Just got my 880M twins!

Just download Hwinfo 64... Install it and click on the Sensors tab.. Reset the sensors before you start the game by clicking on the clock symbol and after clicking on the GPU's core clock reading, a plot will come up.. let HWinfo 64 in the background while you run the game.. after the session, take a screenshot of the plot..

Otherwise you can use MSI Kombuster and all the 3 clocks will be recorded...

 

You only need GPU Clock?

 

You can clearly see there is no taxing by the max voltage being 0.887v

At that voltage my cards run so cool that I don't even need the fans (temps stay under 87C with both fans on their first step). Card = not taxed = low voltage = no heat.

Sent from my HTC One_M8 using Tapatalk

 

I can't view deleted posts outside of the Alienware forums, so I can't check for you. (I have no moderator rights except in Alienware forums.) My guess would be something in your posts is not in compliance with the rules. You are required to use a specific form for the Marketplace. If that format is not followed to a tee they will delete it.

 

Checking my voltage table list from what I was able to extrapolate watching on screen display with voltage and core, 0.887v isn't on the list but the max stable I got at 0.875v was 836MHz core so it's somewhere around there.



Sent from my HTC One_M8 using Tapatalk

 

Each step is 13 MHz and 12.5 mV, so 0.887 V would get you almost exactly 850 MHz (849 MHz to be exact). Which is stock 780M clock, no?

 

Unless he's running a custom vbios, yes.

Sent from my HTC One_M8 using Tapatalk

Didn't realize non boost was still considered a boost bin.

Sent from my HTC One_M8 using Tapatalk

 

Look at this photo.

 

That photo doesn't really tell you enough. The "average" column doesn't provide any kind of meaningful insight unless you are dealing with static values. It includes calculations of no-load and power saving scenarios, so it's really utterly worthless. You can remove that column because all it is good for is creating confusion. The table only shows current, lowest and highest values. The high and low are important, but what you really need to know is what is happening in between those points.

You need to double-click the lines in the table to open a graph and then watch what the clock speeds do under load, or set up OSD to watch in real-time on-screen. Using OSD is ideal. When your FPS drop you can see in real time what happened with core/memory/voltage, etc. You can also see whether the GPUs are running the clock speeds you want them to, or doing their own thing "dynamically" adjusting (which is not a good thing) to load demand.

 

I'm telling you they're running 850 core at the most and if they're not I'd appreciate a vbios dump.

Grab 3DMark 13 (just called 3DMark now) and reset your he info before starting the test and run firestrike extreme... Then post average clocks...

Sent from my HTC One_M8 using Tapatalk

 

Yeah, my bad. I meant steps. Shouldn't have said Boost.

On a Kepler card with GPU Boost enabled, core clock and voltage moves in discrete 13 MHz and 12.5 mV steps from base clock and voltage. Both up and down. If, when adjusting core clock speed, you set a non multiple of 13 MHz offset value, it automatically rounds up or down to the nearest valid number.

With a modded vBIOS that removes GPU Boost, you can set any core clock speed you want.

 

Write what and how can I turn, step by step instructions. Tomorrow I'll do the tests.

 

Your games aren't taxing the card enough.

Anyway if you go to the Alienware forum and search OSD or RTSS you'll find his glorious tutorial on how to see everything in real time.

Sent from my HTC One_M8 using Tapatalk

 

Here's a look.



Everything I have works very well and turbo boost, too.

 

(Cheers Mr Fox, didn't realize you could do that in HWInfo (the graphs), I've been using other means of measuring, but that's a very useful option too!)

 

Looks good, but you should ideally (to put it into context) tell us how long that test was run for, what the test was, along with doing a graph of the GPU clocks & GPU load (in addition to the graphs you've already provided).

EDIT: Didn't realise you'd actually provided graphs of clocks, they're not clear though, you would need to change the y-axis so that it is easy to understand what the actual value of the clocks are. You'd still want to provide graphs of GPU load too though, and answer those other questions above too.

 

Yeah, they want me to remove my GPU's and take a picture of it with my nickname on it. Stupid if you ask me...

 

Just saying, I'd have no confidence buying anything from you if all you do is whine whine whine and can't even follow basic marketplace rules which are there for a reason.

 

If that flat-line is a long period of gameplay, that looks outstanding for 880M. Maybe post a screen shot of your GPU-Z tab showing the vBIOS version and if folks like Ethrem and DumbDumb that are having fits with 880M have a different vBIOS you could dump yours with GPU-Z, zip it and attach it so others can download it.

 

Here is a new graphics after playing 2 hours.

 

That looks perfect. I bet some 880M users that are having bad experiences would appreciate it if you share your vBIOS with them. If yours is the same and flashing it does not fix their GPU malfunction problems, then there must be a hardware difference of some kind. Looking at your graphs, it would appear NVIDIA may have fixed whatever was wrong with 880M.

