With new cards coming out...

Because, if I make them work in an Alienware... That may cause serious problems for you.

 

Ehhhh... There was obvious degradation. I'm not sure what the cause was but the graphs don't lie. When I received the replacement machine, the two cards were totally synced with each other between core and temp and now there is a very large disparity between them with the slave being up to 100MHz off in the core on the same game that I tested the machine with when I got it back. The two cards are nowhere near synced now and I haven't even opened up the machine since I got it back and it has sat on the same table in my room which is always kept between 20 and 21C. I baby my machine and don't let the cards run higher than low 80s and it still happened.

I'm sure with hard mods and your expertise, we could get a fresh pair working totally stable but very hot and there's no guarantee they would stay stable.

I'd rather just be done with it. 780M are proven solid, if I have to pay for it, so be it.

Sent from my HTC One_M8 using Tapatalk

 

Yea, no idea why but it seems like the SLI on the 880's is just totally fubar.

I've rarely heard issues on single 880M's, and they seem to be more or less 'in spec'. But SLI? Yea, no idea why so fubared.

Any chance they may be willing to up you to 980's?

 

You have to realize this....You and I are two completely different people. I have been doing this for a very long time. So I have a completely different out look on the "impossible" And what can or can not be done. And since I have already proven that one 880M can work perfectly fine. (Also with boost enabled). Now might be the the time to try two. Or else just wait for the 980M's to hit the street and just jump on them.

As to 780m's working....That was another projected we worked on and that's why they get the performance they are meant to today. If you were here at the start of that project, you would know they did not work! And everyone who got them was trying to get back on 680M's, till they finally started to work. This is nothing new. Although we hope with the next gen gpu, they skip past that not working right part at stock.

 

I think it's going to be fun to see how the 980Ms are going to work. I remember the good old days when I got a 280M in a D900F and I installed windows and drivers and the card just played everything at max and I was happy. Man, those were the days.

Not that I dislike what eventually came out of the 780M vBIOS modding (and of course, the other versions) but I'd like a non-gimped, fully-enabled, properly-working mobile GPU flagship release again please & thank you.

 

Its true that single card systems seem to have far less issues. Even disabling SLI on my machine the master starts throttling when it was the slave before. It boggles the mind.

As for an upgrade, I have no idea. I had already planned to upgrade to 9xx cards and try to work out a deal with xotic anyway, it just so happened that these cards decided to go kaput before that.

I do understand this and that's why I said with your expertise, I could probably get them working how they should but that it would just be easier to get the 780M.

I know about the drama over the 780M, I did my reading but they are fixed now and from what I read they still were nowhere near as broken as the 880M cards. These cards are a mess. I'm amazed I never ran into the "lets run like an integrated graphics card" issue.

Sent from my HTC One_M8 using Tapatalk

 

We shall see my friend. Right out the box, run a few test to see if it throttles at all. Right now im running 771/1250 boost 850/1250 clocks on a 780M. Perfectly stable and completely optimized for stock. So I will know exactly what your suppose to get. And I do have some of the same games you have as well I think.

And you never ran in to the "lets run like an integrated graphics card" because you were never really able to do anything other than stock. You would need to have a few hard crashes for this to kick in. And maybe a 334 and up to 340 driver. (speculation on that) Although I haven't seen it on recent drivers with the 780M's which is interesting and a blessing! But trust me. When it happens. The card will run like that on any machine it's put into until fixed. Which is a software/flash/bios rest type of fix.

At the end of the day though....lets just hope all is well this go round for you.

 

I meant the new pair, not this pair. An obvious degradation has occurred with this pair, its just a matter of time before the memory errors get worse and there is an all out failure.

We will see what happens with this, xotic is working to see what options there are and then when I've seen everything laid out, I will make my decision on what to do.

I've done more than my fair share of work to get the first pair of cards I had working and just wasn't willing to go through all that time and trouble with this pair since its a memory issue.

