Intel Core i7-6850K 'Broadwell-E' CPU benchmarked, compared against 5820K Haswell-E @ 4.2GHz

[Shawn Knight]

Intel’s next enthusiast-grade line of processors, dubbed Broadwell-E, is due out sometime this year. Four chips are expected – the Core i7-6800K, i7-6850K, Core i7-6900K and Core i7-6950X – and thanks to a user from the Overclock.net forums, we’ve got an early look at what the 6850K is capable of.

A user by the name of Maintenance Bot has posted photos and benchmarks of the 14-nanometer Broadwell-E part which features six cores and 12 logical threads. The engineering sample is clocked at 3.60GHz by default although for the purpose of testing, it was overclocked to 4.2GHz to match the speed of the 5820K it was compared against.

Also see: Intel 'Kaby Lake' Core i7-7700K CPU details leaked in benchmark results

Before jumping into testing, Maintenance Bot measured the two chips and found the newer model is just 1.12mm thick versus the chunky 5820K at 1.87mm. Eagle-eyed viewers may also notice a new IHS (integrated heat spreader).

Running a 32M calculation in Super Pi mod1.5 XS, the 6850K completed the task in eight minutes and 27.854 seconds compared to eight minutes and 38.886 seconds from the current-generation chip. In Cinebench R15, the new chip turned in a score of 1,311 versus 1,191 on the older chip. And in 3DMark Firestrike, the Broadwell-E CPU was good for a score of 9,440 while the Haswell-E chip received a score of 9,353.

All testing was performend on an ASRock X99 Extreme3 board with 16GB of Kingston DDR4 RAM and a Zotac GeForce GTX 980 Ti.

While the hexa-core i7-6850K certainly looks impressive, most enthusiasts are anxious to see what the 10 core / 20 thread Core i7-6950X can do. It’s expected to retail for around $1,500, firmly putting it out of reach of most hardware geeks on a budget.

Images courtesy Maintenance Bot

Permalink to story.

https://www.techspot.com/news/64698-intel-core-i7-6850k-broadwell-e-cpu-benchmarked.html

 

[theBest11778]

So now the "Enthusiast" Platform will be 2 generations behind instead of just 1. Intel needs to wake up and realize "Enthusiast" means "Newest" and fastest IPC. One generation behind, but offering more cores/threads was tolerable, but now being bumped back 2 generations, and this iteration doesn't even offer the mediocre 10% increase we're used to at this point.

No thanks... I think I'll stick with the faster, cooler, and cheaper mainstream CPUs.

 

[dividebyzero]

So now the "Enthusiast" Platform will be 2 generations behind instead of just 1. Intel needs to wake up and realize "Enthusiast" means "Newest" and fastest IPC.

HEDT hasn't meant that at all for some time. HEDT is about bandwidth, connectivity, and multi-threaded application performance.
The mainstream platform(s) are limited in PCI-E lanes and bandwidth. Multiple storage and I/O options generally require concessions with regards those options and multi-GPU. Likewise system memory capacity and overall memory bandwidth are also limited in the mainstream.
If your prime requirement is gaming with 1-2 graphics cards, modest drive requirements (especially if limited in PCIE dependent M.2/U.2/PCIE SSD), and general consumer content creation/productivity (I.e. most of the PC user base) then the mainstream platform serves well. Anyone wanting something more akin to a workstation (+ multi (3-4) GPU and benchmarking in general) system geared towards more intensive applications that actually scale well with multi-threading is often better served with HEDT.

Basically, if you don't see the advantages of the HEDT/WS/Server platforms, you don't need them.

