + memory combination. (just a smaller pcb with less features)
 
update: i've tried setting the fan speed to 75%. amazingly enough, the card is now apparently stable with 1150mhz core and 1.274V. it completed a full run of heaven v2.5 with no crashes, with temps maxing out at 67C. before, it would crash within minutes. i'm not sure what changed here, whether the increased fan speed allowed the VRMs to stay cool so it becomes stable, or the increased GPU cooling that's doing it.

unfortunately, 75% fan is screaming loud, making it unpleasant to actually use the computer.

next, i'm going to try stock voltages at the same frequency at 75% fan, then i'm going to drop the 75% fan. hopefully this will allow to isolate exactly what's wrong.
 
moriz said:
update: i've tried setting the fan speed to 75%. amazingly enough, the card is now apparently stable with 1150mhz core and 1.274V. it completed a full run of heaven v2.5 with no crashes, with temps maxing out at 67C. before, it would crash within minutes. i'm not sure what changed here, whether the increased fan speed allowed the VRMs to stay cool so it becomes stable, or the increased GPU cooling that's doing it.

unfortunately, 75% fan is screaming loud, making it unpleasant to actually use the computer.

next, i'm going to try stock voltages at the same frequency at 75% fan, then i'm going to drop the 75% fan. hopefully this will allow to isolate exactly what's wrong.
Click to expand...

I plan on adding an Artic Cooling 7970 3 fan cooler to my card as soon as possible. You can cut a large amount of heat out with it, and it's near silent. These cards love to clock high when you can dissipate the heat.
 
OCGuy said:
You consider that OC "woes"?

Man people are spoiled....
Click to expand...
That's unnecessary.
moriz said:
update: i've tried setting the fan speed to 75%. amazingly enough, the card is now apparently stable with 1150mhz core and 1.274V. it completed a full run of heaven v2.5 with no crashes, with temps maxing out at 67C. before, it would crash within minutes. i'm not sure what changed here, whether the increased fan speed allowed the VRMs to stay cool so it becomes stable, or the increased GPU cooling that's doing it.

unfortunately, 75% fan is screaming loud, making it unpleasant to actually use the computer.

next, i'm going to try stock voltages at the same frequency at 75% fan, then i'm going to drop the 75% fan. hopefully this will allow to isolate exactly what's wrong.
Click to expand...
It sounds like you have a very low-leakage chip, in which case cold is going to be your friend (but not necessarily voltage). What's your GPU's stock voltage?
 
Stock voltage is 1.174v. ASIC score is 74.1%, so it's not particularly low leakage. It's pretty ordinary, actually.

Didn't get too much chance to test last night, but the card is not stable at 1150mhz with stock volts and stock fan speed.
 
MathMan said:
Furmark is good for thermal limit testing, but that doesn't mean it does a good job of testing as many critical paths on the die as possible. As others have pointed out: modern games are much better at that, having DX11 etc. which, AFAIK, Furmark doesn't exercise at all.
Click to expand...

This. My 7970 was stable at 1125/1575 @ 1.075V in Furmark after 30 minutes, but would crash in games after 10 minutes. In the end, it's all about stability in games anyway, so using that as a medium for testing makes sense.
 
Update: just tried 75% fan, 1150mhz core, stock voltages. Heaven crashed, but lasted all the way to scene 23/26. I'm going to do another run with additional voltages later, but it looks as if it is a combination of both GPU temps and VRM temps then...
 
moriz said:
Stock voltage is 1.174v. ASIC score is 74.1%, so it's not particularly low leakage. It's pretty ordinary, actually.

Didn't get too much chance to test last night, but the card is not stable at 1150mhz with stock volts and stock fan speed.
Click to expand...
moriz said:
Update: just tried 75% fan, 1150mhz core, stock voltages. Heaven crashed, but lasted all the way to scene 23/26. I'm going to do another run with additional voltages later, but it looks as if it is a combination of both GPU temps and VRM temps then...
Click to expand...
Very odd. 1.275V shouldn't trip the OCP, so I don't think that's the issue. Are you keeping your vRAM at stock while overclocking the core alone?
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top