GTX 980m Limited by CPU....

4 different processors with 980m

NVIDIA GeForce GTX 980M video card benchmark result - Intel Core i7-4710HQ,ASUSTeK COMPUTER INC. G751JY

NVIDIA GeForce GTX 980M video card benchmark result - Intel Core i7-4790S Processor,Notebook P770ZM

NVIDIA GeForce GTX 980M video card benchmark result - Intel Core i7-4860HQ Processor,ASUSTeK COMPUTER INC. G751JY

NVIDIA GeForce GTX 980M video card benchmark result - Intel Core i7-4930MX,Notebook P15SM-A/SM1-A

 

and Sky Diver

NVIDIA GeForce GTX 980M video card benchmark result - Intel Core i7-4710HQ,ASUSTeK COMPUTER INC. G751JY

NVIDIA GeForce GTX 980M video card benchmark result - Intel Core i7-4790S Processor,Notebook P770ZM

NVIDIA GeForce GTX 980M video card benchmark result - Intel Core i7-4860HQ Processor,ASUSTeK COMPUTER INC. G751JY

 

If you do not run out of power OR hit the heat limit, you will stay (under load) at the maximum 4-core boost clock you own. Throttlestop will force it to run at the max speed whenever you want it to, even if a game isn't using enough of it to run at the max 4-core boost clock.

If using 2 cores or 1 core, it uses the 2-core and 1-core boost clock which is usually higher than the 4-core boost clock. For sandy bridge and above, the general rule of thumb is that the boost clock is +800MHz over the base clock for 3 and 4 cores.

 

massixline, something like 1fps is like +400 points or so on the score tables... so that's pretty much also negligible. Though interesting that the 4930MX is the lowest scoring... that CPU or GPU is definitely throttling somehow; it's the strongest CPU out of the bunch. So yeah, not much of a FPS bonus there, right?

 

You should be able to raise the TDP limit with XTU. I can get my 4900MQ to do 4GHz without throttling if I set the appropriate current and turbo power limits.

 

Now its time for me to say:
I told you so...

Thanks for the headsup. I had a feeling that 680M SLI power would somewhat be too much for this HQ CPU. Anandtechs 980M review show the same and the GPU is missing maybe 10% of what it should have performed.
Will absolutely try to get the IPS SKU with 4980HQ or any other notebook with similar CPU

 

Are the 4900mq qs you see on eBay going for 430 dollars going to work well in my 18? Will it basically be a plug and play type thing?

Sent from my Nexus 5 using Tapatalk

 

I think the lower score on the MX is likely due to Clevo... Their machines need modded BIOS to have PCIe 2.0 which probably matters even more Maxwell than it did for Kepler.

 

what are your current and turbo power limits on that cpu? i have the same one and just got the unlocked bios. i dont know which bars to move and what numbers

 

Even after setting the TDP limits to 100W, Turbo Boost Time Window to max 128 seconds, and current limit to 80A, it still had TDP throttle regardless.

 

Did you unlock turbo power limit MSR? That way you can keep it at the short duration boost indefinitely, otherwise it'll fall back down after 128 seconds.

Probably need an unlocked BIOS for this.

 

It won't even last 128 seconds, only about 30 regardless of what I do. That's only with CPU at 100% with stress test. With gaming it tends to stick near 3.5GHz but it isn't always at 100% load.

 

Yeah without the MSR lock disabled, adjusting the turbo time limits might not do anything.

 

Seems like you might need to disable the package turbo power lock if you have that option in XTU.

I poked around in my unlocked BIOS and the closest thing I could find is a "CFG lock" option. It was set to enabled but my 4900MQ doesn't TDP throttle so that's likely not it. Maybe Prema's unlocked Clevo BIOS comes with the TDP lock disabled by default.. *shrugs*

Btw just did a TS 1024M run and my 4900MQ didn't throttle even with the turbo boost time set to 28s as shown below.

See below, but I should add that each chip is different, so there's no guarantee what worked for me will work for you. Also, the power and current limits can be set arbitrarily high since they're just that, limits. The CPU will only take as much as it needs. Voltage however is a very different story, and I strongly recommend you do some reading before messing with it.

 

LOL I think we're talking about the same thing, just called by different names depending on the BIOS/XTU.

 

But whatever that "CFG lock" was in the BIOS, it was set to Enabled, yet my 4900MQ still didn't TDP throttle, so that can't be it.

 

This sort of stuff makes me mad. Why can't laptops have good-looking and functional UEFI interfaces with helpful tooltips.

 

Lowest common denominator...

 

I honestly think anybody who has a gaming laptop with MXM chips and socketed i7 CPUs should, by default, have things unlocked, or at least the option to pay like $5 extra for an unlocked BIOS. I don't really see the problem. Those who know what it is will gladly pay for it; those who do not or don't care for OCing will leave it be.

