3930k vs 3960x Whats the difference???

[azed3000]

I did a bit of research and couldnt find any substantial difference in performance between these two processors, their overclocks and everything are almost identical. So whats the point of spending 500 dollars extra on a 3960x if the 3920k does the exact same thing?? Apart from the unlocked multiplier, what else is there?? They both top out at 4.7ghz without resulting to extreme cooling methods, they both bench around the same area, both support same chipset, cant seem to find anything more superior about the 3960x. Am i missing out on something or...?

 

[sarinaide]

Nothing but a lot of money.

If I am to advise you, I would forget about socket 2011 its a novalty that wears off a few days after purchase, and hits your bank account hard, factor in resale on ebay, amazon and the works will see you lose half value within days should you choose to sell it. In all Socket 2011 is only for the money is not a option select, if you are then by all means, if you value your money and what you get then I can suggest looking elsewhere.

 

[Kamen_BG]

They both have unlocked multipliers so there is no difference there.The only real difference between them is 3MB L3 cache.But since they've already got quite a lot of cache this doesn's bring any benefit.
The other differences are that 3960K's are better binned. That means that they can usualy work with a bit more voltage.Therefore they can potentialy clock a bit higher (~20 mhz).
And the most important thing is that you can brag that you've got the best chip out there.To some people that might actually mean something but to me... it doesn't.

 

[sarinaide]

Yip, epeen status. Its bad when a ivybridge setup is enough to show up just how bad Socket 2011 is, this wasn't the case with 1366 where my 990X blew away all comers.

 

[wr6133]

Any socket 2011 setup is a waste for 99.9999999% of home users. Unless you have a professional need for a workstation to be used in a production environment (and then there are argueably better options) its all wannabe e-peen.

 

[sarinaide]

I can get a SR-x with Xeons for less and they can do any productivity based application faster, and its all still cheaper.

SB-e is pointless, its like choosing between a heated butter knife or steak knife and persisting to cut the same block of butter.

 

[azed3000]

Yea im not planning on buying a 2011 socket, just wanted to know why theres a 500 dollar price range difference between the two, i thought their might be some legitimate reason i may have missed, clearly i was right in thinking its just stupid marketing.

If i were to upgrade, i would just chuck in a i7 970 (upgraded from my 950) and run that at 5.2ghz stable on normal high end cooling methods, something the 2011 chipset cant do for some reason. 5.2ghz i7 970 is equivilant to a i7 3930k at 4.7. I searched up some youtube videos before on i7 3930/3960s and people topped out at 4.7 to 4.8ghz stable, while a lot of other people got their 2600ks and gulftown processors much higher (5ghz+). Ivy bridge surprisingly tops out at 4.7 as well, no clue why. Sucha useless upgrade from 1366 and 1155 in my eyes. http://www.youtube.com/watch?v=c6zUE-dlPfs

 

[Kamen_BG]

Actuallt core i7 970's don't clock that high.They topped out at around 4.4 GHZ if i remember correctly and they are a bit slower per clock than Sandy Bridge-E.
And not all LGA 2011 processors are pointless.It's just the six cores.The Core i7 3820 is arguably a bit worse than both the 2600K and 3770K but it provides a very good upgrade path and LGA 2011 is just a higher end platform than LGA 1155 (2.5 times the PCI-E lanes!)

 

[sarinaide]

more lanes doesn't make it better, some Z77 boards can hold the same highest end possible 3-4 way setups as a LGA2011. The only use for a LGA2011 is high levels of number crunching like stock trading even then a Xeon is better.

 

[azed3000]

Actuallt core i7 970's don't clock that high.They topped out at around 4.4 GHZ if i remember correctly and they are a bit slower per clock than Sandy Bridge-E.
And not all LGA 2011 processors are pointless.It's just the six cores.The Core i7 3820 is arguably a bit worse than both the 2600K and 3770K but it provides a very good upgrade path and LGA 2011 is just a higher end platform than LGA 1155 (2.5 times the PCI-E lanes!)

Gulftown processors are all capable of 5ghz + i7 970 is just a 990x but with a limited multiplier, just like the 3930k.

http://www.youtube.com/watch?v=f5n6E1PWHA4
http://www.youtube.com/watch?v=klPXYTCA5yI

Videos show i7 970 running at 5ghz with a mainstream air cooler, not even water cooled and still hits 5ghz, 2011 cant do that.

 

[bemused_fred]

Literally the only differences between the two are a 100MHz higher clock (on an unlocked processor), a 15MB L3 cache as opposed to a 12MB one and the 3960X allows for 64GB of RAM, unlike the £930x.

 

[spookyman]

Just to rip you off.

Actually both processors will allow for 64 gigs of RAM. I have a 3930k and it can use 64 gigs.

Sandy Bridge E is great for workstations running multi-threaded apps. Also the available 40 PCI-E channels can be handy too with Graphic cards. It was designed as a workstation processor. If I was a gamer, no its a waste of money. Still its cheaper then the Sandy Bridge E Xeon based processors.

