Intel Ivy Bridge-E Core i7-4960X IPC and OCing Review @ [H]

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,534
Intel Ivy Bridge-E Core i7-4960X IPC and OCing Review - We debut Intel's next $1000 Extreme Desktop processor, the Core i7-4960X, this time with Ivy Bridge architecture and a couple of extra cores thrown in for good measure. It is a beast of a CPU for those that can actually harness its power and bandwidth, but how much better is it than Sandy Bridge-E and Haswell at the same clocks?

Edit!!!

Added full size die shots at this link, and this link. About 8MB each.
 
Last edited:
Nothing that special then compared to SB-E, might be worth upgrading from my 3820 to 4930K.
 
Well I am still on my 930, might get this as a christmas present to my self when I finally overhaul my whole system.
 
So it overclocks worse than Sandy Bridge-E.
I think a typical overclock with 3930K is 4.8GHZ, doesn't it mean 3930K is better than 4930K? I know 4930K has a bit higher IPC but still.

Wont Asus release Rampage V Extreme soon?
 
So lower temps than Haswell but no real improvements to overclocking (maybe 100 - 200 MHz) and lower IPC.

22nm chips just don't like high clock speeds, regardless of whether they sit under solder or TIM.
 
My 3930k is running at 4.9. It feels like i've had this thing forever and i find it incredible that there still isn't something worth upgrading to. Not sure if that's good or bad ;p
 
I think we will see very few IVBE do 4.8GHz on water, very few, on pretty much any board. I think about 75% of product in the first launch stepping will do a solid 4.6GHz if you pour at least 1.4v to it. I feel like the IMC on the IVBE will almost all give a solid 2133MHz with rated DIMMs, but we will not see those pop 2400MHz like we did on Haswell, at least not very often. VCCSA facilitates memory frequency, so push it up to 1.4v if you are loading all eight memory banks. But with the matched sticks I had we could push 8x4GB on the X79-Deluxe. FWIW also, I had good success using the 4 Way Optimization with AI Suite III. That worked as well as I could do manually when pushing the BCLK up.

Have had no sleep, hope that made sense. Will check this post again later.
 
22nm chips just don't like high clock speeds, regardless of whether they sit under solder or TIM.

That is the most disappointing part to me. It seems the 22nm tri-gate silicon is better than the 32nm process at stock clocks but not any better at all when overclocking. And Intel ~17 months this is not fixed at all.
 
I always look forward to Kyle's reviews with every new CPU launch. However, like the last past 5 years, I can't say I'm all that thrilled with the incremental improvements Intel has been passing off as progress. Without benchmarks, no one would even even notice any improvement from SB-E. Overclocking seems the usual disappointment as well.
 
That is the most disappointing part to me. It seems the 22nm tri-gate silicon is better than the 32nm process at stock clocks but not any better at all when overclocking. And Intel ~17 months this is not fixed at all.

I am not sure "fixed" is the issue here. I am not so sure they Intel is in a position that it wants to spend the resources to push those desktop processors any higher in terms of clock speed.
 
Good review of a piss poor processor. Very few advantages over Sandy Bridge-E. The lack of overclocking prowess means Sandy Bridge-E in most cases will be able to out-overclock Ivy Bridge-E and this fact alone should overcome the slight IPC advantage that Ivy Bridge-E has. PCI Express 3.0 is a minor thing, and the Xeon E5-1600 equivalents to the i7s have had official support all along!! This strongly suggests an artificial and intentional limitiation by Intel so that they had something they could market as an advantage for Ivy Bridge-E.

What it boils down to is that Ivy Bridge-E is a waste of time for the HEDT crowd. The only reason we're seeing Ivy Bridge-E at all is because it does offer tangible benefits to Enterprise customers in the form of greater efficiency and greater core density. While Ivy Bridge-E is a native 6 core die, don't doubt for one minute that Intel wouldn't hesitate to use defective 10 and 12 core Xeon dies that have 4 or more cores (or the associated cache) defective to get rid of them. If Xeon yields are good, Intel will simply run more 6 core wafers to make up for the lack of defective Xeon dies.

But, given the near total lack of improvement of Ivy Bridge-E for HEDT users, I don't think Intel will have to worry about running many 6 core wafers as I predict demand for them will be tepid at best.;)
 
Kyle really needs to not bottle up his thought process, and tell us what he actually thinks. :)

Kyle Bennett said:
But let me say this, while Intel has been beating the drum about this being the "Ultimate Desktop Processor for Gamers," I think that is a lot of horse shit. This Sandy Bridge E is not going to do much anything for gamers if I am making the right guess based on what I have seen, possibly with one exception, and that is multi-GPU, multi-display gaming.
 