 

Would be interesting to know what registers are set.

Maybe because too many people don't understand what they are changing and it ends up in a support call because they can no longer boot their laptop.

 

^yep exactly

I remember you pulled your unlocked MSI BIOS because too many people had issues with turning the IGP off and not realizing it would black screen their monitor since their machines didn't have a MUX correct?

 

If somebody knows that little and is paying for a $1500-3000 laptop and doesn't even TRY to look for help before doing something, that's on them to be honest. Besides, I don't know exactly how one can change BIOS settings with regards to CPU overclocking (especially on CPUs with a maximum turbo boost) so that they can't make it boot. That has to be a special kind of tinkering.

 

Well unfortunately it happens and IMHO that kind of spoils it for the rest that have a bit more knowledge about such things.

Set the current limit / voltage too low and the laptop doesn't boot.

"What's a CMOS reset?", "Woot, I have to dismantle the laptop to do that!"

Anyway, back to the original throttling problem. On my own hardware the 4700MQ will PL2 core throttle at 47W (package at 50W) even if set to 100W / 100A so it's interesting this doesn't happen for n=1. Might or might not be able to see something if a registers dump was taken.

 

I don't get how the hell you get 4GHz @ 84C max... Makes me fuming mad that my 4940MX won't do the same. What's your chips's VID @ 4GHz?

CFG Lock is just used for Hackintosh so that it doesn't trigger a reboot as soon as the kernel boots. Its something Apple implemented with Haswell to make it even more of a pain for Hackintosh.

 

I'll check VID when I get home, but yeah please ignore the temp, I had my cooler fan blowing directly into it, and I'm using Liquid Ultra on top of that. If I turn off the cooler fan the temp runs all the way up to 91-93C range. Although if I forced max fans I could probably keep it below 90C without the cooler, will have to try and see.

I'd be happy to help if you could walk me through it. Alternatively I'm using the latest version of Prema's BIOS so that could also be a starting point.

I should probably add I used TS purely for demonstration, never had to actually use it since my CPU doesn't power throttle with Prema's BIOS.

 

If you would run the attached and zip back the text file that would be great.

 

Here are some more average FPS results of the 970M/980M compared to pretty much every competition card:

Spoiler

Credits to LinusTechTips. These results are taken from an ASUS G751 and Aorus X7 Pro.

The desktop cards were running with a 5960X. The laptop cards were running with a 4710HQ.

 

4 cores of the 5960X were disabled and it was downclocked to 3 GHz, so if anything, the laptop CPU's had the advantage with their higher 4-core turbos.

 

Done, see attached.

Let me add that I've observed a strange behavior with XTU that's been present since day one. Sometimes, it doesn't matter what settings I use, the processor would either refuse to go past stock turbo, or worse get stuck at 3.4GHz for no reason at all. Usually this can be solved by changing the offset voltage and core multi back and forth a few times, but when the settings just absolutely refuse to stick I have to restart the laptop for it to return to normal. Observed this again when I loaded the 4GHz profile from XTU, but thankfully fiddling with the volt and multis a couple times fixed this issue. Did a TS 1024M run just to be sure, and took a dump right after.

And Ethrem, hopefully the picture below answers all your questions:



This time I forced max fans with Fn+1, and didn't use the U3 cooler fan at all. Somehow this worked out even better lol, as temp slowly rose up to about 77-78C and then stayed put. Ambient temp was 22C (72F) btw. Should probably add I'm still using the modded bottom cover with holes cut out.

And apparently my 4900MQ is capable of 64 GHz lol

 

LOL SIXTY FOUR GIGAHERTZ?

Yeah n=1 I wanna trade CPUs pls. I want a 64GHz 4900MQ, I'll have bragging rights for the century!

 

Got nothing on my 110 GHz!

 

Yes it does... your chip does 3.9GHz @ 140mV less than mine. Your max VID was 1.072v there and you had a -50mV undervolt. Add them together and you get a stock VID of 1.122v vs my 1.262v

My chip with a -86mV undervolt is still pushing 1.177v VID @ 3.9 which is still 105mV more than your tweaked chip.

I'd have to see more stock VID figures from others but I'm guessing that 4900 and 4910 chips will be pretty similar which would confirm my suspicions that they are binned for lower voltage in order to meet their 47W TDP limit and thus are better candidates for high every day overclocks.

 

Imagine how funny it would've been if the power read 1.21x10^9 W....

Makes sense, and also corroborates what Mr. Fox had remarked very early on that "seems that some 4900MQ chips can actually outperform 4930MX at the same voltage, so it's a mystery why they weren't binned as 4930MX instead".

Oh and btw the 4900MQ was at 4GHz. Shows up as 3.99GHz because of the stupid spread spectrum which I forgot to disable.

 

Yeah mine is the same way.

