Can I have a 2 way SLI GTX 660 with a Phenom II X4 965, and will it bottleneck the GPU's ?

Roxstrambler

Honorable
Apr 30, 2013
8
0
10,510
I want to have a 2 way SLI system with GTX 660's, and I was wondering if the Phenom II X4 965 will bottleneck the GPU's, if so what will it be like?
 
Yes it will severly bottleneck the gpu's. Even a newer FX chip at a higher frequency gets bottlenecked badly, so the phenom II's will be even worse. "Some" games will not, but alot of the newer games will.
http://www.tomshardware.com/reviews/fx-8350-core-i7-3770k-gaming-bottleneck,3407-5.html

I plan to get a second HD7950 soon, but I will be upgrading my cpu+motherboard first. In some games my X6 1090T(at 4.0ghz mind you) bottlenecks a single HD7950 a little bit. Not by much, but a second card will barely show an increase at all. If at all...

I'm not tring to discredit anyone, but the benches I linked show the intel chips putting up ALOT higher numbers while in crossfire even at crazy high resolutions where the cpu shouldn't matter as much. That is a cpu bottleneck if I ever saw one.
 


From the conclusion of the artcle I linked: "From now on, we'll need to limit the use of AMD's flagship to systems already bottlenecked by their graphics cards. A less expensive CPU is more attractive when it isn't affecting performance negatively." They are basically saying AMD's chips arent strong enough for more powerfull graphics subsystems.

This is also AMD's top tier cpu at the moment overclocked to 4.4ghz so the Ph II wouldnt have a chance since you have to be extremely lucky to get a PH II to 4.4ghz.

My next chip will be Intel, unless AMD pulls a rabbit out of their %$@*.



 

ish416

Honorable
Jul 5, 2012
771
0
11,360




That link you posted doesn't prove anything that wasn't already known. F1 2012 and Skyrim don't use the AMD FX architecture well at all. They basically use two cores then barely anything on the others. Thus, the Intels will be supreme as they have much better IPC and real physical cores (not counting hyper-threading) as opposed to AMD's modules.

BF3 and Crysis 3 show what the FX series are capable of when they can be properly utilized.

That said, my 965 @ 4,336 Mhz performed nearly as well or better than my stock I5 2500K.

3DMark Cloud Gate with 7870XT
Phenom II 965 @ 4,336Mhz - Score 14527 - http://www.3dmark.com/cg/208847
I5 2500K @ Stock (3.4Ghz) - Score 14551 - http://www.3dmark.com/cg/253776

Not what I would consider a huge difference.

Yes, the 2500K is an older CPU but the I5 3570K is what, at most 10% faster than the 2500K and on average 5% faster.

3570K vs 2500K
http://www.tomshardware.com/charts/cpu-charts-2012/compare,3157.html?prod%5B5755%5D=on&prod%5B5760%5D=on

So unless there is some SLI/Crossfire voodoo going on somewhere, why wouldn't the single card configs that I posted above also be representative of what these CPU's would be capable of while running two or more cards? I understand that there might be more of a gap between them but I wouldn't expect it to equal no gains unless it was something application specific. Will you see 100% scaling, no. I wouldn't expect the performance difference between a 965 @ 4,336Mhz and the 3570K @ stock speeds to be more than 20% in a worst case scenario when running a dual card system.
 


yeah... um.
changed my mind i'll just leave this comment be

I'll just say this~ I'm rolling with a 965, overclocked up to a modest 3.78ghz, and my system isn't bottlenecked in the slightest. of course my monitors refresh rate is just 75hz. So anything over 75fps i wouldn't see anyway.

In multithreaded games which are not coding disasters like Skyrim or Civ5 (both games are specifically optimised for intel cpus btw), the 965 is more then enough cpu to power pretty much any gpu you tack onto it.
 

maxalge

Champion
Ambassador


I had a 1005t @ 4ghz, you are bottlenecked.


Yes you would, in smoother gameplay thanks to better animation refresh, and since you are at higher fps the drops are not as bad/noticeable.


The 965 begins to bottleneck anything higher than a 7850, not to mention the higher frame latencies.

Its roughly equal to a i3 2100 when @ 4ghz.
 


That's what I thought I read in the under $200 cpu shootout on Tom's a few months back. I will admit that the AMD chips hold their own in about half the games going around these days, but intel holds their own in ALL titles. CPU dependant or not. Right now I recommend Intel chips for no bottlenecks at all from the CPU in any title, and I have a hard time recommending any AMD chip since GPUs are getting faster and the gap most likely will widen over time in CPU intensive games. When crossfire/sli come into the picture it will most likely matter even more so.

OP: You will be bottlenecked in some games, and not in others. As time goes on you will probally want to look into a swap of cpu and motherboard to see any further performance increase from your SLI/Crossfire setup. Making the CPU switch now and just buying one single powerful card is another option to consider as well. Then when you decide to grab a second card you wouldn't have to worry about a bottleneck, and by that time the price will have most likely dropped quite a bit as well. Just something else to consider.



