i7-7700k + gtx 1080ti poor performance
Hello!Faced with the landings FPS in some games, I decided to watch the tests of this interaction in games on Youtube. To my surprise the results in tests were significantly superior to my system. The short of it (more in profile)
i7-7700k Gigabyte GeForce GTX 1080ti, 16 gb RAM, Win 10 Pro x64 (license), PS 750W, play in 1920x1080.
Will try to be brief about the essence of the problem:
It all started with the fact that in Crysis 3 after entering the liberty dome FPS drop from stable 60 to 30-40. In order to find the answer to this topic: http://forums.playground.ru/crysis_3/tech/crysis_3_gtx1080ti_35_fps_nuzhna_pomosch-953044/
In this thread people are actively shared with me tests of similar systems in various games. In one of these tests (video) demonstirovalos third Witcher, where I have noticed 100-110 FPS in Novigrad. In turn, I in Novigrad have 75-80 FPS max, on the same (highest possible) settings. The game license.
Surprised such a difference I decided to Google other available to me at the moment games. It came to DA: Inqusition. I stumbled on a video with a completely identical system: to make it easier on everyone I will say that the average FPS in this video 110-130 frames on maximum settings in full HD. I have certainly never been, and now more.
Deciding that business in curve Windows (with win 10, I faced for the first time after I updated the computer) I think the only solution may be to attempt a clean install of the system. Reset the system, and freewheeling only at the moment all the same Dragon Age Inquisition (currently swinging parallel to the third Witcher).
According to the results of reinstalling Windows... nichrome has not changed.
Actual screenshot from OED MSI Afterburner - as it is. Settings ultra, anti-aliasing and so on, all on high, in General above nowhere - as well as in the test on Youtube.
Spoiler
As you can see, loading the GPU to 99%, then Fpsu to grow nowhere else. What is the reason for the difference of my performance with the performance of the exact same system in the tests - for me it remains a mystery exactly why we are asking for your help.
From the following: Paying attention in the OED on line LIM1: Voltage. Googled in the Internet there is absolutely no information on this subject, but over the hill a little was found. Does this crap that supposedly the card has reached the limit of consumption on voltage and performance to grow nowhere else (and this limit is not associated with PD, this is what that dick knows what limit, at least I know enough). Foreign recommendation is simple - using MSI Afterburner to raise the Core Voltage by 10-20%. Made for 10, not helped, has not changed anything. Even thought the FPS dropped a couple units. The above increase was not, I am afraid that this may be something to burn in overlocking understand more than a monkey.
In General, friends, I need your help, what's the matter don't know. Why is the performance so low, although it may be higher? (judging by the tests so well above).
W_A_R_P 73
Someone more, someone less. I stock only 68-70 degrees. Add another video card and +10 degrees of being punished, and is the minimum. You are lucky with the temperatures.
Example of the difference upper / lower:
With DLC in overclocking SLI 60 degree easy. Well, the question SLI and the temperature here still STANDS Milivoje annealed, he did DLC on the top card dropped to 50 on the descent 1080. And it was over 70 degrees, can't remember exactly.
Short people. THANK YOU ALL FOR PARTICIPATING, PROBLEM SOLVED!
For future generations who will be googling the solution to this problem:
Following a three-day *bli found that the reason lies not in the software system and hardware. What is happening does not affect the system, drivers, settings, BIOS. It was established that the cause is PCI-E x16 v3.0 in which is inserted the graphics card! The system to determine how the video card is connected via PCI-E x1 v 1.1. I moved the video card into PCI-E x8 3.0 v and got a significant performance boost, from which he concluded that the problems with PCI-E x16 slot on the motherboard. When you return the card to its original position PCI-E x16 system again identified the card through the PCI-E x1 v1.1, and thus the performance drops again. I began to actively googling this problem, and all the resources pointed to 4 reasons:
1. Faulty PCI-E x16 slots on mother
2. Dirt/dust/pastaroni element in the PCi-E x16 slot on the mother
3. Problems in the socket CPU namely the incorrect installation of the CPU and the result could bend the contacts of the socket
4. Improper installation of the video card (SAG, SAG).
As a result, I decided to do the only true old - fashioned way- if something doesn't work, take it apart and assemble again.
in front of the TV and with grub, I slowly slowly dismantled the system unit to the screw. Cleaned out all the dust which has accumulated over half a year since the purchase of the computer. Got even the CPU from the socket to check (bent contacts were not found, although one of them looked suspicious - the tip was glistening not all).