Either way, despite the argument on the other thread, it does seem that the 49xxMQ chips are the better way to go as I suspected. I wonder if I can find a decent deal on one and if it would even be worth it. Since I can't use CLU on this machine, I still probably wouldn't like temps at 4GHz.

I had to repaste last night, my GC Extreme pumped out.



Sent from my HTC One_M8 using Tapatalk

 

I think you could probably still keep temps below 90C if you forced max fans. In my experience CLU makes a difference of 5-8C under load so if those temps are anything to go by and assuming you get a similarly efficient chip your load temps at 4GHz probably shouldn't break 90C.

 

Max fans are so helpful it's amazing. Those alone are a full 10+ deg drop in temps for me sometimes. Far less for you n=1 who has the extra airflow through your backplate

 

I force max fans when I'm doing something with the CPU overclocked, the system fan doesn't spin up fast enough and it doesn't hit a high enough point to keep the temps under control. They still pass into the 90s @ 3.9GHz though. My heatsink doesn't make enough contact for CLU either, I already tried it.

 

Imo, it's because very few of the people who do software tweaking at these companies know what the functions actually do.

The way boost and throttling is done on laptops in general tends to sustain that idea. It is possible, for example, to clock bus-speeds and ram timing separately with the xtu interface. And to have "sustained load" tweaks that favor maintaining normally high clock speeds on all cores (ensuring that none of them drop, and suddenly increasing thread response on occupied cores..). Or have lower base clocks and favor getting one core to boost more often (intel doesn't have a "tdp" target that decides whether throttling happens, but you can use the element temperatures and pick something that avoids raising the temperature, but maintaining the boost). But it's not done, and the majority of the laptops out there essentially run with something that resembles "safe" desktop tweak settings thanks to the recommendations from intel..

And you run into the situation you have here, where the processor avoids throttling until it hits the hard-limit at 105 degrees. Or else a cut-off set in the firmware package at for example 90 degrees, depending on load.. It's not strictly necessary to have that, but..

 

Thanks for that. Had a look, nothing unusual unfortunately.

 

Cool, good investigative work HTWingNut! The good thing is that it looks like the 4710HQ would be fine for 60Hz/60fps gaming, yet given you've seen restrictions on frame rate above the standard 60Hz standard on some games then I think I'd want a faster CPU if I had a 120Hz screen and a 980M. Good work!

 

Great investigative work. If you had a 980M SLI an 120Hz you'd need a i7-49xx to get the most from it.

 

I am getting a 980m SLI from eurocom since they still offer a 120hz screen. I'm torn on whether I should get the 4810mq or a 4900mq. The same configuration that costs $3k at a reseller like Xotic PC would cost me $3800 from eurocom. Luckily, they offered me price match but only for identical components. I can't seem to find another reseller that offers the i7 4900 mq for a cheaper price., and the upgrade from a 4810mq to 4900mq costs $350 at eurocom. Any suggestions?

 

when i overclock my cpu, my laptop randomly shuts off and restarts. no blue screen of death. no bad temps. just idling. idk why....

 

voltage...

 

I wonder if Clevo doesn't have a limit...

Mine is doing it with dual power adapters if I have 4.2GHz on my CPU and run SiSoft Sandra. It will start the test and then shut off.

 

Switch to the latest hwinfo64, so the maximum core values ​​are adjusted properly. Which you wrote (?) :thumbsup:

Edit: I meant Octiceps its Hwinfo64

 

Could you take a screenshot of your XTU settings?

Are you running a CPU only benchmark? That's just plain weird. I can confirm that with all turbo bins maxed out on my 4900MQ, it can consistently draw about 80-90W in CineBench without ill effect, and XTU even dips into the 110-120W territory but was still OK.

Incidentally, should mention I've seen my 4900MQ hit 122W peak draw when benching XTU. That's like maybe 12W less than my 4930K at 4.5GHz. So much for Haswell power efficiency rofl

 

Yeah I don't know what it is, it's only that bench. Sisoft is extremely stressful though, it freezes the whole computer until it's done with its CPU tests so maybe it's hitting thermal shutdown? No idea.

Either way, 4.2GHz as my daily driver is looking nice. Fans go max when I need to do something but it idles at 50C and the fan is basically off while stock you always hear the fans.

Sent from my HTC One_M8 using Tapatalk

 

I still think you should get that heatsink checked out. If nothing else another 200MHz couldn't hurt.

And yeah I used to use Sandra for stress testing like 10 years ago, and it did a pretty good job, although sometimes it was hard to tell if your computer locked up or it Sandra was just stressing it too much.

 

I just set the TDP to 70W to keep it from going into throttle territory but it runs 4-4.2GHz in just about every game I throw at it, its just the synthetics that it off.

And yeah the Mandel test in AIDA still throttles @ 105W lol