 

[azed3000]

Yea thats what i thought, for extra 500 dollars theres very limited difference between the two. I just wanted to find out the actual truth between the two, and why there was such a huge price difference (more than double). The way intel releases their processors, chances are their next gen mainstream cpu would match or even demolish a 3930 or 3960x, just like the $400 i7 920 did with the $1000 intel core 2 extreme.

 

[azed3000]

Best answer selected by azed3000.

 

[cjl]

Gulftown processors are all capable of 5ghz + i7 970 is just a 990x but with a limited multiplier, just like the 3930k.

http://www.youtube.com/watch?v=f5n6E1PWHA4
http://www.youtube.com/watch?v=klPXYTCA5yI

Videos show i7 970 running at 5ghz with a mainstream air cooler, not even water cooled and still hits 5ghz, 2011 cant do that.

That's a 32nm processor with 1.5+ volts running through it - regardless of cooling, that's a sure way to brick your CPU. Sure, you can boot up at that speed briefly for bragging rights, but that's way, way above the safe voltage for that processor. Realistically, you would be looking at more like 4-4.4 GHz depending on how lucky you got with your chip's VID and leakage.

As for LGA2011? I'm not willing to put 1.5 volts through my chip to see, but I can hit 4.4 GHz at 1.23V, so I wouldn't be all that surprised if my chip could hit 5GHz if I crammed that much voltage into it.

 

[azed3000]

Actually a few guys ran them stable at 4.8ghz as well, 4.4 was the max for the bloomfield 45nm processors, gulftown being a hexacore 32nm processor didnt produce as much heat can could be pushed much harder. Its all over the place, google it =). Keep in mind i7 970 is technically the same as a 980x and a 990x which all can hit 5ghz.

 

[cjl]

Actually a few guys ran them stable at 4.8ghz as well, 4.4 was the max for the bloomfield 45nm processors, gulftown being a hexacore 32nm processor didnt produce as much heat can could be pushed much harder. Its all over the place, google it =). Keep in mind i7 970 is technically the same as a 980x and a 990x which all can hit 5ghz.

Sure, they're stable, but at what voltage? On 32nm, anything higher than about 1.35-1.4 volts or so can degrade the processor, so even if it's stable, over time (which could be weeks or months), it will degrade to the point where it may no longer even be stable at stock speed. I really haven't seen much evidence for stable, 24/7 overclocks on reasonable voltage that were much higher than about 4.4.

As for it being the same as a 980x or 990x? Sure, it's the same die, but the 980 and 990 are higher binned components, and even with the higher binning, they still wouldn't run 5GHz on reasonable voltage unless you got stupendously lucky with your chip. They would absolutely run 5GHz stable for benching, but 24/7 overclocks require a bit more care with the settings.

 

[azed3000]

Sure, they're stable, but at what voltage? On 32nm, anything higher than about 1.35-1.4 volts or so can degrade the processor, so even if it's stable, over time (which could be weeks or months), it will degrade to the point where it may no longer even be stable at stock speed. I really haven't seen much evidence for stable, 24/7 overclocks on reasonable voltage that were much higher than about 4.4.

As for it being the same as a 980x or 990x? Sure, it's the same die, but the 980 and 990 are higher binned components, and even with the higher binning, they still wouldn't run 5GHz on reasonable voltage unless you got stupendously lucky with your chip. They would absolutely run 5GHz stable for benching, but 24/7 overclocks require a bit more care with the settings.

Yea thats true, but with proper cooling, 1.4volts isnt that harsh on the cpu. Ive been running my 950 at 1.45volts @4.5ghz for 1 year straight without a problem, running it on a corsair h80. The die degrades due to high temps which is a result from the high volts(just like you said). But yea youre right in saying its not stable for 24/7, 4.8 is pretty close to stable though.

 

[cjl]

Yea thats true, but with proper cooling, 1.4volts isnt that harsh on the cpu. Ive been running my 950 at 1.45volts @4.5ghz for 1 year straight without a problem, running it on a corsair h80. The die degrades due to high temps which is a result from the high volts(just like you said). But yea youre right in saying its not stable for 24/7, 4.8 is pretty close to stable though.

A 950 is 45nm though, not 32nm, and as a result, it's more resistant to high voltages. In addition, although high temperatures exacerbate the degradation, voltage alone can ruin a CPU, even if the temperatures are completely reasonable. 1.5V on a 32nm CPU is sufficiently high to cause such degradation, even if the cooling is good.

 

[azed3000]

A 950 is 45nm though, not 32nm, and as a result, it's more resistant to high voltages. In addition, although high temperatures exacerbate the degradation, voltage alone can ruin a CPU, even if the temperatures are completely reasonable. 1.5V on a 32nm CPU is sufficiently high to cause such degradation, even if the cooling is good.

