Monday, May 15th 2017
Intel Readies the Core i9 Brand Extension
Intel is reportedly giving final touches to a new line of high-end desktop processors under the Core i9 brand extension. Until now, the company used the Core i7 brand extension broadly, to cover both the top-end parts of the mainstream-desktop (LGA115x) segment, and the high-end desktop (HEDT) segment, consisting of the LGA1366 and LGA2011-series sockets. With the advent of the new LGA2066 socket, Intel will be launching two distinct kinds of products - the Core i7 "Kaby Lake-X" quad-core series; and the Core i9 "Skylake-X" 6-core, 8-core, 10-core, and 12-core processors.
The Core i7 "Kaby Lake-X" will include the much talked about Core i7-7740K and i7-7640K quad-core processors (there's no Core i5 Kaby Lake-X). These chips will feature up to 1 MB of dedicated L2 cache per core, which is four times that of the existing i7-7700K chip. The i7-7740K features 8 MB of shared L2 cache; while the i7-7640K features just 6 MB. Interestingly, the i7-7640K also happens to lack HyperThreading, while the i7-7740K features it. The i7-7740K will ship with higher clock speeds than the i7-7700K, with 4.30 GHz core, and 4.50 GHz Turbo Boost. The i7-7640K features 4.00 GHz core, with 4.20 GHz Turbo Boost. The Core i9 series is a whole different beast.
The Core i9 series will consist of four parts, the Core i9-7800X six-core, Core i9-7820X eight-core, Core i9-7900X ten-core, and the Core i9-7920X twelve-core. All chips feature HyperThreading, and 1 MB of dedicated L2 cache per core. The i9-7800X features 8,448 KB (8.25 MB) of shared L3 cache, and comes with clock speeds of 3.50 GHz core, and 4.00 GHz Turbo Boost. The i9-7820X eight-core chip features 11,264 KB (11 MB) of shared L3 cache, clock speeds of 3.60 GHz core and 4.30 GHz Turbo Boost. The Core i9-7900X ten-core chip features 14,080 KB (13.75 MB) of shared L3 cache, and clock speeds of 3.30 GHz, with 4.30 GHz Turbo Boost. Clock speeds of the top-dog i9-7920X twelve-core chip are unknown at the moment, but it comes with 16,896 KB (16.5 MB) shared L3 cache.
The i9-7800X, and the i9-7820X, along with the "Kaby Lake-X" based quad-core parts, feature 28-lane PCI-Express gen 3.0 root complexes, so your 2-card SLI/CrossFire setup will run at just x8 per card. The i9-7900X and i9-7920X feature 44-lane PCIe gen 3.0 root complexes, which enable the 3-way/4-way multi-GPU setups you originally bought any HEDT chip for. Further, the "Kaby Lake-X" quad-core parts feature just dual-channel DDR4 memory interfaces. The entire Core i9 "Skylake-X" series features quad-channel DDR4 memory interface. Socket LGA2066 motherboards will feature quad-channel wiring, with up to 8 DIMM slots, but when a Kaby Lake-X chip is installed, two memory channels are inactive.
Intel is expected to launch most of the Core i7 "Kaby Lake-X" and Core i9 "Skylake-X" lineup in June 2017, with the top-end i9-7920X following on in August, 2017.
Source:
AnandTech Forums
The Core i7 "Kaby Lake-X" will include the much talked about Core i7-7740K and i7-7640K quad-core processors (there's no Core i5 Kaby Lake-X). These chips will feature up to 1 MB of dedicated L2 cache per core, which is four times that of the existing i7-7700K chip. The i7-7740K features 8 MB of shared L2 cache; while the i7-7640K features just 6 MB. Interestingly, the i7-7640K also happens to lack HyperThreading, while the i7-7740K features it. The i7-7740K will ship with higher clock speeds than the i7-7700K, with 4.30 GHz core, and 4.50 GHz Turbo Boost. The i7-7640K features 4.00 GHz core, with 4.20 GHz Turbo Boost. The Core i9 series is a whole different beast.