I would have kept my first pair if I didn't have one of the cards show up as revision FF no temp sensors or anything then randomly come back and start working again.

Maybe I just have bad luck and the third set would work flawlessly but I want to see all my options laid out and then make a decision because this is inconvenient beyond belief. I already wiped my machine too so either way it's not getting any use until something is done.

And yeah I had a ton of crashes. That's what has been happening lately - artifacts and then driver crashes.

Sent from my HTC One_M8 using Tapatalk

 

I don't know about that obvious degradation stuff. That could of very well been user error.

 

Right. User error. When just using my machine like normal and babying the hell out of it. Check.

I may not be a video card modding God like yourself but I'm not an idiot either. I've been building computers since I was 12 and I know when something isn't functioning properly and how to diagnose it.

Sent from my HTC One_M8 using Tapatalk

 

And what would happen if i were to prove your cards work? then what?

 

Why would you want 880ms anyway? 780m has much better reliability and with the better bios, good overclocking potential as well. Also, they don't have temperature issues lol.

 

Not gonna happen if it is the inductors causing the problem as you say. Not without a hard mod. I'm not willing to go to stupid lengths to get something I spent a lot of money on to work right, I want it fixed and I have a reasonable expectation to have it fixed. Something changed in the month that I've had the cards. When I sent the core and temperature graphs to xotic, it was clear as day to them and it is clear as day to me that they are not functioning at the same level as they were a month ago. Logic tends to dictate that when things go south so quickly, they will continue to go south until the point of catastrophic failure unless the situation is remedied so I am seeking remedy through my warranty on the machine.

I didn't spend almost 5k to have a machine that I have to extensively modify to get working within the specifications that it was sold with. End of story.

Degradation becomes apparent when you look at a system in the exact same operating conditions it was in one month ago behaving drastically differently. A card with a 954MHz base clock should never run at 850 with a 95-98% load.

Sent from my HTC One_M8 using Tapatalk

 

If yhou are asking me...it would just be for points on the bot for sli. LOL.

I'm going to let it go. You just don't understand. You're to busy playing defense my friend.

 

I think maybe sometimes you don't realize your posts are taken as being aggressive, hence the defensive stance.

I see your point, I'm just irritated that the cards literally went from working in acceptable parameters to being completely unacceptable literally overnight. It has me pissed off. Don't mind me, I'm angry.

Sent from my HTC One_M8 using Tapatalk

 

No worries, we all have those days....

What you have experienced i have done more times than i can count. When i thought cards were done for only to find out they were not. when things were working perfectly fine only to shut down for the night and come back nothing works. Down clocks, slow downs, over heating, not enough power or voltage to to much voltage. And so on...
I get that. And in your case if you're not running a dual psu you wouldn't be able to get to much further either. These machines can eat up 500+ watts. 330 just on stock and psu's usually max out around 400W to 440W. The rms ones that is. Still enough power to enough ones "working" system

 

I actually have dual PSU (I bought two PSU when I got the machine and bought the converter box separately) and that was one of the first things I tested. I really don't have an explanation for the difference. I already wiped the entire machine but these are the two graphs I sent Xotic, one the day I got my machine the other was taken just the other day when I had a crash in watch Dogs that artifacted and froze the system. I used Bioshock Infinite for both.

Here was brand new. Very little throttling and both cards synced together



Here is what happened when I ran the same test after the Watch Dogs failure.



See my concern? I ran Valley and Heaven as well and they both were pretty similar and look at the core temp on the slave despite the dropping core clock.

I used DDU to uninstall the drivers and reinstalled them. I tried 337.88, 340, and the new 344 drivers with all the same effect. Starts off high and quickly scales down while temps continue to rise and eventually there will be artifacts and the driver will crash when the GPUs max at 87C before fans kick up and knock the temps back down.

Some games are worse than others, Bioshock was my benchmark to go by since it can consistently keep the cards over 95% load without the embarrassing throttle that a benchmark would have had to achieve around the same load.