Really?? i never knew that =/ Always thought it was heat resulting from high voltages that degrades a cpu. Then i dont understand the point of going from 45nm to 32nm processors? why dont they just manufactuer 45nm processors for the ivy bridge etc, since they suck at high voltage, wouldnt that make it more overclockable?? never actually understood the whole nm architecture stuff.

 

[cjl]

Really?? i never knew that =/ Always thought it was heat resulting from high voltages that degrades a cpu. Then i dont understand the point of going from 45nm to 32nm processors? why dont they just manufactuer 45nm processors for the ivy bridge etc, since they suck at high voltage, wouldnt that make it more overclockable?? never actually understood the whole nm architecture stuff.

The reason why shrinking the process node is a good thing (i.e. going from 45nm to 32nm) is twofold:

1) A smaller transistor can run the same speed at a lower voltage. This saves power. For example, my 3960x can run 4.4GHz at 1.23 volts (and I can run 4.5GHz at around 1.275 or so). Your 950 requires 1.45 volts for the same speed, even though it has 2 fewer cores. That is largely because of the smaller process node. A 65nm processor (e.g. Core 2 Quad Q6600) can't even reach 4.4 GHz in the vast majority of cases, and takes substantially more voltage at lower speeds (but can safely be run at 1.5-1.6 volts, due to the larger process node).

2) A smaller transistor means more of them can fit in the same die area. Since manufacturing cost is roughly proportional to die area, this means that more transistors can be used at a given price point. This is why a processor with 4 cores and a built in GPU on 32nm or 22nm (Sandy or Ivy bridge) costs the same as a dual core without even a built in IMC on 65nm (Core 2 Duo). Specifically, a 4c Sandy Bridge die with graphics is 216mm^2 with 995 million transistors, a 4c Ivy Bridge with GT2 graphics is a diminutive 160mm^2 with 1.4 billion transistors. An old Conroe Core 2 Duo had a similar die size to Ivy Bridge (143mm^2), but it only had 291 million transistors.

To put this another way, due to the process shrink, an Ivy Bridge quad manages to fit 4 cores, a memory controller, a PCIE controller, and a decent low end GPU in only 12% more space than a Core 2 Duo took to fit two cores and a front side bus, and it uses comparable power too (despite a significantly higher clockspeed).

Basically, a process shrink means you can do one of two things:

1) You can make effectively the same processor, except it costs less and uses less power
2) You can make a more powerful processor that costs the same as the old one, and doesn't use any more power

It seems like a die shrink is a pretty good idea to me, as long as you're careful about the limits of each new process.

 

[azed3000]

The reason why shrinking the process node is a good thing (i.e. going from 45nm to 32nm) is twofold:

1) A smaller transistor can run the same speed at a lower voltage. This saves power. For example, my 3960x can run 4.4GHz at 1.23 volts (and I can run 4.5GHz at around 1.275 or so). Your 950 requires 1.45 volts for the same speed, even though it has 2 fewer cores. That is largely because of the smaller process node. A 65nm processor (e.g. Core 2 Quad Q6600) can't even reach 4.4 GHz in the vast majority of cases, and takes substantially more voltage at lower speeds (but can safely be run at 1.5-1.6 volts, due to the larger process node).

2) A smaller transistor means more of them can fit in the same die area. Since manufacturing cost is roughly proportional to die area, this means that more transistors can be used at a given price point. This is why a processor with 4 cores and a built in GPU on 32nm or 22nm (Sandy or Ivy bridge) costs the same as a dual core without even a built in IMC on 65nm (Core 2 Duo). Specifically, a 4c Sandy Bridge die with graphics is 216mm^2 with 995 million transistors, a 4c Ivy Bridge with GT2 graphics is a diminutive 160mm^2 with 1.4 billion transistors. An old Conroe Core 2 Duo had a similar die size to Ivy Bridge (143mm^2), but it only had 291 million transistors.

To put this another way, due to the process shrink, an Ivy Bridge quad manages to fit 4 cores, a memory controller, a PCIE controller, and a decent low end GPU in only 12% more space than a Core 2 Duo took to fit two cores and a front side bus, and it uses comparable power too (despite a significantly higher clockspeed).

Basically, a process shrink means you can do one of two things:

1) You can make effectively the same processor, except it costs less and uses less power
2) You can make a more powerful processor that costs the same as the old one, and doesn't use any more power

It seems like a die shrink is a pretty good idea to me, as long as you're careful about the limits of each new process.

Wow that actually cleared things up alot. Makes so much more sense now, i searched that up on wiki a while back and i didnt understand what it meant, you're explanation was much better =p. That would also explain why laptops with 32nm processors run much more better than the 1st gen i7 in laptops, most of them had heating problems due to the cpu generating too much heat from all 4 cores. From what ive seen, the lower nm processors are capable of having a higher stock clock than the higher nm processors, for e.g. my 2670qm runs 2.2ghz (turbo boost disabled) while my friends 1st gen i7 740 clocks at 1.73ghz althought it was capable of much more, because the higher clocks generated too much heat.