 


Are you really trying to compare a PH II 965 @ 4.336mhz vs. a Stock i5 2500k in 3dMark? Gee my chip at 4.0ghz should be better than any intel i5 then with 6 cores fully utilised, sadly that's only in the case of 3dmark and a few games that are properly optimised. Most are not since they are console ports. Most of the Phenom II chips out there can barely reach 4.2ghz, and then you need aftermarket cooling to even get to that point or beyond. Most of the Phenom II chips that I have seen are only hitting 4.2ghz or higher on the newer stepping as well. The older ones were just barely getting to 4.0/4.1ghz if you were lucky. How much faster would the difference of been in 3dmark if both chips were at that speed, say 4.3ghz? You know, apples to apples comparison....

The OP wanted to know if there would be a bottleneck, and yes in some situations there will be with a Phenom II. A newer intel I5/I7 chip wouldn't present this issue in games. The gap gets smaller with FX chips but not by very much. This is what I have said from the very beginning, but a few people come in and say no it wouldn't, and then say only a couple of games will be affected as if they don't matter. Well what about games coming in the future? There is bound to be more titles showing the same bottlenecks as the ones that already do.

PS: I said earlier that 2 x 660's would be severely bottlenecked, and it probally isn't as bad as I made it sound, since they are only mid range cards. For that I am sorry. More powerful graphics subsystems would present a bottleneck more clearly, and that is why I was trying to make my point clear.
 

ish416

Honorable
Jul 5, 2012
771
0
11,360


Yes, I believe I just did compare a 965 @ 4,336 Mhz vs a stock I5 2500K and the difference is very small in 3DMark. A 24 point difference is very insignificant. I understand that 3DMark is a synthetic test and not 100% accurate in the real world. It will however, represent what a system is actually capable of under ideal circumstances.

I did post a link to an article that anandtech just put out and it shows that a system with similar performing components to the OPs will scale very well in SLI.

Link - http://www.anandtech.com/show/6934/choosing-a-gaming-cpu-single-multigpu-at-1440p

The AMD 960T CPU is slightly slower than a 965. The GTX580 typically performs within 5% +/- of the GTX660.

This article shows great scaling in all games except Civ5.

Also in that article, it also shows exactly the bottleneck that sincreator is referring to. Look at the 7970 crossfire numbers, they don't scale nearly as well as the GTX580 on the lower end CPUs.
 


you're right, you're lucky if the PhII x4 965 OCs on air over 4ghz stable. I could get mine up to 4.2, but there was no cpu cooler on the planet that could keep that chip from melting down, and i've got no idea if it was stable (i killed that overclock the moment i saw the temps idle)

Frankly, i couldn't get an overclock on my PhII over 3.8ghz that wasn't so hot the system was unstable alone under stress from the temps. Purhaps with a water cooling system i could work it, but not with air.

But comparing an i5 and a PhII at the same clock speeds isn't "apple to apples". The i5 runs a lot more instructions per cycle. so at the same overclock it will be the faster chip. The poster was just making the point that the PhII could reach i5 stock speeds, that there wasn't that huge of a difference between companies products.

His pointing out that two of those cards in Sli will scale at 85%, means there basically was no bottleneck from the chip.

AMD cpus ARE slower then Intel. But they are generally "good enough" with ANY gpu out there. When overclocked propperly most AMD cpus more then carry their own weight. Most of the benches i've seen that show AMD cpus lagging with GPU performance occures with RADEON gpus and stock CPUs.

NVidia gpus are not really sensitive to cpu speed. AMD Radeon gpus ARE sensitive to cpu speed.



Nvidia gpus are not cpu sensitive. And the 660 in SLi is plenty strong enough to stomp all over a single 7970... might even get up to Titan speeds actually

I get what you're trying to say. and i get you don't like the performance of your system. but lets not color the facts with your angst.
 
The reason I will be upgrading to an Intel chip is because I want my system to not be held back in any game I decide to play, when I decide to grab a second card. As it stands right now I am happy with my system with a single GPU. If I was to grab a second card and played a game that barely showed a performance increase because of my CPU it would feel like I wasted the money on purchasing a second card.

Either way this thread hasn't really solved any question that the original poster asked. Just a good debate from both sides of the coin, IMO. I wish some of the Mods/Writers of these forums and the website would chime in with their opinions. Maybe I'd learn something new as well. Maybe I won't be effected in the way I assume, and the performance will be better than I think it would be? Anyone reading this recently upgrade from a PH II x4/x6 to a 2500k/3570k running SLI/Crossfire that has any numbers to share? Maybe it actually would save me a bit of money if I'm wrong, and I can grab a second card without worrying about a new CPU and motherboard.