2600k 3GPU vs 3930k 2GPU

[abeeftec]

I tried this in another forum and of course was accused by people who either just purchased the z68 platform or have owned it of spreading false information

This is NOT intended to insult anyones system in any way.

It is clearly for the educational purposes of the screen shots themselves and in that I will leave any further comment past this first post out unless a question is asked. Please, If you want a benchmark ran with both systems you will have to hold on until I get a new X79 board in. I blew up my board by condensation build up while putting it on Phase cooling.

In every benchmark the 2600k has 3 EVGA Superclocked GTX570s and the 3930k Has 2 EVGA Superclocked GTX570s. In some cases the 2600k is sped up 100mhz more then the 3930k.

Please note that Reviews are subjective to the hardware they are written about. They are Not the law when different variables are in play. I see many people posting information based soley on reviews. One thing I have found to be true, The X79 has more bandwidth and can handle 2 GPUs better thenthe Z68 can handle 3. My Z68X-UD7 has 2x16 lanes(one of the ONLY Z68 boards with 32 lanes) and with 3 GPUs is running in X16 X8 X8 whereas my X79 with 2 GPUs is running in X16 X16. I do think that there is no way to get around the fact that the 3930k with 2 GPUs is a better gaming machine then a Z68 with 3 GPUs. It runs smoother and it is wide open. And there is a noticeable difference between the two when running games. This is my opinion having used both playing games such as,Tribes Vengeance, Tribes Ascend, Crysis 2 And Portal 2. Those are the 4 games I have played on Both systems and though the 2600k is absolutely smooth as can be with never even a hint of stutter in all 4 games with 3 cards, The 3930k seems a bit more fluidly smooth and does get More FPS at the same clock speeds with 2 cards then the 2600k does with 3. CPU and Chipset does have a lot to do with Utilization of Graphics. I cant explain why it seems more fluid but it does.

Now, Do not get me wrong here, I am not dissing the 2600k. I own the system. I am just saying that for those who do not haggle about price/performance ratios and who want to Run SLI or TRI SLI, The X79 is the best platform available for this.

3930k 4.1ghz 2 570 3Dmark 06 I messed up when taking this screen shot and changed all the CPUz shots so the speed cannot be seen but it is 4.1ghz

[/IMG]

2600k 4.2ghz 3 570 3Dmark 06

[/IMG]

3930k 4.4ghz 2 570 3Dmark 11

[/IMG]

2600k 4.4ghz 3 570 3Dmark 11

[/IMG]


Let me make clear to everyone who reads this that some things in this post are my opinion. I also want to say that the 3Dmark11 Shots are skewed results because
I did not know when I ran it that I needed to INPUT the amount of cards running so it really severly messed up the physics and combined score. I will redo them once I get my X79 system up and running.

More to come as soon as I bench the 2600k at higher speeds. I already have the 3930k results as I benched them just this past week all week long.
I am doing the 2600k Benchmarks over again even though I have a ton from Jan. and Feb because there has been patches made on these benchmark programs and the video drivers were different since the beginning of the year and I want to use the same exact version for each set up along with the same exact video drivers.

EDIT: Please dont hijack my thread with arguements about why I am wrong here. If anything is wrong it is not me. I simply loaded the systems and ran the benchmarks. I cannot tell the benchmark what numbers to throw out when they are completed

 

[Denithor]

First, your pictures aren't loading correctly.

Second, how do the systems compare if you do a direct apples-to-apples matchup? IE Z68+2x570 vs X79+2x570. Because it's definitely harder to juggle the throughput on three cards than two, giving the X79 system a distinct advantage right from the start.

 

[abeeftec]

First, your pictures aren't loading correctly.

Second, how do the systems compare if you do a direct apples-to-apples matchup? IE Z68+2x570 vs X79+2x570. Because it's definitely harder to juggle the throughput on three cards than two, giving the X79 system a distinct advantage right from the start.

It took me a few minutes to figure out how to even get them to show up on the page.

I have not ran the direct comparison but I can do that with the 2600k with no problem. But I can tell you from running two and three cards earlier in the year I am getting MORE FPS from 3 cards then 2 on the Z68 platform so what is the point of doing that? It is still not coming out on top in overall score using 3 cards. Why would it do better with 2? Though if it is just an equal comparison you want to see then I understand. But I can tell you without running them again that it will spread the gap even further.