 

That looks good, but you're still not providing all the info:

1) Change the value of the vertical y-axis of the Graphs down from 3000 to something sensible like 1000 (considering we're expecting your core clocks to be in the region of 850-993Mhz) - that way we can see more clearly what your clocks are.

2) What game were you playing?

3) Provide a graph showing GPU load/usage as well as the ones you have just provided.

4) Job Done!

[5) You could do what Mr Fox suggests re letting us know your vBIOS version & uploading that too.]

 

Whine and whine and whine? Wow, someone having a bad day..

 

I'm having a great day, thank you very much . Just making the observation that in the time spent complaining, you could've already finished listing the item correctly and been done with it.

 

Okay. While this is all well and good, I still need you to really stress the card. Call of Duty: Advanced Warfare and 3DMark 06 are not going to do it; not even close. Hell I get 60 degrees in SLI with that game on max graphics; on Single GPU I got about 72 like you did, except I was also recording via Shadowplay. I need to see how it works in a game like Far Cry 3, or BF4. Both games must be set to ultra (at 1080p, and yes this includes anti aliasing) and have the FPS unlocked, and BF4 needs to have its resolution scale set to 150% or higher. That way your card should begin a nice heating process, and we'd really find out if you hit your thermal barrier or encounter downclocking or not.

 

My day is good too! Anyway, Im not home at the moment so thats why I cant remove the GPU's in order to meet the rules.

Anyway, 880M's are still for sale...

 

Ill say this.. my 880s run games ..how well i dont know.. i got the stupid horizontal line tearing that shows up in most of them..
now how do my 880s compare to my 780s? they dont.. and if what i see is true then you got a diamond in the ruff and should throw down a stock bench run on vantage or some thing to see if they do like every one elses and go to crap..

i wanna see your vbios.. please do as mr fox has stated and upload em.. thanks..

p.s. ive grown over the years of being here and tbh it makes me WANT TO SCREAM FOUL AND BS looking at your screens.. how can 1 person have a perfect working set of 880s while millions of other dont?
with that said im not going to.. but please run a vantage bench and upload your vbios thanks..

 

here is his gpu-z vs mine.. it is different.. i need to see that vbios sir..

Gyazo - bcff29a81b82e8d6bdeac56944fec981.gif

 

Yours isn't stock either though... the default clock should be 954, but yours is 993 constant.

 

NOT THE POINT LOL.. his card is performing a thousand times better than mine and every one elses according to his graphs.. also the vbios version is different from the 4 stock vbioses i have.. its one i havent seen,,

 

Well yeah, I understand, but that's why I keep telling him to use it in a hot game which he refuses to do. Showing us Advanced Warfare and 3DMark 06 is not going to help anyone. Until he shows us in a game like the ones I keep calling at the settings I keep calling and shows us that it won't touch 90 degrees or cap at 92 & begin throttling, then it's kinda pointless. My cards don't hit 90s in BF4 at 4K res with ultra preset, but they sure as hell pass 80s. I told him 150% because it's 1 card vs 2 and I was indeed recording at the time, but my cards were 99% load and got quite hot.

So in essence, I want him to push the card properly and see if it remains cool. If it does, then fantastic! We get a vBIOS dumb and everyone stuck with 880Ms are happy and all is right as rain, then maybe johnksss & svl7 can use that vBIOS as a base and release an updated modded vBIOS for the chips.

 

Who actually puts the vbios on the cards to begin with. Nvidia ships them with a universal vbios and each company loads there own or nvidia puts it on there for them.

 

If somehow the vbios got fixed for the 880M that would be sweet for everyone who owns the cards now. Just have to get the person with new card to upload vbios for everyone else.

 

Here is the new graphics, after 3 hours of playing BF4 settings on Ultra. I think more evidence is not necessary. Thank you.

 

Excellent, very good results, thanks for shrinking the y-axis on the core clock graph so we can see clearly the clock. A very stable core clock at boost, as well as a high average boost clock according to the HWInfo table. If that was run for 3 hours of BF4 at ultra, I think that's good evidence, that would be 100% GPU load for most of the time right? Would you be able to upload the vBIOS so that others can flash it?

EDIT: took a closer look at the core clock graph: it's not locked at 993Mhz the whole time (hardly ever at 993Mhz actually), as there are only 9 grid demarcating segments for the y-axis as opposed to 10 demarcations, so it looks superficially like a higher core clock. In reality the 9th demarcating segment of the y-axis is 888Mhz, so you were always above 888Mhz in that graph, which tallies with your 899Mhz average value in your table. So, in summary it's not locked at 993Mhz, but a lot better than most people's 880M right. Would still be useful to upload your vBIOS I reckon. Good stuff!

 

View attachment GK104.zip
vBios.

 

LOL look at that fluctuating core clock. Much ado about nothing.

 

Don't like don't look. My card works perfectly and I am satisfied 100%.

