• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

780 x 2 possible with this PC?

IMO people yell bottleneck a bit too fast ! :)
I ran 3x580 with Ci7 870, 3x 670 with Ci7 3820/3930K all stock clocks and that at 1920x1200 in the beginning and I still got great scaling and very good boosts - running with 16x or 32xQCSAA has been a bless.

Now going to 2560x1440 on my own PB278Q the 3x 670 and now 3x 780 really show their worth and the scaling is even better in higher res'. I'm still running stock speed and yes I could get more performance with OC'ing - you always can !
Bottleneck or not I'm still getting stellar performance with 3x 780 and a stock clocked 3930k and I'm very happy with that.

My advice is don't listen too much to all those bottleneck answers because you will always hear them and no matter how powerful setup you put together there's always some kind of bottleneck to find.
I've been running SLi setups permanently since 2006 and only one setup I had extremely overclocked (AMD 64 X2 4600+ [email protected] on Corsair Nautilus 500 H2O, MSI GeForce 7900GTO SLI@Voltmodded core 872MHz, Mem 1020MHz) to "overcome bottlenecks", but hereafter I've been only running stock clocks because I still feel I get great performance. All the time (endless stability testing) and effort (also read money) I put into the extreme OC'ing didn't really make games run "better" than with stock clocks, they just ran faster.

With a good amount of GPU grunt you can always turn on the additional eye candy that the driver allows or via SweetFX. Higher levels of AA or combined AA models make jaggies a thing of the past... and I really don't like jaggies !.

The difference with the OP is he is using an older 920 which is no where near as efficient as your 3820 and 3930k.

I don't tend to run my CPUs very fast most of the time either and find 4.0ghz is fine.

What I will do when I am off work in 4 days time is run some benches with both an i7 9xx and also a 3960X both at 3.8ghz the same as the OP is using. I will also run the 3960X with 2 cores disabled to see if that makes any difference. I will use 2 of NVidias fastest out of the box cards so there will be no question about the results.
 
The difference with the OP is he is using an older 920 which is no where near as efficient as your 3820 and 3930k.

I don't tend to run my CPUs very fast most of the time either and find 4.0ghz is fine.

What I will do when I am off work in 4 days time is run some benches with both an i7 9xx and also a 3960X both at 3.8ghz the same as the OP is using. I will also run the 3960X with 2 cores disabled to see if that makes any difference. I will use 2 of NVidias fastest out of the box cards so there will be no question about the results.

Yep, but even before this with my 3x 580's I used them on an Asus M4N82 Deluxe motherboard, 8GB DDR2 PC6400 and a stock AMD Phenom II X6 1090T.
I then upgraded to 8GB DDR3 1333MHz, Asus P7P55WS, Core i7 870 before ending up with 16GB 1333MHz, Gigabyte X79UD3, Core i7 3820.
I had these 3x 580's through 3 different platforms and I never really felt they the cards "weren't performing" on either platform eventhough there was a significant performance difference between the Phenom II X6 1090T and the Ci7 3820. Playing at 1920x1200 with max 16x/32x AA I never really saw that hughe difference going from the X6 1090T to the ci7 3820... it was really only synthetic benchmarks showing the gains.
I'm pretty conviced the OP will still get a good gain, and he can always transfer the cards to an never platform and get the extra boost that gives should he consider that in the future ;)

However I think it would be cool if you'd make such a test and also enabled max eye candy and AA levels !... also to see if there PCI-E bandwidth wise would be something significant holding against him.

Hmm now thinking of it I have an ASUS CROSSHAIR IV FORMULA motherboard + AMD Phenom II X6 1090T coming to me monday and I have 8GB 1333MHz laying around.... I could throw this into my "test/review rig" and pull out some 780's and try some SLi :D
 
Last edited:
Depends if his Cpu will run at 4.5 stable enough for benchmarking.

I think it can just about do 4.3ghz as it has got an old corsair H50 on it, if I put a decent watercooled loop on it I would do a lot better as it does not need much volts.

What I expect to happen is at 3.8ghz the i7 9xx and the 3960x on 4 or 6 cores will bottleneck a pair of GTX 690s. The difference is the SB-E will be able to manage more fps as it is a bit more efficient. Once the CPU clockspeed gets to 4.3ghz there will still be some bottlenecking on both but not much.

I will test with max settings @1080p with

Heaven 4
3dmark11 - only the graphics tests
Tomb Raider
Sleeping dogs
Firestrike - only the graphics tests
SE V2

If anyone can think of any others I have used I will do those too.

I do the above on tuesday when I am off work.