Plus The point was giving the 2600k an advantage. The only Program I have seen so far that doesnt seem to raise at all in FPS with 3 cards over 2 on the Z68 platform is 3Dmarks06.Well not at all but very little gain with 3 cards if any. Which is why I have ran other benchmarks as well. Because there is an increase in FPS and score with 3 cards over 2 on 3Dmark11 with Z68. I am as shocked at this as anyone else is and cannot explain why this is the case. But I have even tried a different 2600k, different cards, Cards rearrainged in different slots and even reloading of the OS and drivers. The results are the same. The X79 overall scores higher in every benchmark Furturemark makes using less cards.


If you can explain to me the best way to load screen shots I will certanly correct it. I really dont post in tons of forums and the other forum I posted in was very easy to just load the pics right out of your PC instead of needing to put them on some site like photobucket.

Edit: I just thought about something after posting this response... In what way to you mean " I am giving the X79 system and advantage"
How does 2 cards give that platform and advantage over 3 cards running? Is not TRI SLI better then SLI?

 

[faseman]

Synthetic benchmarks will do that. I don't use 3DMark but it probably gives too much weight to CPU power. Try playing some games, the tri-SLI will win most benchmarks.

 

[Shmee]

also, what resolutions would you be benching at in games? Recommend highest possible. I would really like to see how much difference 3 way SLI on either system is effected by PCIE bandwidth as well as CPU power, if at all.

 

[abeeftec]

I do believe it is PCIe bandwidth that is the issue in some ways.That and the CPU is just faster overall. The 3930k at 4.5ghz had way higher scores then the 2600k at 5.0ghz on many benchmarks.

I remember a year ago clocking X58 with these cards alongside p67 on a UD7 board and wondering why the benchmarks appears to be so strange. It was almost as if the 3rd card in the p67 board was a mute point. I cant remember for sure exactly why but I just remember thinking then that my new set up had issues and trying to search out why but found it did not. I do know that I found Bclk overclocking had something to do with raised Graphics performance that p67 did not seem to have by multi climbing. I am not saying there were not gains or that z68 wasnt faster overall. Just that the gains were not as high percentage wise Multi climbing as they were Bclk climbing. I could be wrong there but it seemed I got a higher percentage of "gains" with X58.

I normally bench on whatever is considered default by Futuremarks when using their products. And normally with 3D marks 11 I go with the Extreme settings.. The only reason I did not do it with 3dmarks11 at this point was because when my X79 system got wet I lost all the 3Dmark 11 benchmarks. The only one I had saved was the Performance test results at 4.4ghz with 2 cards in a storage drive the day before. So I ran that with the 2600k with 3 cards to see.
I will do it again once i get my board back. I am dying to see if there is any difference though I cant see why there would be using the same cards. At diff resolutions I would think it would have a similar gap between the two systems. It's not as if the cards are different. Is it not true that some cards handle higher resolutions better? I never thought the same cards would perform differently on different systems at different resolutions in that regard.

There seems to always be overall system gains by raising the CPU. Better graphic performance, Higher FPS. So I suppose the CPU power is raising overall performance. I think it would be even higher if the BCLK could be raised as it was in X58. I speculate Intel took so long to release this because they wanted to stop BCLK overclocking as opened up as it was so people would start buying their Extreme series and stop getting gains over the Extreme by raising the BCLK on lower end CPUs like in X58.

 

[Vesku]

The 3930K with 2 570s has a 3DMark11 GPU score of 11748
The 2600K with 3 570s has a 3DMark11 GPU score of 17018

I don't know enough about the esoterics of 3DMark11's final P score to really know why the combined is so low for the 2600K, but I think gaming results of both running 2 570s would be more informative.

I appreciate the effort you have gone to with this post, though.

 

[Grooveriding]

Looks obvious to me. In your 4.4 benches the 2600K system's GPU score is higher than the 3930K's, as it should be. The 3930K's CPU score is higher, as it should be. This is what is giving it a better 3Dmark score.

3Dmark takes advantage of cores and threads, so you basically have a system that is 50% faster CPU wise than a 2600K in 3DMark.