 

How does your 3DMark Sky Diver score look compared to this? Result

Well, considering those fluctuations are over a 3 hour period it seems like an improvement. If you are running a stock vBIOS there is going to be some fluctuation on any Kepler GPU because of NVIDIA's retarded "adaptive" performance and dynamic boost pü-pü-ca-ca. Compare that to only 26 minutes on this core clock graph and I think it could be an improvement in 880M misbehavior.

Speaking of "adaptive" are you telling your GPUs how to act with NVIDIA Control Panel or allowing them to tell you how it's gonna happen? If the tail has been wagging the dog, you might get better results by tweaking everything for performance versus efficiency and sparkly eye-candy fluff. Also use EVGA Precision X, Afterburner or NVIDIA Inspector to prioritize power over temps and max out both sliders (at least 130% on power target) so they GPU isn't allowed to make any decisions. If you use NVIDIA Inspector, be sure to set each GPU individually, since a sync option does not exist.

 

Denial is a long river.

 

The REAL benefit here is that his card didn't cross 75 degrees and vsync was off. I don't know if he used 150% resolution scale to heat it up as I asked him, but even so, that seems to be better than many users' cards. Most peoples' 880Ms would hit 92 instantly.

Still not a perfect vBIOS though, but maybe a step in the right direction. If he doesn't heat up and putting the vBIOS in another user's machine heats it up though, then he has a good CARD and not just a halfway decent BIOS.

 

It not passing 75 means he used some extreme cooling solution. One of three things happened. Either his system is not set up properly and thus was not taxing the cards, he has a different vbios, or he used a more extreme cooling solution like CLU and an A/C unit..

Sent from my HTC One_M8 using Tapatalk

 

Well I play BF4 with my chips in the 60s, so 74-75 degrees for a 880M sounds fairly "par for the course", which is why I specifically asked him for 150% scaling or above, but he didn't say if he used it.

Either way, the vBIOS is there for downloading, so feel free to try it out Ethrem. If you don't heat up (and he swears he's using stock cooling too) then great. If you DO heat up, then it means he either has an extreme cooling situation and is not telling us, or his card is "good" and most others are bad, I.E. he got one of the best samples.

 

Well that's what I'm saying... BF 4 is not demanding enough. Fire up Bioshock Infinite on max settings watch the temps skyrocket. There's no way nVidia started pulling the chips from a different bin because the mobile chips can't reach the frequencies their 680 desktop leftovers can. At least not without a voltage bump.

Sent from my HTC One_M8 using Tapatalk

 

mitya_alba's vBIOS (vBIOS 80.04.f5.00.07) is the same vBIOS version I reported 2 weeks ago.

At that time, Mr. Fox asked for my vBIOS (vBIOS 80.04.f5.00.07), when I reported a stable "recently deployed" 17r1-880, 4910 running benches and keeping max temp at 73C. See Mr. Fox's and my posts below.

mitya_alba's replacement 880 cards have the new vBIOS.
So (apparently) do recent 17r1-880 systems coming from Alienware (like mine).

Posts from thread
http://forum.notebookreview.com/ali...are-18-880m-sli-4910-oc-d-4-10-fps-issue.html

 

Hmm well a silent fix is better than nothing I suppose. And I wouldn't flash an Alienware vBIOS on a Clevo card lol no, but if you really wanted to give it a go flash one card first and see if the computer boots with that one card only as the master. This way at least you can flash it back should the vBIOS brick your card.

 

Alienware vbios works fine on a Clevo card, what do you think was used to make the mod?

It sounds like nVidia fixed the hardware, maybe I should have Sager swap my pair and see... Really hate to be taking a risk on 880s again though.

Sent from my HTC One_M8 using Tapatalk

 

Interesting, always thought the incompatibility between Alienware cards and Clevo machines was due to the vBIOS. Or maybe there's some hardware difference as well.

 

The difference used to be that Clevo didn't have true MXM support until the SM and SM-A from what I understand but beware if you have hybrid boot or have it on BIOS only, you can soft brick with the newer ECs so I always have to flash in Windows and reboot to avoid that. For some reason the mod doesn't like Non-UEFI boot.

I'm over fussing with these cards though. I doubt the vbios will fix anything... If anything, they increased "efficiency" ala Maxwell and just made clocks even more unstable to keep heat down which would mean that the average core clock would actually be lower than what I have now.

Also keep in mind SLI machines were the throttle monsters...

Sent from my HTC One_M8 using Tapatalk

 

His good temperatures could just be the heatsink he has & how well it's been pasted - there's variance inbetween units of the same model even - if that's what you meant by him having a good "card" in relation to his temperatures. (Because the operating voltage vs Mhz is stipulated in the vBIOS right, or does the same vBIOS change the voltage vs Mhz stepping automatically based on the ASIC quality for that individual unit?)