Accordingly changed chudamani, cleaned the coolers.
Well, actually collected all ago.
And you won't believe all the fuck worked! I don't know what kind of magic was realized then, but my grandfather was right.
In short the GPU Z detects all korektno, the performance of games just grew. The Witcher 3 gave stable 100-120 FPS (including in Novigrad) in the forest at night in the rain slipped up to a maximum of 87 during the battle.
I think the problem is completely resolved, however, the final reason (besides the PCI-E x16) is not installed.
THANK YOU ALL FOR PARTICIPATING! ONCE AGAIN CONVINCED THAT THE FORUM PLAYGROUND IS THE MOST RESPONSIVE AND INTERESTED IN THE PROBLEMS OF OTHERS PEOPLE!
Special thanks to Steve Taller for extra help with recommended utilities reset configuration MSI AB and other detailed advice!
VOVAN WOLF
Well, the question SLI and the temperature here still STANDS Milivoje annealed, he did DLC on the top card dropped to 50 on the descent 1080. And it was over 70 degrees, can't remember exactly.
Remember, like a bad dream that happened a year ago. The temperature of the upper were vidyahi for hours in very demanding games 92C( Frequency kernel as a result of this temperature thrown off by default. Even turbo the acceleration was not kept. Although artifacts were not and any departures during the five months. But I took no chances and bought a dropsy (because summer was). Now the temperature of the upper 50-55C hard in team fights with good acceleration! Even less than the lower!) Manual overclocking is all right. Dough in dropsy of course a lot was wasted! BUT IT WAS WORTH IT!!!!!!!)
W_A_R_P 73 wrote:
How can you say that the Titan Z the most powerful graphics card today, inferior to it only 1080ti.
th fierce fowl??? The Titan Z it's an old map on the Kepler architecture consists of two titans black on one PCB, each chip has access to 6 gigs of video, maybe 12 total in two chips, I don't remember exactly. titan z has lost two in a sly titans black. Single Titan-Z conceded two 780ти in sly, maybe you have a titanium X Maxwell, then that's another card entirely and the use of 4-Wei sly little scalability when moving from two cards to three of the antler, and 4-I finally almost for show, and I've already written, and then wrote to you if you upload a video of gameplay to show the performance, do it with levels of MSI AB to show the frequency of percent, chip some, the consumption of video memory, and RAM, download and vidyahu percent, but you still post only performance with the Fraps fps one. You MSI AB to put the religion does not allow?
STANDS of Milivoje wrote:
her fever upper were vidyahi for hours in very demanding games 92C
When 780Ti in Sli stood, also heat on the top was. Solved the problem of crazy hands. For the bottom Board is cut from a sheet of aluminium the so-called backplate and stupidly put it directly to the Board (PCB) has bought into the system one 80 mm fan and set it to the fins of the graphics cards on the side where the inside of the case drove fresh air Cabinet cooler. In the end, the top card got less cold air, not heated by the bottom card. Temperature at 75 was under load then. Comp has successfully worked to sale. The new owner had recommended it and leave, but he looked at me like I was a moron and the whole system to take did not become. Now with the Assembly not know)
if we consider Tynan Zet as the graphics card that maybe she was the most powerful to the level 1070-80. but with the same success was dual Radeon 290. although the advantage of titanium a good amount of memory on the chip
and on expediency peck-peck-peck. Titan z one piece was more expensive than 2 Titan with the same chips. the 3-layer map and its characteristics are somewhat lower than conventional titanium. it was more profitable to just buy 4 normal titanium Tetra if (the second if we add mulberries Mat Board with plkh Aki ASUS h99 sun-e or ASRock h99 extreme 9 (that will fit in the difference of 2x titans z vs chetye ordinary ordinary лга2011 nurse and nurse for 30-40K) then each card individually would be more advantage seat in its expresse and isolated power and cooling) and this build was viable until some time PEK-PEK but probably still able to policymaking games where SLI is working fine
MunchkiN 616 wrote:
if we consider Tynan Zet as the graphics card that maybe she was the most powerful to the level 1070-80.