Try benchmarking an actual game the tri-sli system is going to have a much better framerate than the dual-sli 3930K.

 

[abeeftec]

The 3930K with 2 570s has a 3DMark11 GPU score of 11748
The 2600K with 3 570s has a 3DMark11 GPU score of 17018

I don't know enough about the esoterics of 3DMark11's final P score to really know why the combined is so low for the 2600K, but I think gaming results of both running 2 570s would be more informative.

I appreciate the effort you have gone to with this post, though.

Where are you getting the score for the 2600k with 3 570s at 17018?
I have never scored that with this system. not even close

 

[abeeftec]

Looks obvious to me. In your 4.4 benches the 2600K system's GPU score is higher than the 3930K's, as it should be. The 3930K's CPU score is higher, as it should be. This is what is giving it a better 3Dmark score.

3Dmark takes advantage of cores and threads, so you basically have a system that is 50% faster CPU wise than a 2600K in 3DMark.

Try benchmarking an actual game the tri-sli system is going to have a much better framerate than the dual-sli 3930K.

I would think the combined score comes from the test towards the end where I dont think anyone gets high FPS.(the name slips my mind) Look at the difference in those two scores. The score with the 3930k is killing the 2600k by way more score then 50 percent more. And it appears that has a larger impact on overall then an entire GPU does.
In Vantage, with PPU disabled the 2 cards still beats the 3 cards. It still doesnt make sense to me. IF these CPUs are really that much different (and they are) then Chipset performance has a lot more to do with gaming performance then most people think(And I believe it does). I mean, the X79 chipset has brought back high performance across the board for sure.
Honestly, When I went back to the Z68 system after I blew up my X79 board with water, I noticed a definite lack of performance overall from the Z68 for the first time since I had it. But I spend all day on my PC so I notice any slack in it. It made me instantly order another X79 board and a couple more 3930k's.

 

[faseman]

Run some real-world benchmarks with actual evidence of your claimed performance differences. Synthetic benchmarks are not true indicators of what you will get out of your PC.

 

[AdamK47]

I'll run some game benchmarks. What do you want to run? You run yours, I'll run mine and post the results. I'll even keep my CPU at a modest 4.3GHz during the benchmarking.

Intel Core i7 3960X @ 4300MHz
Asus P9X79 Deluxe @ 43 x 100MHz
16GB Corsair XMS3 @ 2133 DDR
Three GTX 580 in Tri-SLI @ 850/2200
120GB Vertex 3 SSD
Two 3TB Deskstar 7K3000 in RAID-0 - 6TB
1TB SpinPoint F1
Pioneer BDR-206 BD-RW
Sound Blaster X-Fi HD USB
Cooler Master HAF-X case
Prolimatech Megahalems cooling
Corsair AX1200 power supply


The Synthetics:

 

[Vesku]

Where are you getting the score for the 2600k with 3 570s at 17018?
I have never scored that with this system. not even close

From the screen you posted saying '2600K at 4.4GHz with 3 570s'.

 

[mrjoltcola]

Hi OP, welcome to AT.

I think you will find ATers are the most friendly when it comes to factual information and scientific approaches to things, much more of a mature forum than certain other forums I know, yet at the same time, intolerant of FUD, misinformation, and apples-to-oranges comparisons.

As to your comparison, I'd like to see Apple to Apples. We can all speculate, but giving the 2600K a leg up really doesn't give us much factual data. Too many variables to consider, Amdahls law, non-linear CPU/GPU scaling, etc. There are algorithms and implementations that run worse the more CPUs you throw at it.

So post some comparisons head-to-head at similar CPU clocks, similar GPU setup (2x SLI) and I'm all ears. Like AdamK47, I also have a 2600K and a 3930K with multiple GPUs, so I'd be happy to help with some benchmarks of my own. I was planning to do this already, but had a bad Rampage board, and just got the replacement in yesterday.

