Architecture And Performance
Ivy Bridge-E and the Core i7 4960X brings one key advantage over the Core i7 3960X and that is that it shrinks the manufacturing process size down from 32nm to 22nm.
The Intel Core i7 4960X boasts an impressive 1.86 billion transistors with 15MB of cache shared across all six cores. The Core i7 4960X is a native six core which means there are no extra 2 cores that have been disabled like on the Core i7 3960X which was an 8 die CPU.
The Core i7 4960X has 1.86 billion transistors according to Intel which is strange as this is less than the Core i7 3960X which had 2.27 billion. Maybe the removal of those two extra cores have something to do with that.
If you’re wondering what the die lay-out actually means, well here you can see what part is what. You have six fully fledged cores hooked up to 15MB of total shared L3 cache. The Core i7 4820K and Core i7 4930K will have some of this cache disabled but all three CPUs share the improved I/O and memory controller.
What can you expect in terms of performance? Well Intel has some information for us to share with you before we even delve into benchmarks. Apparently compute performance is up 5% over the Core i7 3960X.
In other things outside of Compute we should expect to see in the region of 4-10% more performance, which is roughly typical of the generational improvements Ivy Bridge gave over Sandy Bridge on LGA 1155.
Compared to the current Intel Core i7 4770K Intel claim the Core i7 4960X is faster in gaming, 3D modelling and data/finance work by quite some margin. The Core i7 4960X does apparently fall behind by around 18% in everyday usage due to the lower single threaded performance of the Ivy Bridge architecture compared to Haswell.
In terms of how Ivy Bridge-E compare on the raw details we can see the specifications below. As mentioned the models below the Core i7 4960X have reduced cache. One thing to note is that all Ivy Bridge-E CPUs have higher clock speeds than the CPUs from the LGA 2011 Sandy Bridge-E generation. We can see the much higher price points and TDPs compared to the Core i7 4770K too.
As you say it’s something that an everyday user doesn’t need and those who do buy them, will all be about willy waving. The price of the chip coupled with the archaic X79 chipset renders it redundant anyway.
I agree to an extent, the lack of SATA III and USB 3.0 isn’t ideal. But then PCIe goes through the CPU so thats up to date. Board vendors can add marvell/asmedia SATA III/USB 3.0 implementations so its not really redundant.
It is useable but for the price you’re expected to pay, it really requires a new chipset. Marvell & Asmedia chips are just stopgaps at best.
For the price you’re expected to pay it should come with a Butler who sets the new CPU up for you and provides you with free ice cold beer.
That being said, It could be a cheaper, more cost effective alternative to Intel’s Xeon’s.
Why use a $200 AMD cpu against a $1000 intel one? why not use a 9590? or even a 12 core socket G34 AMD cpu for similar price?
The FX-8350 at 4.8GHz is better than an FX-9590. Why? Because the FX-9590 is 4.7GHz base clock and 5GHz turbo. It only goes to 5GHz turbo on one module, maybe you’ll get two if you’re lucky. Most of the time when all four modules are used the clock speed is 4.7GHz. As far as the socket G34 goes they are industrial/business grade CPUs. Not meant for consumers, it would be a pointless venture. Plus we’ve already established that anything more than four cores is barely utilised except in productivity. So 12 cores at 2.3GHz (on the top of the line Opteron 12 core model) will probably fare worse than 8 cores at 4.8GHz IMO.