7970M performance

While it is no GTX 780M or 980M, it still isn't too terrible. Latest drivers + overlocked GPU and CPU
http://www.3dmark.com/3dm11/9470470

Still waiting for a worthy AMD successor, or a windows 7 +legacy bios fix for the 980M

UPDATE:
Higher score:
http://www.3dmark.com/3dm11/9795282

 

Nice scores

 

I agree... Extremely good score. Far more than I was able to achieve with my 7970m's.

 

My 7970M could do 975/1475... Any higher and I used to get shutdown..Probably needed 330W PSU... That is a good score though

 

That is exceptional for 7970M CF. Your Physics score is really low for the CPU clock speed. That should fall in about 2,000 points higher (low to mid 11K) at 4.4GHz. You might have a power throttling issue. If you use OSD and watch, are those clocks dropping in the Physics test? Bumping up your Pri Plane and long/short duration turbo power settings might make a huge difference in the CPU performance. If you can sort that out I think you would be well over 14K on the overall score.

 

This was my score with an unlocked 3720QM @ 4.3GHz (was not a good overclocker) you should be able to beat that easily once your cpu power is sorted
http://www.3dmark.com/3dm11/8232268

 

Gents, I haven't got a clue. I thought the physics score was low. Took another benchmark, pri plane maxed out
http://www.3dmark.com/3dm11/9486615
Graphics score is nice, higher than the last now at 15519 instead of 15512, but the physics score is LOWER?

Unfortunately, these 7970Ms (nor the first one in my M17x R4) were particularly good overclockers. Most stable is 950/1300 at stock voltage. The R9 M290x overclocks much better though, I've been able to get to 1050Mhz on stock voltage.

Out of curiosity fox, what settings do you run your 3920XM? I'm not sure what flex VID works best for me.
C-states are disabled, XTU shows a constant 4.40 Ghz flatline.

 

I would just bump up the flex a bit and see if you get a performance increase, every CPU is different so you will have to find out what works best for yours.
Make sure the MSR lock in the bios is disabled, while your there you might as well crank up the power settings in the bios as well just to be sure

 

I'm not sure what is going on. MSR lock disabled, pri plane primary and secondary maxed out, throttle set to 90C, C-states disabled. XTU stress test shows that 4.40Ghz at flex 16 is stable, the clockspeed never dips during the duration of the test.

 

A number of things can come into play. Even certain Windows Updates can have an adverse effect on CPU performance. BIOS settings or running background services are the most likely culprit. Maybe your settings are too "hot" with Primary Plane maxed out. Secondary Plane is for Intel HD Graphics, so that can be left at zero.

My approach is to set my BIOS for my max stable overclock, or the next step down from max stable, then use ThrottleStop to create three additional profiles for lower clock speeds. I fine tune the lower clocks with XTU to find the sweet spot, then mirror those settings in my ThrottleStop profiles for one-click (or keyboard combo) clock changes. Below are my BIOS settings as well as what I use for 4.3GHz. Using ThrottleStop for settings lower the the BIOS values works great. Trying to use ThrottleStop to establish settings higher than what is set in the BIOS doesn't work nearly as well.

 

The thing is whatever I tr, my 3940XM doesn't hold 4.3GHz steady even when I set everything properly in throttlestop.. I get such a low Physics score too hence...

 

Is it hitting thermal or power limits when it drops?

 

Well power limit is 67W but it sometimes drops after 61W... Does the unlocked BIOS have the option of disabling this limit?

 

What?!? Seriously, you're still using a stock BIOS? You'll never get full performance using a stock BIOS. There are too many settings that need to be tweaked. Why on earth would you be using a stock BIOS when you have the option of full system access? Some Alienware 17 and 18 owners would kill to have it, LOL.

 

Yes you need to increase the TDP limits up nice an high to avoid hitting the 67W/57W turbo limits.

 

I have got my 3920xm up and running and I am having the exact same problem as Raidriar. Low physics scores on 3Dmark11. I'm at my wits end, my crummy unlocked 3270Xm 4.3GHz scores better then my 3920xm @ 4.7GHz. It just will not go over 90% total cpu usage. The weird thing is that Vantage and 3Dmark06 and anything else that stresses the cpu can get 100% utilization but not 3DMark11!!!

 

Same deal, my 3720QM outperforms the 3920XM, and I have no idea why.

 

Are you getting good results in vantage?

 

I don't have vantage installed, but I cannot for the life of me understand why the physics score is so low in 3dmark11. I checked XTU, it was 4.2Ghz constant the whole time, with a max temp around 78 degC. I tried with DPTF on and off, no difference. No background processes running.

 

I've got the same bloody problem with my 3940XM.. it should be fine in benches but Physics score is too low...

 

It could be a benchmark quirk, 3dmark 11 is pretty old now, what about 3dmark firestrike?

 

Even Firestrike has this problem.. I don't know but does RAM speed affect the Physics score in 3D mark?

 

Yeh I had the same in Firestike too, Vantage works flawlessly though.
The only issues I have seen regarding ram and physics score was ensuring that you are running in dual/quad channel

Edit: I did some research, poor optimization of the task scheduler with the physics tests has been acknowledged by futuremark staff in the following threads (may or may not apply to windows 7)
Thread 1 Thread 2
These articles go though registry/control panel settings to optimize background or program performance, I'll be trying when I get home tonight:
Control panel Method
Registry method

 

I will try with vantage at some point. Must be some sort of strange bug, I have 4x4gb for 16GB ram.