 

[PowerGPU]

OK you r getting very low scores for the 2600k with three 570's here is a video I did with my three machines and comeing soon three way crossfire with 7970's with a 3930k

http://www.youtube.com/watch?v=LyTvazcMC3E&list=UUxhFeDHHW4Eoo4LEVmOPoPw&index=4&feature=plpp_video

 

[Zap]

The X79 has more bandwidth and can handle 2 GPUs better thenthe Z68 can handle 3. My Z68X-UD7 has 2x16 lanes(one of the ONLY Z68 boards with 32 lanes) and with 3 GPUs is running in X16 X8 X8 whereas my X79 with 2 GPUs is running in X16 X16. I do think that there is no way to get around the fact that the 3930k with 2 GPUs is a better gaming machine then a Z68 with 3 GPUs.

All those 32 lanes in your Z68 still gets shoehorned into 16 lanes. There is no getting around that fact. The "extra" lanes are from PCIe multipliers or whatever they're called. Apparently this board uses an NF200. It basically allows for Tri SLI, but otherwise does not increase overall system PCIe bandwidth, and indeed probably adds a tiny bit of latency.

The 3930k seems a bit more fluidly smooth and does get More FPS at the same clock speeds with 2 cards then the 2600k does with 3. CPU and Chipset does have a lot to do with Utilization of Graphics. I cant explain why it seems more fluid but it does.

Maybe more memory bandwidth? All the textures have to load from system memory and maybe the superior memory bandwidth of X79 helps?

 

[abeeftec]

I have just found out that Futurmark has thrown a monkey wrench in the way you run their Newest Program 3DMarks 11

It used to be that you run their programs at default. And wether you run 1 2 3 or 4 cards it just upped the score.

It appears now they have changed it on 11 to you MUST SET Your amount of Graphics cards. As I said, I dont do much posting in forums because of the immaturity that goes on so after years of running their benchmarks I did not know this.
So my 3Dmarks 11 score is TOTALLY misinformed results wise.

This still does not explain why the 3930k with 2 cards is beating the 2600k with 3 cards in Vantage and 06 Marks

Below is a screen shot of 4.5ghz with the 2600k on Performance settings AFTER I set the number of cards. But the 3930k Results above are skewed with the same issue.

So once I get my board back this week I will redo the runs and mark up the cards and see what happens.

It is pretty frustrating for Futurmarks to change the way they do things after years of running their benchmarks!

Let me just say to those who think that this is not a good comparison.

Running the tests with the same cards would be a more even comparison sure. But the reason here why the 3930k is beating the 2600k with more cards is simply because of how Futuremarks scores I believe. They put a lot of emphasis on COMBINED SCORE and very little on a ton
of graphics scoring. It does not make sense to me why they do that but I have found that your graphics score can be way up but if your combined scoring is down you lose a ton of score.

I have every intention of running the systems with the same amount of cards and posting them as well. I just was a bit perplexed and amazed at the results of Vantage runs and 06 marks runs. I also plan to do gaming benchmarks as well. And may put this Z68 board on the cooler express also! But now I am currently waiting for a board to come in for X79 to resume testing and benching.

@mejoltcola Its actually not still 16 lanes. the NF200 chip gives 16 more lanes.

I dont think it matters if the lanes are connected directly to the chipset or not. If you add a USB chip onboard it
still gives LIKE performance to an on chipset USB controller. I would think that adding PCIe lanes do the same..

To everyone here....
MUCH appreciated that there is a place where the flaming and nonsense is not tolerated. I may have finally found a forum to post in after YEARS of staying away because Forum moderators in other forums LOVE to come fly in and ban the guy being attacked by immature Kids breaking rules themselves. And then they leave them all alone. Used to make me sick.

Thanks for being a decent community!

I do find it strange that checking off 3 GPUs versus not checking anything ONLY affected the combined score. It did not affect the Physics only score OR the graphics score.
This is very strange I think that you must change the GPU setting at all but why would it not affect graphics or Physics if it wasnt noticing 3 GPUs before and now it is?

 

[abeeftec]

OK you r getting very low scores for the 2600k with three 570's here is a video I did with my three machines and comeing soon three way crossfire with 7970's with a 3930k

http://www.youtube.com/watch?v=LyTvazcMC3E&list=UUxhFeDHHW4Eoo4LEVmOPoPw&index=4&feature=plpp_video

Why did Futuremarks say your system results were Low compared to similar systems when you had your cards overclocked like mad and the CPU clocked up to 4.5?
Usually it my score is low until I get over 4Ghz on my CPU and then I get the green plus.