The i9-7800X, and the i9-7820X, along with the "Kaby Lake-X" based quad-core parts, feature 28-lane PCI-Express gen 3.0 root complexes, so your 2-card SLI/CrossFire setup will run at just x8 per card. The i9-7900X and i9-7920X feature 44-lane PCIe gen 3.0 root complexes, which enable the 3-way/4-way multi-GPU setups you originally bought any HEDT chip for. Further, the "Kaby Lake-X" quad-core parts feature just dual-channel DDR4 memory interfaces. The entire Core i9 "Skylake-X" series features quad-channel DDR4 memory interface. Socket LGA2066 motherboards will feature quad-channel wiring, with up to 8 DIMM slots, but when a Kaby Lake-X chip is installed, two memory channels are inactive.
Intel is expected to launch most of the Core i7 "Kaby Lake-X" and Core i9 "Skylake-X" lineup in June 2017, with the top-end i9-7920X following on in August, 2017.
34 Comments on Intel Readies the Core i9 Brand Extension
Can't wait to see the prices....:fear:
Could we perhaps wait for the official (press) releases before sharpening the pitchforks?
"The i9-7800X, and the i9-7820X, along with the "Kaby Lake-X" based quad-core parts, feature 28-lane PCI-Express gen 3.0 root complexes"
This is not accurate, Kaby Lake-X quad-core parts will have 16 PCI-E 3.0 lanes...
Edit: It might also be worthwhile to state that 7900X and 7820X will support Turbo 3.0 (both up to 4.5GHz single-thread), while 7800X and lower will not. It's safe to assume that 7920X will also support Turbo 3.0.
However, it's not unexpected intel's first reaction is a hasty one. I'm not expecting a proper answer to Ryzen till 2018 at the earliest.
If this is true then Intel's really muddying the waters with their naming scheme.
Intel already have the mainstream covered, Kaby-Lake is superior to Ryzen's cores.You know Skylake-X was designed long before the release of Ryzen. Keep in mind that even Broadwell-E beats Ryzen in most benchmarks, so there should be no doubt that Skylake-X will be up for the task.
Personally, I don't have a problem with either lineup. But I predict intel will move to counter AMD's "core war", while AMD will try to up their IPC. Gotta love competition ;)
First you forgot to call out "at gaming".
Second, you forgot to call out "mainly 7700k".
Oh, and last but not least, you forgot to mention the whopping single digit difference one gets on average while on 1080Ti card.
If all that comes from pathetic tomshardware piece, you are forgiven.
Are there any?? Any test that have been made with this??
For example, I could easily spawn a thread for each physical core, let it compute paths, but only query the result once in a while. Am I really using all the cores in this scenarios? Cause a look at the task manager would say that I am.
Another example, I could have 32 threads doing useful things, but each only needing 10% CPU time. Thus all threads can run just as fine on 4 cores (with room to spare).
The only way to test this, would be to run a benchmark, look at the number of cores used and then disabling cores in BIOS, run the benchmark again, see if it makes any difference.
And then you find out games don't actually use that many cores. And then supporters of multi-cores will tell that's only because the hardware wasn't there... It goes on and on.
As a matter of fact, ever since Haswell-E (i7-5820K), the six-core has been the best buy:
i7-5820K($389) vs. i7-4730K($339)
i7-6800K($434) vs. i7-6700K($339)/i7-7700K($350)And how is Ryzen innovation when it's limping years behind Intel?
The HEDT platforms are not "year-old dies" either, the overarching architecture might be 1-1.5 years behind the mainstream, but it's way more refined and has better features:
- Better memory controller, more cache and better core intercommunication.
- Double memory bandwidth.
- Better vector performance, HEDT usually gets most of these features first.
- Broadwell-E and newer: better boosting, which works better with more cores.
So generally HEDT is not lagging behind in terms of IPC.
"Haswell-E" is not the same as "Haswell" with some more cores slapped together.
If you look at the progress from Sandy-Bridge -> Kaby-Lake vs. Sandy-Bridge-E -> Broadwell-E, it becomes quite clear that the HEDT is the only platform doing significant progress. The "IPC improvement claims" are very inflated due to special features such as AES acceleration (which most people rarely use), subtract those and you're left with ~6-15% real IPC improvements, and subtract vector extensions (which helps a lot in select applications), and you're left with <5% generic gain, mostly attributed to the prefetcher. So for generic performance in typical applications the gains from Sandy-Bridge to Kaby-Lake is practically only increases in clock speed, which are not that substantial anyway.
So while the mainstream has been stagnant for many years, HEDT is at least progressing with more cores. The relative performance gap between HEDT and mainstream has never been greater, and it will grow with Skylake-X.
115x for pentium/celery/i3/i5
20xx for i7/i9
3647 for big xeon