 

This is my vantage score, not sure if it is ok or not:
http://www.3dmark.com/3dmv/5245844

 

Yeh thats on par with my scores

 

Hello guys, i would like to know if my physics score is normal for my 3630qm overclocked with the Bus Clock ? http://www.3dmark.com/3dm11/9594383

 

It's lower than my stock 3630QM (~7600).

You overclocked using BCLK? You do know that disables Turbo Boost, right? Your CPU is actually running slower than before, unless your BCLK is at least 135 MHz to compensate for the loss of 800 MHz from Turbo Boost.

 

Huuhh seriously??!! I didnt know that it disables turbo boost... Alright thanks man, i will try without overclocking to see what it gives.

 

Yeah, Turbo Boost only works on default BCLK

 

Without overclocking im getting worse performances... (Physics Score : 6600).I dont really know why its so low... Ive looked back to my old 3dmark11 scores, and ive noticed that i had 7800+ physics score...(about 9 months ago). What can cause this problem ?

 

log your your run with HWiNFO make sure your all the multipliers are activated during the physics test, you will also be able to temperature and anything else that may be out of place.
Your gpu score is phenomenal though!! how much extra voltage are you running on your cards?

I have checked my Firestike results with what others are getting and they seem to be ok, even though the cpu is never 100% utilized.... must be a windows 8 issue then
Physics and Combined results are still horribly pathetic in 3DMark11 though

 

Have you switched from Windows 7 to Windows 8?

 

Even in Windows 8 I was getting 1000 points higher than you in 3DMark 11 Physics with 3630QM. Your TB must be disabled or CPU is throttling some other way.

 

I just made it with the hwinfo log, during the physics test, the cpu runs at 3.2GHz, 80% usage, no throttling, max cpu temp 80 celcius degrees...
For the gpus, they were running at 1.1v ( asus gpu tweak overvolt)

First i had Windows 8, and i think it worked fine ( i had around 7900 physics score), now im running on Windows 8.1...

But ive noticed something, in my old 3dmark11 scores, my max turbo core clock were 3.43GHz now its 3.19GHz... maybe you 're right on the turbo boost issues

 

New result:
http://www.3dmark.com/3dm11/9795282
Still find that the physics score is very low. Graphics score is on point.

3Dmark13 result, currently the highest for 7970M setup, but i'm not sure why. I only have 925mhz and undervolted to 1.025, I'm sure other guys can break the score. I'm just not ready to risk my 7970M for an overvolt and overclock
http://www.3dmark.com/3dm/6925473

 

For only 4.2GHz that Physics score looks just fine to me. Under Windows 8.X it would be about that even at 4.5GHz.

 

Oh that is good to know, never mind then!

On a separate note, I feel that my CLU application on my 3920XM has worn out or something has gone wrong. I woke one day to find idle temp in the 60s and load temps in the upper 90s, but when I pulled the heatsink, the stuff was still liquid on both the die and the die contact area on the heatsink. I have no explanation as to why temps jumped for no reason. Even a reapplication (90% alcohol on die and heatsink + sanding heatsink) has not really remedied temps. With no extra flex and 42 x 4, temps reach 75C routinely. and sometimes even 80. Is this normal?

 

Nope, I have not seen that. My CLU is still working marvelously on the M18xR1/R2 and 18. If it is running too hot with CLU, chances are it's going to run even hotter with something else. It stayed soft and liquid-like for me for over a year. The only part that gets hardened is where it is exposed to air, and that does not hurt anything.

Something had to change to cause the temps to be different, but it does not sound like it is attributable to CLU. It may be an air flow issue. Is the machine elevated so it can breathe in an unrestricted supply of cool air? Maybe the fans are not working right. What are the ambient temps in the room where you are using it? Are the load temps that high even if the fans are forced to run full blast?

 

It could be the heat pipe has developed an issue.

 

More to add to the mix, throttlestop is reporting temperatures that are fine, typically around 65-68 degC at 4.3ghz x 4 on full load at 0 flex where XTU is reporting 75-80 degC. It is almost always a 10 degree difference between the two reporting.

 

I don't use XTU for that measurement and it may not be accurate. XTU is actually designed more for desktop motherboards/chipsets, so it is possible the sensor readings are off. Try HWiNFO64 and see if the thermal reporting is in line with what ThrottleStop is reporting.

 

HWinfo is giving readings 5 degrees in between xtu and throttlestop. gahhhhh I don't know who to trust?!

 

How strange. Maybe the polling rate has something to do with the variations (being slightly off in real time with a lower refresh cycle). Are the max temps off by 5°C also? If so, I've not seen that kind of variance between ThrottleStop and HWiNFO64 sensor data. They are usually the same, or extremely close.

 

It might be differences in how the Tjunc temp is being set.

 

I have seen such varrience between throttlestop and XTU/HWinfo64... XTU and HWinfo64 are similar but Throttlestop is like 10C less even when max load is on for a while where temps are evidently higher..

 

Meaker is right, if the TJ max is not set correctly in the throttlestop ini it will report incorrect temperatures.

 

umgh how do you set it though?

 

Open the Throttlestop.ini, the TJMax setting should be the fifth line down, I believe the TJMax is 105C for Ivy-bridge but I would check Intels specification on your specific cpu just to be sure.

 

It may not matter which line it is on. ThrottleStop 8.00 doesn't even have the TJMax line in the INI file on my systems. Maybe @unclewebb decided to eliminate it. It was the first line in the INI file for me with ThrottleStop 7.00.