My master plan has worked.

Buy X79 + 3820... hold out for IB-E... and there's the 4930K, just as better-than-the-3930k I was expecting.

New CPU will be procured soon.
 
My master plan has worked.

Buy X79 + 3820... hold out for IB-E... and there's the 4930K, just as better-than-the-3930k I was expecting.

New CPU will be procured soon.

Pick up a Xeon E5-1650 instead...should overclock better than the i7-4930k and it has PCI-E 3.0 as well...;)
 
Still rocking a Rampage IV Formula and 3930K......

Mine will run at 4.5 GHz all day long.......with the exception of the extra PCI-e lanes I guess I'm good for a while more.

I guess any upgrades are going to be cosmetic, like one of those new Enthoo Primo cases and some of that bendable Primochill tubing.:D

I enjoyed reading the article, I was hoping to put one of these in my RIVF but I think I'll pocket the 1K and buy something else.
 
Last edited:
Sad, I had hopes of seeing at least a 4.7-4.8Ghz o/c on the 6 core parts..I wanted to replace my 3770K/Z77 combo but now am at a loss..I wonder what the 4820Ks are going to end up doing..
 
Kyle,

not quite 5 Ghz but i'm on a 4.2 Ghz Gulftown and this is still meh!!!

Processors are so boring now.

I will have to wait for haswell E it looks like.
 
Kyle,

After reading the review of this processor - Great job as usual, by the way - and looking at the $1,000 price tag, it got me thinking - Has there ever been a comparison of the extreme processors across the generations? What I'm actually interested in, is if I bought an extreme processor a few generations ago, would it still be relevant as a CPU today, and at what price-point of a current generation CPU would it compete at. Also, if I went and spent $1000 on the new latest and greatest, what value might I get out of it when it comes to this "relative performance depreciation".

Thanks!
 
Kyle,

not quite 5 Ghz but i'm on a 4.2 Ghz Gulftown and this is still meh!!!

Processors are so boring now.

I will have to wait for haswell E it looks like.

My uneducated guess is that Haswell-E won't offer massive performances over the previous generation either (single digit % per core), but merely add two cores. Maybe that's just my cynicism from being disappointed over and over again.
 
Kyle,

After reading the review of this processor - Great job as usual, by the way - and looking at the $1,000 price tag, it got me thinking - Has there ever been a comparison of the extreme processors across the generations? What I'm actually interested in, is if I bought an extreme processor a few generations ago, would it still be relevant as a CPU today, and at what price-point of a current generation CPU would it compete at. Also, if I went and spent $1000 on the new latest and greatest, what value might I get out of it when it comes to this "relative performance depreciation".

Thanks!

Well, I sort of touched on this subject in the conclusion as it pertains overclocking these parts and where I thought a solidly overclocked 1366 part is still relevant. Also it truly depends on what you are doing with it application-wise. But I think as you look across the last few generations of architecture, it still really just comes down to what processor clocks you are running....and again what software you are wanting to use. Clock is still king.
 
Interesting write up. Without a true high end competitor, Intel doesn't have to get really innovative these days. I will say that I am super happy with my recent upgrade to a Haswell processor. Since I don't do major encoding other than an occasional vacation movie, this processor is more than adequate for my gaming needs.
 
Kyle:

So just like with the switch from regular Sandy to Ivy, Sandy-E to Ivy-E gives an IPC improvement that is almost exactly offset by a reduced ability to overclock. Sandy to Ivy was 5.0GHz down to 4.8GHz and Sandy-E (on water) is 4.8 down to 4.6 on Ivy-E. At least when regular Sandy transitioned to Ivy all the extra space on the die was taken up with more iGPU "cores". Beyond a 20-50 watt power saving, what is the actual benefit of the new chip?

Assuming one does not believe Intel that the new Ivy-E are all purpose made dies rather than gimped Xeons, the only real change on the die from Sandy-E to I-E is a 266MHz bump. Considering how much larger Ivy-E chips are than Ivy... what is taking up all that space?!
 
So, basically, no surprises. I'm still looking forward to that 4820k: more I/O than Haswell, no iGPU to deal with, will be a bit cheaper than the 4770k and have more cache than the Haswell chip. I can hardly wait to see some benchmarks from it. I hope it overclocks better than the 4930k. The 6 cores aren't too attractive due to price.
 