 

[theBest11778]

HEDT hasn't meant that at all for some time. HEDT is about bandwidth, connectivity, and multi-threaded application performance.
The mainstream platform(s) are limited in PCI-E lanes and bandwidth. Multiple storage and I/O options generally require concessions with regards those options and multi-GPU. Likewise system memory capacity and overall memory bandwidth are also limited in the mainstream.
If your prime requirement is gaming with 1-2 graphics cards, modest drive requirements (especially if limited in PCIE dependent M.2/U.2/PCIE SSD), and general consumer content creation/productivity (I.e. most of the PC user base) then the mainstream platform serves well. Anyone wanting something more akin to a workstation (+ multi (3-4) GPU and benchmarking in general) system geared towards more intensive applications that actually scale well with multi-threading is often better served with HEDT.

Basically, if you don't see the advantages of the HEDT/WS/Server platforms, you don't need them.

I do see the benefits of more PCI-E lanes and greater storage options, AND greater core count. However, the platform being artificially gimped for some reason perplexes me. I need an 8c/16t cpu running at least 4.6ghz with the best IPC possible for my home server, but I refuse to pay top dollar for "new" old tech. Intel needs to skip broadwell on their "high end" consumer line just as they did mainstream to keep pace. At this point it's more beneficial to run highend client pcs vs a single server with low power clients. For me this isn't ideal. If Zen is at least on par with ivy bridge's IPC performance I'm willing to just dump Intel and support another option. Once again, still not ideal.

 

[dividebyzero]

I need an 8c/16t cpu running at least 4.6ghz with the best IPC possible for my home server, but I refuse to pay top dollar for "new" old tech.

Well good luck with that. Transistor density and transistor switching power have basically curtailed high OC's as process nodes have shrunk. What you are asking for is basically unattainable unless so much of the die is fused off that the unused logic acts as a heatsink for the active blocks (or more esoteric cooling is used)- hardly a recipe for financial success.

Intel needs to skip broadwell on their "high end" consumer line just as they did mainstream to keep pace.

No they don't. HEDT is a drop in the bucket with regards revenue that the Xeon-EP/-EN/-EX parts bring in. You advocate Intel trying to recoup $7-10 billion per fab on two generations of product per process node. Good luck with that too, since litho power consumption alone is astronomical - every architecture requires more metal layers going forward. Broadwell uses two more layers per chip than the previous Haswell (13 vs 11), and the trend will only continue. Higher fab retooling/refurb costs, increased energy costs from higher power lasers and more chip layers don't lend themselves to "skipping" a generation or two of product with which to amortize the cost and maximize ROI....and it will only get worse. Intel's initial order for 15 NXE:3350B's litho tools alone equates to a $1.5bn investment and is nowhere near the numbers they'll need for 7nm, and thats assuming they get the all the tools they need since TSMC, Samsung, Glofo, IBM, UMC, and STM all need orders filled as well as upgrading existing 3300's to 3350B specification.

If Zen is at least on par with ivy bridge's IPC performance I'm willing to just dump Intel and support another option. Once again, still not ideal.

Very much not ideal. You just said you needed 4.6GHz from an Intel chip. Samsung/Globalfoundries 14nmLPP is a relatively low power process node in comparison to Intel's P1272 (14nm). There is a damn good chance that Zen won't even reach 4.6, and if it does but its IPC is worse then you are looking at 4.6+....an even more unlikely scenario.

I think everyone would love to have huge IPC jumps, high guaranteed clocks, and all the cores that can be squeezed into a chip, but those days are gone on silicon. IPC increases play second fiddle to ISA extensions and the onus falls on software coders, clocks are at the mercy of transistor switching power and heat dispersion within the silicon, and core count is just a function of yield and production cost. Wishing and raging against semicon's won't alter the laws of physics or financial necessity.

 

[wiyosaya]

So it sounds like current gen "E" owners really do not have that much of a reason to update. It also sounds like an Ivy "E" owner like me, having recently upgraded to a used E5-1650v2, would have something to gain, but not all that much compared to the cost of the upgrade.

As I see it, I am better off upgrading my GTX 580 to either a Pascal based card, or when Pascal comes out, a used 980 Ti or similar.