Sent from my HTC One_M8 using Tapatalk

 

Yes, I see what you mean. It has happen to me more than i care to count..

What program are you using for that?
And you have nothing that looks like this?

 

I use GPU-Z to make logs of each GPU (one instance per GPU) then I use a tool called Generic Log Viewer to sort the results out from GPU-Z into a usable graph.

I used HWiNFO64 and RTSS to have an OSD and I monitor both cards (temp, core, voltage), my CPU max core temp, and my frame rate.

Sent from my HTC One_M8 using Tapatalk

 

Good deal.
Now set it for hot keys so you can record and take snap shots. (Meaning. When you get your new machine)
You can then take a game or benchmark that can be ran in windows mode side by side. Open up something to move voltage and clocks and watch in real time what happens. That goes for cpu cores & speed. Not all games use quad cores and some even run better on dual core. (That is debate able though)
Graphs are fine and all, but it didn't tell you exactly what part of the game it was having problems. Some games will have a spot in it where even if you run 4 cards, it will still run that same low fps in that one particular spot. Example. 3dmark vantage when you first enter the tunnel, just before the hard left turn. Which would be fine.
Then start making on the fly changes to see what happens of if you can get away with lower voltage or more voltage. lower clocks or higher clocks.

Here is my 780 running factory clocks of 771/1250 and boosting to 993. Something we talked about before....

 

I did a lot of work last night including taking apart the assembly on the card that was acting up. Well one of the inductor pads was only half way on. I cleaned up the stock goop they used and replaced it with GC Extreme and fixed the alignment of that pad and installed 344 driver and did some testing. It didn't crash, still hot as ever but no crash.

This will work for now until the new cards drop. They still throttle but I ran heaven for 3 hours and it didn't crash, average core was 940.6 on the slave and 954 on the master which isn't too terrible.

So the system is usable again, at least for now. I am not going to do any mods and just wait until the new cards drop and work on getting those.

I find it odd that it didn't have the problem to begin with but that's how it goes.

Sent from my HTC One_M8 using Tapatalk

 

Well that's good news!

 

If pressure on that pad was not great it might have shifted maybe?

 

Yeah but it doesn't fix the temps or the throttle but I can game now. I'll take it.

I suppose that's possible. I should probably look into replacing the thermal pads, Sager didn't bother to replace any of them and one of the memory pads has a tear in it. Probably not enough to affect anything but I was visually inspecting the pads and I saw that the one that goes on the smaller inductor was actually flattened out visibly compared to the other two so I squeezed it together a bit to get the thickness back and it seemed to hold.

Maybe it's just me but if I had someone have a hardware failure that caused them to send in the machine and resulted in another expensive pair of cards being put in said machine, I'd replace all the pads too. -_-

Sent from my HTC One_M8 using Tapatalk

 

The factory pads will always be on the side of guaranteeing contact, I have always had gains from replacing them with slightly thinner ones where I can manually check contact is still good.

 

Temps and throttle would entail you going the extra miles to do something about that.

 

For stock it's not necessary with a decent paste job, but it does not hurt. Someone like me is always going to hand tweak it and it takes a bit of time and effort. Do be careful though as you must ensure everything does make contact.

 

Side note:
Make sure both cards read 8x or 16x if that's what the clevo reads. The M18x r2 reads 8x for it's lanes to the gpus. This will also create havoc on a system.

 

I can confirm it's meant to be x16 SLI off and x8 SLI on for our clevo models. I actually think consumer-grade haswell (and all the previous consumer-grade core i chips) doesn't have enough PCI/e lanes to do x16 in SLI, but I could be wrong about that.

 

Yes all non x platforms (which are the desktop quad channel ones) don't have enough lanes from the CPU to do 16x/16x. Desktop boards can add in a PLX chip but they tend to be a bit power hungry and add latency, not worth it for a simple two card setup.