 

[abeeftec]

I'll run some game benchmarks. What do you want to run? You run yours, I'll run mine and post the results. I'll even keep my CPU at a modest 4.3GHz during the benchmarking.

Intel Core i7 3960X @ 4300MHz
Asus P9X79 Deluxe @ 43 x 100MHz
16GB Corsair XMS3 @ 2133 DDR
Three GTX 580 in Tri-SLI @ 850/2200
120GB Vertex 3 SSD
Two 3TB Deskstar 7K3000 in RAID-0 - 6TB
1TB SpinPoint F1
Pioneer BDR-206 BD-RW
Sound Blaster X-Fi HD USB
Cooler Master HAF-X case
Prolimatech Megahalems cooling
Corsair AX1200 power supply


The Synthetics:

Great Vantage score!
Can you post a CPUz with your Benchmarks? The 06 marks score looks a bit low on your benchmark with 3 580s..
Here is my 3930k with only 2 570s at 4.36...

In your sig it appears your cards are overclocked as well? Is that right?
Seems to me that 06 marks isnt very accurate.
Your system with 3 overclocked 580s should be killing my system at nearly the same speed with only 2 570s not overclocked.
Granted my system is clocked a bit higher but only by 60Mhz. My scores didnt raise that much at a 100Mhz increase

I do still have some benches from my system before it blew up from water but I will have to wait before I clock the 3930k again because though I have another board coming it will no be here until mid week or the end of this week. Maybe we should drop 06 marks if its that off.
I have a feeling it doesnt notice the 3rd card. I did a test about a year ago with 2 and 3 GPUs and I think the 3rd GPU scores only very slightly higher. So that would leave us at a closer margin of course with me being 60Mhz faster on the CPU


I noticed you have a Vertex 3 SSD in your System.

Watch this wildness! I had the luck to be able to get a second 120gb Vertex 3 max IOPS drive for 80 bucks!
So I raided them together for this system
Here is the results of that...

WBC enabled

WBC disabled

Disabled it appears the read speed is slighly faster. Does not lose that much speed at all though with the WBC enabled and the write speed is through the roof.

Just wanted to show you the advantage of one more Vertex 3. It used to take 3 or 4 SSDs raided to get that type of increase in speed.

@Faseman..
To your first post.. You are most likely correct. It does seem to give most of the score according to CPU power
And to your second post..
As you can see I didnt post these to dis the Z68. I own one of them myself. It was my experience. And if you knew me, You would know I will stop at nothing benching day and night and testing to make sure that my results are accurate. I said in the bottom of my first post that I just load the systems and run the benchmarks. If they are wrong, It is not my fault nor my intention. And If I find I have made a mistake, I am not to proud to admit and show where my mistake was.

I do this by myself sharing results with only those who purchase systems from me normally because forums have been a nasty experience. Please dont get on me with comments like "My CLAIMED Results" as if I said something with the intention of being misleading.
I read the rules here and I absolutely love that a forum is taking the approach of not tolerating flaming and rude remarks to other people. I dont intend to post to argue. I post for one reason. To give information and also to get information.

 

[Zap]

@mejoltcola Its actually not still 16 lanes. the NF200 chip gives 16 more lanes.

I dont think it matters if the lanes are connected directly to the chipset or not. If you add a USB chip onboard it
still gives LIKE performance to an on chipset USB controller. I would think that adding PCIe lanes do the same..

Was this supposed to be directed at me?

Not sure you are understanding what I'm trying to explain so I'll try an analogy.

16 PCIe lanes is like eating dinner with your one mouth, one hand and one fork. Adding an NF200 is like using your other hand and a second fork. You still only have one mouth and can still only take one bite at a time!

It may add some capabilities, but ultimately you have not changed the bottleneck.

I had the luck to be able to get a second 120gb Vertex 3 max IOPS drive for 80 bucks!

New or used? Either way that's a great price!

I read the rules here and I absolutely love that a forum is taking the approach of not tolerating flaming and rude remarks to other people. I dont intend to post to argue. I post for one reason. To give information and also to get information.

Wow, someone that actually LIKES our rules and don't think them draconian. :thumbsup:

 

[abeeftec]

Was this supposed to be directed at me?

Not sure you are understanding what I'm trying to explain so I'll try an analogy.