Titan z-one worse than 780ти in the sly, in sly 780ти inferior to 40 percent even 980ти, until she finally 1080 as the moon, if 2вэй sly in the games works, the growth of 3rd and 4th card is very weak
Steve Taller
I made a video? - Made
I showed the settings? - Showed.
I showed fps? - Showed.
I showed performance in very heavy games? Showed. (where would one 1080ti sucked horse))))
I only use PrepCom. MsiAB and GPU-Z in addition to what is shown and the frequency is wrong and the amount of video memory 12, and 4Gb.
That's not all.
Because MSI AB I have crashes in games. I need it? No. And why? Fps visible? Visible! Performance elegant for vidyahu four years ago. What else is needed for happiness, Carlo?)
Steve Taller
Carlo, I look again carefully the video (everything is in 4K). And then say someone who lost in the games)))))
Steve Taller wrote:
maybe you have a titanium X Maxwell
Wat???? You know better than me what I have in the system)))????????
This sucks I really don't need, Lord!)
fatallezzz
Solved the problem of crazy hands. For the bottom Board is cut from a sheet of aluminium the so-called backplate and stupidly put it directly to the Board (PCB) has bought into the system one 80 mm fan and set it to the fins of the graphics cards on the side where the inside of the case drove fresh air Cabinet cooler
Powerful!)+++
I'm so stirred up! I thought I one such pervert! But it turns out that is not after your story. But to me this does not work( At least though the experience gained, and time spent)
fatallezzz
In my profile photo, which some with dropsy, and where not. Sorry did I picture, when the cut sheet of aluminum and also put it on the PCB! (to dropsy)
W_A_R_P 73
there are no miracles, scalability 4 Wei sly little on the Titan black in sly in 1080p 60 frames a drawdown, and you have 4K 75 frames of MSI AB are millions, one of you crashes in games. 1080ти the way in 4K without Sparky pulls normally, you do not like MSI AB, Aida is also possible to monitor in games
Steve Taller wrote:
the increase from the 3rd and 4th card is very weak
misleading. growth always EN-number of times the if below the processing capabilities of the process and the cache priority, in other words if you come up with such an application where the Titan throws 7фпс 4 titanium (and for savennah games I think you can try to load one map to 10-12 fps) will give about 28 fps under the condition that the cards have enough free buffer memory and the processor as a whole is ready to serve in 4 times more Drava Callas than with a single card.
the increase in resolution above that of the free limit available bandwidth Express and the controller connects to the bridge notwithstanding, leads to poor masshtabnost and increase those delays that conventionally, the buffer of priority. so shipping should net grafonola.
another thing noticed is that the cross fire and SLI in Tetra work sometimes, not always though on idea should work through a recruitment method where the latency further increases. but there is pravelo 30фпс and full BH, peck-peck
MunchkiN 616 wrote:
where Titan throws 7фпс 4 titanium
MunchkiN 616 wrote:
will give about 28 fps
Never Sli nor crosshair will not give an N-multiple of the number of cards increase. Even if you take the banal SLI of two cards. Roughly speaking one card is 100%, then at best two cards in sly it will be 160%. In other words, if you have one card 100 FPS, this does not mean that the two of you can have 200 (with 160), but with 4 400. And again, I repeat - at best. Practice and the description shows that SLI in most cases gives only 30-45% increase, and that if the application supports in principle the technology (which is actually a little while, and those that are not always stable).