A 3960X on 4 cores is near enough a 2600k.
 
Depends if his Cpu will run at 4.5 stable enough for benchmarking.
Good luck getting 4.5ghz on a bloomfield for 24/7 use. Verry very hard to get a chip capable of that, ive owned a 4.2ghz i7 920 and still have a 4.3ghz capable i7 930. Both were bottlenecks with sli gtx 670's. I had to go to ivybridge 3770k to get the best out of those cards, (3570k first). No difference with my current 4770k in sig with those cards.
 
I think it can just about do 4.3ghz as it has got an old corsair H50 on it, if I put a decent watercooled loop on it I would do a lot better as it does not need much volts.

What I expect to happen is at 3.8ghz the i7 9xx and the 3960x on 4 or 6 cores will bottleneck a pair of GTX 690s. The difference is the SB-E will be able to manage more fps as it is a bit more efficient. Once the CPU clockspeed gets to 4.3ghz there will still be some bottlenecking on both but not much.

I will test with max settings @1080p with

Heaven 4
3dmark11 - only the graphics tests
Tomb Raider
Sleeping dogs
Firestrike - only the graphics tests
SE V2

If anyone can think of any others I have used I will do those too.

I do the above on tuesday when I am off work.

A 3960X on 4 cores is near enough a 2600k.

Thanks Kaap , same cooler as my i7 960 over 1.4volts mine starts to hit low 80's and i'm too lazy to w/c it properly
Dont expect you to blow it up just curious i guess how much it will change
 
Thanks Kaap , same cooler as my i7 960 over 1.4volts mine starts to hit low 80's and i'm too lazy to w/c it properly
Dont expect you to blow it up just curious i guess how much it will change

I will give it a run on Tuesday when I am off work and can move the cards about but I will post a few benches with my 3960X @3.8, 4.3 and 4.8 with two GTX 690s in a bit.
 
The more you overclock the CPU, the less the bottleneck would be.
I would agree with the faster the cpu the more frames per second you might get, for example i know on my system that if i'm testing crysis 3 with a i7-3820 + gtx 780 2-way sli on 1920x1080 @120Hz the stock default i7-3820 @ 3.6 GHz (cpu boost 2.0 @ 3.8 GHz) the framerate in crysis 3 is quite low and poor performance. But bumping up the 3820 by 400 MHz to even an average low overclock to 4.2 GHz the framerate increases and performance in crysis 3 is so much better. I may take some screenshots later to compare the framerates. Although i'm not 100% sure if the minimums have given me the smoother performance, maybe that's what it is. But at the higher resolution like the original poster has you would most likely need a very fast cpu to help keep the experience smooth.


EDIT: Just tested Crysis 3 at maximum settings plus MSAA @ 8x, 1920x1080 @120Hz. Test with CPU i7-3820 @ 3.8GHz and @4.4GHz, 3 frames per second difference, just not worth it. Of course that might depend on the scene being drawn/rendered whilst in game. I suppose you would have to go way beyond 4.4 GHz to actually make the overclock worth while. Maybe it's all about overclocking the graphics cards that actually makes it worth while?. can't be bothered with the crap, stick with what i got, i reckon i would always recomend gpu upgrade if its gaming at 1920x1080 and above.
 
Last edited:
Yet he states he has a 60fps monitor....
And like i said the bottleneck is only above 60fps below that there isnt one as the cpu is not restricting gpu usage.
Except that a i7 920 ISN'T going to hold a solid 60fps if one was playing a demanding game online.

Max frame rate hitting above 60fps means very little if average and minimum dip down to 30-50fps frequently.
 
Good luck getting 4.5ghz on a bloomfield for 24/7 use. Verry very hard to get a chip capable of that, ive owned a 4.2ghz i7 920 and still have a 4.3ghz capable i7 930. Both were bottlenecks with sli gtx 670's. I had to go to ivybridge 3770k to get the best out of those cards, (3570k first). No difference with my current 4770k in sig with those cards.

Hence why I queried it ;)
 
My old [email protected] and 7970 xfire results no bottleneck here and 139fps.This is pre mantle as well. Maxed at 2560

4 threads think this BF3



8 threads



i7 920 more than capable with old tech lol. This was a cherry picked 920 that would do 4.5ghz all day long, but i chose to run it at 4.4ghz just to keep the voltage a little lower 3.75v at 4.5ghz on air.