16 PCIe lanes is like eating dinner with your one mouth, one hand and one fork. Adding an NF200 is like using your other hand and a second fork. You still only have one mouth and can still only take one bite at a time!

It may add some capabilities, but ultimately you have not changed the bottleneck.

I do understand what you meant I think. And I do believe the bus is the same size either way. I just dont believe there is a limitation that would make a system with 3 cards be at a disadvantage to one with only 2 cards. That would sort of make having the 3rd card pointless and a waste of money.

I do believe that you can only cram so much info down the pipeline but It definately ran better with 3 cards then 2. It scored a higher score with 3 then 2 is what I mean.

My point in the whole thing though(and this may be what you mean) is that the z68 board does not handle the 3 cards the greatest and I do not think it was meant to handle 3 cards like a true express board. I most certainly dont know everything about it. Those are just my thoughts after having ran the board for a year.
It seems they could have given us more lanes on the z68 and they certainly could have at least put an NF200 chip on the upper end X79 boards to get us 4 x16 lanes which Gigabyte has done on their upper end boards until x79.
Just seems to me that they are doing a lot of skimping on the x79 even though it is great to finally have a true express chipset again!
Think of the limitations on the x79 in regards to what upper end X58 boards had. Sort of seems like they are moving backwards to me in features. But who knows!
Maybe Intel had a plan to sell people this series in two different chunks. When I first looked at the features of the X79 it looked like it was going to be a total upgrade of everything across the board. And then all of a sudden some issue came up and everything got stripped to being hardly anymore then the Z68 chipset with the exception of the 40 lanes and quad memory channel. Maybe they are planning to give all that other stuff they cut on the revision. I hear that PCIe 3.0 isnt even in the current 3000 series CPUs.

New or used? Either way that's a great price!

That was brand new from Newegg! I was able to finagle the price using a few codes and it was on sale with a rebate and I had a 30$ Gift certificate. Before i knew it I had it for 80.00! Of course the way my luck had it I had to wait an additional week to use it in raid0 because my first drive broke the day before this one came in and I had to RMA it. Luckily OCZ had it back to me within a week!

Wow, someone that actually LIKES our rules and don't think them draconian. :thumbsup:

Man I cant stand forums for the most part. It seems everyone is trying to be cool at others expense all the time. In every forum dont say something against the mainstream because that will get you hunted by people!

EDIT: took out a part about another post in another forum. Not necessary.

Sorry for long posts. Thats another thing, Everyone acts like if your post is long then your a troll! Seems lazy to complain about something like a long post that actually explains so no one has to guess what you mean.

 

[AdamK47]

In your sig it appears your cards are overclocked as well? Is that right?
Seems to me that 06 marks isnt very accurate.
Your system with 3 overclocked 580s should be killing my system at nearly the same speed with only 2 570s not overclocked.
Granted my system is clocked a bit higher but only by 60Mhz. My scores didnt raise that much at a 100Mhz increase

Yes, all three are at 850/2200. I guess my system just sucks at 3DMark 06.

 

[mrjoltcola]

Man I cant stand forums for the most part. It seems everyone is trying to be cool at others expense all the time. In every forum dont say something against the mainstream because that will get you hunted by people!

Its an indicator of:
1) The maturity level of the people on the forum.
2) What sort of example the site admins / moderators set.
3) What sort of moderation, if any, exists on the site.

If the site has [] in its name, I avoid it nowadays. Not many reviews can hold a candle to Anand's anyway.

 

[abeeftec]

Yes, all three are at 850/2200. I guess my system just sucks at 3DMark 06.

LOL, I had to laugh when I saw this comment!
I am fairly sure its only seeing 2 and a 5th of a card.
Plus, Do you notice an improvement in overclocking the cards?
I overclocked mine to 822 and my score got lower in the next run. But those tests can fluctuate hundreds of points per run.
You should try taking a card out to see if its just not noticing the card.
Plus, my system was 60mhz faster on my results. That and a couple
points here or there per run and they would most likely be pretty even with my CPU being a bit faster when ran.