That is, even now let's have the owners of 1080 cards (without Ti). According to benchmarks (I do not know how to practice) 1080ti gives 30% better performance.
The owners of the 1080 is much better to upgrade to 1080ti and sit with one card than to buy a second 1080. And certainly tembolee where it will be better to take the next generation of cards (the amps? Volts?) than to buy a second 1080.
fatallezzz
can cite as an example the tale of the Witcher 3 and Givors that 690 is 2 680 sitting on one Express (in my case 2.0 which is comparable to installing 2 cards in common systems on Intel) and connected on a printed that is classic to if.
in padmaka 3 on ultra settings results with enabled and disabled SLI roughly coincided with the error in a couple fps of 40-45 personnel, compared to about 20-30 in single GPU. after manipulation of travona and increase the load on the GPU as a result of these manipulations upon closer pristegivaniem plot the performance on the GPU was 15 fps in SLI 30 respectively with an error of less than fps. similar results were given where the load on the GPU was boosted. including assain kred 4 where the load was boosted by the forced fold 8x anti-aliasing alpha transparency = мсаах8 in the game and all limit respectively to 30 fps and 15 fps on a single GPU. according to the game engine where he was able to use SLI and operacijam with shadows that are pretty well immersed the GPU can also be noted a two-fold increase with the error in a couple of fps. and most vintage cards of the type Radeon 4870х2 the same story in a number of games.
as an example a recent Eleks which if not supported at the time of testing and was threaded into manual so has an approximate power gains in the range of 80-90% on the second card because of A weak processor and problems on the part of the platform B - lack of VRAM the graphics card which I personally observed at a moderate lack of it leads to the loss of fps is especially pronounced in multi-GPU.
4 GPU-system I have not tested but I assume that pravelo 30фпс there also works
conclusion. if pritendovat at lower fps and resolution full BH and the like (for above me poorly known) it makes sense to upgrade to 2 video cards with lots of memory than on a single card but that is not faster than 2 times the results of the 1st weak of the existing card. another thing is that the amount of memory seems insufficient 1080 so the long-term course wins 1080ти. however, if 1080 and 1080ти had 16GB of VRAM and the task was to sit pridelnyh long on this bond within 30фпс and full BH choice would be a 2 card.
the essence of it is that the cards really don't give a 100% increase at speeds above about 40 to 60 and skraastad higher the gain is even less so starts to wag a number of extraneous factors and circumstances from progames of the application and problems super multi concurrency that will manifest if you take a map with a very large number of unified blocks and throw on the calculation of something simple
Sli is a good thing, I yuzal until may 2017 starting from 2010, i.e. 7 years was the user of the ligaments, I can say that the technology works in many games the scalability of the two maps was almost 100%, but the scale of the transition to 3 and 4 cards is not the same and not every game engine can benefit from 3 and 4 cards, the latest titles 2016-2017 stopped being friends with megacephaly and scalability has fallen dramatically or there are games where sly and the cross do not plow at all, the same quantum breakpoint, the same AK the origins, the same Agents of mayhem, mafia 3 and many other games where these figqy does not plow all games Windows stor type gears of the thief and our company Motorsport etc. doesn't use 2 cards, so I decided not to bother and to take one top out of the top of the amp if the difference with 1080ти will be 60% or more it is best to take the amp of course
PCI-E x16 slot closest to the CPU .I have the distance between it and the graphics card a few millimeters, although the graphics card with a thickness of 42 mm. And why do the main PCI-E x16 butt . With the upgrade torment.
Steve Taller wrote:
there are no miracles, scalability 4 Wei sly little, n
fatallezzz wrote:
Practice and the description shows that SLI in most cases gives only 30-45% increase, and that if the application supports in principle the technology (which is actually a little while, and those that are not always stable).
That is, even now let's have the owners of 1080 cards (without Ti). According to benchmarks (I do not know how to practice) 1080ti gives 30% better performance.
Again slowpoke to repeat the video. Where your 1080ti new modernity merges to complete!