I would presume that one 780 or 780ti would match these results somewhat as my Heaven 4 benchmark for one 780ti matches 7970 xfire (780ti with 4770k at 4.4ghz)
 
Last edited:
My 960 is quite happy at 4.5 assumed all were sorry ,, was bought off a bencher though so i should off looked a bit harder, maybe i should look after it a bit more :)
aint looked at oc/ing them since i got it 3-4years ago so old brain gets foggy
 
Last edited:
3960X results with two stock GTX 690s

Heaven 4

Gentle increases with CPU clockspeed.

CPU 3.8
Q9ACuVx.jpg


CPU 4.3
JzfzMGv.jpg


CPU 4.8
0KUvUqo.jpg


3dmark11 graphics score

3dmark11 and Firestrike surprised me as there was very little difference in score as the clockspeed increased.

CPU 3.8
OllRRJX.jpg


CPU 4.3
uXqLFqw.jpg


CPU 4.8
XKkLHVL.jpg


Firestrike graphics score

CPU 3.8
0GYK2GY.jpg


CPU 4.3
X8iyb6E.jpg


CPU 4.8
dJdzJRB.jpg


Sleeping Dogs

Things got really interesting on this check out the fps go up as the CPU clockspeed increase.

CPU 3.8
Ta5j9Ne.jpg


CPU 4.3
sb1bMds.jpg


CPU 4.8
LQ12FNb.jpg


Tomb Raider

I was not expecting much from this game as it is not very demanding on the CPU.

CPU 3.8
7AMokMj.jpg


CPU 4.3
FB3OwAo.jpg


CPU 4.8
RLZSBoz.jpg


SEV2

Again another game that does not really tax the CPU

CPU 3.8
mmCBoy3.jpg


CPU 4.3
5xSmfIS.jpg


CPU 4.8
8SdibLa.jpg


Does using a Hexcore make any difference, in these last pics I run the 3960X with 2 cores turned off but still overclocked to 4.8. Check out the Sleeping Dogs score compared to the 6 core score, the extra 2 cores make quite a difference on SD but not on the other two games.

e7Dbvb0.jpg


nq7nPan.jpg


sj2uI4L.jpg


So I would say we are getting CPU bottlenecking on Sleeping Dogs which is very demanding on the CPU but the other games don't have a problem.

I would also place BF4 and Crysis 3 in the same group as Sleeping dogs but unfortunately they don't have inbuilt benchmarks to test with.
 
Just enabled Prefer maximum performance & Single display performance mode in the Nvidia control panel and now it's bouncing off the 200 Fps rev limiter in BF4. Gpu usage not dropping as low.
 
My old [email protected] and 7970 xfire results no bottleneck here and 139fps.This is pre mantle as well. Maxed at 2560

4 threads think this BF3



8 threads



i7 920 more than capable with old tech lol. This was a cherry picked 920 that would do 4.5ghz all day long, but i chose to run it at 4.4ghz just to keep the voltage a little lower 3.75v at 4.5ghz on air.

I would presume that one 780 or 780ti would match these results somewhat as my Heaven 4 benchmark for one 780ti matches 7970 xfire (780ti with 4770k at 4.4ghz)
Except the examples of scenes you provided there's hardly anything going on- probably even lesser than some action scenes on singlyplayer campaign. Frame rate pitfall the sharply when there are lots of players/objects in sight, shait loads of things happening and explosions all over the place at the same time. And it's worth bearing in mind that like you yourself said you had your i7 920 on 4.40GHz, and reading OP it is most likely that he would only be running his i7 920 at the current 3.80GHz.
 
Last edited:
Except the examples of scenes you provided there's hardly anything going on- probably even lesser than some action scenes on singlyplayer campaign. Frame rate pitfall the sharply when there are lots of players/objects in sight, shait loads of things happening and explosions all over the place at the same time. And it's worth bearing in mind that like you yourself said you had your i7 920 on 4.40GHz, and reading OP it is most likely that he would only be running his i7 920 at the current 3.80GHz.

This is true i was just showing what the old dog could do, i only played 64 player games these were just taken at random while ingame. When i used vsync it very rarely dropped below the 60fps, i would not want the op to make the same mistake is i did, this





I got the upgrade bug and hit it with over 2k of goodies thinking this was the best way to go, but i used the same GPU the 7970 xfire and there was no improvement at all the 4930k was at 4.4ghz also.

That system was sent back (it was sold with a guarantee of the Cpu doing 4.8ghz, overclockers typo error, that's another story)

Then i went 4770k with sli 780 then sli 780ti only then did i get what i wanted.

Op will have to go 4770k and Sli to get what he wants or do what Kappstad said in the begining try another 780 first and see what happens.
 
Last edited:
Back
Top Bottom