 

[RussianSensation]

EDIT: Please dont hijack my thread with arguements about why I am wrong here. If anything is wrong it is not me. I simply loaded the systems and ran the benchmarks. I cannot tell the benchmark what numbers to throw out when they are completed

Using 3DMark 06 to show that 3930 with 2 GTX570s is faster than a 2600k system is pointless since 3dMark06 is a very CPU limited, and not just a GPU limited benchmark. Also, 3dMark 06 performance has nothing to do with real world gaming performance. So to begin with you have chosen a poor benchmark to prove your point.

Moving on to the GPU limited 3DMark 2011. In the gaming benchmarks, which are the only ones that matter, the GT1-GT4 scores are way faster on 3x GTX570 GPUs, which indicates that a 2600k system with 3x GTX570s is faster than a 3930 with 2x GTX570s in game engine code that 3dMark 2011 uses (and this is what was supposed to happen anyway). The overall score is meaningless since it's adding the CPU scores (Physics) and the Combined score, making it useless for measuring overall performance in games.

So if you are only building a system strictly for games, 3x GTX570s on Z68 even with PCIe being limited to 8x/8x/8x is going to stomp all over the X79 platform with 2 of the same GPUs.

It's already been shown 100x that PCIe 2.0 8x is NOT a limitation for graphics performance by more than 2-3%.

Your assertion that 3930 with 2 GPUs is better than 2600k with 3 GPUs (of similar performance level) is incorrect and even your own benchmarks in 3dMark 11 prove this.

My Z68X-UD7 has 2x16 lanes(one of the ONLY Z68 boards with 32 lanes) and with 3 GPUs is running in X16 X8 X8 whereas my X79 with 2 GPUs is running in X16 X16.I do think that there is no way to get around the fact that the 3930k with 2 GPUs is a better gaming machine then a Z68 with 3 GPUs.

Even if we hypothetically assume that a 3rd GTX570 on the Z68 platform is 100% bottlenecked (i.e., it's adding 0% to the performance of the platform), you would need to prove that 3930 is actually faster in games than a 2600k.....Let's analyze this claim.

Most people who upgraded to 3930/3960 platform and saw very good performance increase in games with 2-3x GPUs were upgrading from LGA1366 platform and not from a Sandry Bridge platform.

But if you look at 2600k vs. 980X/990X, the quad-core SB platform was already a faster gaming platform vs. a 6 core LGA1366 to begin with, while when comparing the 2 sandy bridge platforms, the 3900 series do not have a performance advantage in games over LGA1155.

There is no evidence to support the view that even a single PC game uses more than 4 + HT that 2600k/2700k offer.
There is no evidence to support the view that PCIe 8x botteneck is severe against 16x.
There is no evidence to support the view that 3930/3960's CPU performance advantage is so significant in games as to offset adding a 3rd GPU to the LGA1155 platform.

Since we already know that Z68 board with an NF200 chip does not have a significant PCIe bottleneck for running even 3x GTX580s in SLI, the only way for your argument to be true is if the CPU advantage of the 3930K is MORE than the performance advantage of adding a 3rd GTX570 to a 2600k platform with 2x GTX570s. If you load Task Manager, you might see that 8 or 10 or even 12 cores being used a little bit in a game. That does not mean that a game actually uses more than 4 threads. If you were to load the same game on a 2600k, it would just have a higher utilization in those 8 windows, but that doesnt' mean it's even using up 100% of the 2600k CPU. In fact, there isn't a single game on the PC platform that needs more than 4 CPU cores + HT.

Today, the largest CPU limitation lies in performance per clock (IPC) and CPU clock frequency. This is why in games a 5.0ghz 2500k would easily crush a stock clocked 3930/3960, while 3900 series have no performance advantages in games over quad core SB CPUs.

Please note that Reviews are subjective to the hardware they are written about. They are Not the law when different variables are in play. I see many people posting information based soley on reviews. The 3930k seems a bit more fluidly smooth and does get More FPS at the same clock speeds with 2 cards then the 2600k does with 3. CPU and Chipset does have a lot to do with Utilization of Graphics. I cant explain why it seems more fluid but it does.

Show gaming benchmarks to prove this. Just because you "feel" something is faster, doesn't make it true. You could "feel" it to be faster because that's what you may want to see. Reviews show #s we can compare. Provide #s for some games such as Battlefield 3, Witcher 2, Dragon Age 2, Crysis 2, F1 2010/2011, etc. and then we can see if your argument is credible.