Kyle:

So just like with the switch from regular Sandy to Ivy, Sandy-E to Ivy-E gives an IPC improvement that is almost exactly offset by a reduced ability to overclock.

That is exactly why we do the IPC comparisons as explained in the article. And yes, you are 100% correct.
 
Can't read the review again... guess I'll have to wait until I go home because IE7 won't even open the review side of the site.


So it overclocks worse than Sandy Bridge-E.
I think a typical overclock with 3930K is 4.8GHZ, doesn't it mean 3930K is better than 4930K? I know 4930K has a bit higher IPC but still.

Wont Asus release Rampage V Extreme soon?

Man, what 3930k have you been getting? Average chips will do 4.5-4.6 at 1.45V, a really good chip will do 4.8 and a golden one might do 5.0+. I'm on chip #5 and it seems to be somewhere between very good and golden but my motherboard can't handle it past 4.7ghz at 1.400V.

Interestingly all 3 3820 that I've had did 5.0+ easily. I'm curious how the 4820k will do.
 
So when Haswell-E would appear it will overclock to 4.2 GHz and would have AVX2. The only difference from mainstream HW would be DDR4 RAM support, which when we look at comparisons of DDR3 and GDDR5 would run at 0.8 efficiency of DDR3, but of course at much higher clock. And 2 to 4 more cores.
 
So, basically, no surprises. I'm still looking forward to that 4820k: more I/O than Haswell, no iGPU to deal with, will be a bit cheaper than the 4770k and have more cache than the Haswell chip. I can hardly wait to see some benchmarks from it. I hope it overclocks better than the 4930k. The 6 cores aren't too attractive due to price.

if your main purpose is gaming than haswell is a much better option right now. if the 4820K overclocks like the 4930 than it won't keep up with haswell and outside of 3/4 way SLI/CF, Z87 is a better chipset than X79 still. you're better off waiting for X99 and Haswell-E.
 
5+ at what voltage?

The worst one was stable at 1.51V, but the other two did it with 1.475V. I do know people that have them that can't get over 4.6-4.8 though, I think I just had good chips. However, I haven't had any 5.0+ SB/IB and I've tried many of them. Anyway, my point was I hope the unlocked quad will overclock nicely.
 
So, basically, no surprises. I'm still looking forward to that 4820k: more I/O than Haswell, no iGPU to deal with, will be a bit cheaper than the 4770k and have more cache than the Haswell chip. I can hardly wait to see some benchmarks from it. I hope it overclocks better than the 4930k. The 6 cores aren't too attractive due to price.

This is pretty much where I'm at now too. As a gamer on an i7 930 that is considering adding a 3rd GTX 780 some time in the next 6 months or so I'm hesitant to go Haswell, but pretty disapointed with the performance here. If Haswell offered more PCI-e lanes the decision would be easy. I hope the 4820K OCs a little better than the 49x0.
 
Kyle:

I forgot to ask, but what is the TJMax for these chips? Is it back to 100C or is it still 91C like with the 39xxK chips? Considering how quickly Sandy-E and regular Ivy heat up, that extra 9C may really end up mattering for some users

Lig:

Most of the 3930Ks that I have seen can run at 4.8, but need to be on a full water loop to do so because of the need to run at 1.42-1.47v to hit this speed.
 
Thanks for the review, guess I'll be keeping my 3930K for a while. Might I suggest testing CPUs with BF3 in the future? Supposedly it uses 8 threads, and in fact in my non-scientific testing when I ran it on my 3930K with affinity set to 0-3 it ran much worse than when the affinity was 0-5 even at 1680x1050. Also probably a more relevant game than lost planet. :)
 
Kyle:

I forgot to ask, but what is the TJMax for these chips? Is it back to 100C or is it still 91C like with the 39xxK chips? Considering how quickly Sandy-E and regular Ivy heat up, that extra 9C may really end up mattering for some users

Lig:

Most of the 3930Ks that I have seen can run at 4.8, but need to be on a full water loop to do so because of the need to run at 1.42-1.47v to hit this speed.

TJMax = 95C.

I had no issues keeping this one cool, but I have heard about others getting higher temps than mine, but just in the low 80s. Have not pushed it over 1.43v at this point though.
 
I put up two full resolution die shots of the Ivy Bridge-E in the OP.
 
Back
Top