New Features, Initial Thoughts
Improved Turbo Boost Max Technology 3.0
With the release of the Broadwell-E platform, Intel introduced Turbo Boost Max Technology 3.0 that allowed a single core on those CPUs to run at higher clock speeds than the others, effectively improving single threaded performance. With Skylake-X, Intel has improved the technology to utilize two TWO best cores, rather than just one.
This allows the 8-core and higher count processors from this launch to run at higher frequencies when only one or two cores is being utilized. In the two products that we have clock speeds for, that is a 200 MHz advantage over standard Turbo Boost technology. Intel hopes that this improvement in the technology gives them another advantage in any gaming or lightly threaded workload over the AMD Ryzen and upcoming Threadripper processors.
Cache Hierarchy
Skylake-X processors will also rebalance the cache hierarchy compared to previous generations, rebalancing to more exclusive per-core cache at the expensive of shared LLC. While Broadwell-E had 256KB of private L3 cache per core, and 2.5 MB per core of shared, Skylake-X moves to 1MB of private cache per core and 1.375MB per core of shared.
This shift in cache division will increase the hit rate on the lowest latency memory requests, though we do expect inter-core latency to increase slightly as a result. Intel obviously has made this decision based on workload profiling so I am curious to see how it impacts our testing in the coming weeks.
Overclocking Tweaks
Intel does add three new overclocking capabilities to the Skylake-X processors with this release.
First, we see that Intel has added AVX-512 instruction ratio offsets, allowing you to lower performance of AVX performance to the benefit of power consumption, relieving one bottleneck often cited for extreme overclockers. Second, memory controller trim voltage control was added. And finally, PEG/DMI overclocking is opening up.
The X299 Chipset
Though we stuck with the X99 chipset and socket for Broadwell-E, Skylake-X and Kaby Lake-X will be using a new chipset and a new LGA2066 socket. We have already seen motherboards leaking out and we should have numerous announcements through Computex this week (check back to pcper.com for all the news!) but now we have preliminary details on what changes it offers.
Fundamentally it looks pretty much the same. The biggest change is one of connectivity – the X299 chipset will now offer up to 24 PCIe 3.0 lanes, mirroring the capability of the Z270 chipset. Compared to the X99 chipset, that only included 8 lanes of PCIe 2.0, this is a significant increase. The DMI connection between the chipset and the processor is also upgrade to DMI 3.0, giving us a doubling of peak throughput (4GB/s rather than 2GB/s). That helps to alleviate the bottleneck from the chipset to the CPU, though a highly saturated system utilizing chipset-based connectivity could still hit speed limitations.
I would expect that concern to be alleviated for a majority of consumers though as the 28 or 44 lanes of PCIe Gen 3.0 provided by the processors (Skylake-X) allows for multi-GPU configurations at x16 or x8 speeds with room for PCIe NVMe storage to boot. Either way, it’s impressive that an X299 system with a 10-core or higher processor would have access to 68 lanes of PCI Express in total – 44 from the CPU and 24 from the chipset.
Default memory speed gets a jump from 2133 MHz on the X99 systems to 2666 MHz on the new X299 systems, giving the new platform another potential advantage. We also have 8x SATA 3 channels, 10x USB 3.0 ports and support for 3-way RAID of PCIe and NVMe drives as a part of this system platform.
Closing Thoughts
Today marks only the announcement of the Skylake-X and Kaby Lake-X platform, but it sets the stage for an interesting summer in the high-end CPU space. AMD threw down the first gauntlet by releasing 8-core processors in the same space that Intel had been limiting to 4-cores. Next, AMD announced its intent to release 16-core processors under the Threadripper brand directly targeting Intel’s profitable halo product line under the X99 platform. Intel’s response is now two-fold – release more and higher core count CPUs, up to 18-cores with 36-threads, while also lowering prices in a couple of key areas. (The $599 8c/16t Core i7-7820X comes to mind.)
It’s still not a perfect setup and Intel is leaving its flank open again for another attack. The $999 price tag for the 10-core Core i9-7900X is aggressive compared to Intel’s own launch last year but I would full expect AMD to offer 16-core Threadrippers for that same price, maybe slightly higher. Intel’s 16-core processor is set to run $1699, giving AMD another chance to offer a matching core/thread count part at a lower price. No, I don’t expect single threaded performance to change at all, and Intel should still have the advantage there (maybe more so with increase clocks) but multi-threaded performance could still swing AMD’s way.
It’s a race to release, and a race that enthusiasts should be excited to see play out. We’ll have more details from Intel and reviews of products first, so stay tuned to PC Perspective for that!
What the heck. I still want
What the heck. I still want to get Coffee Lake for 8600k instead of 7640x.
CFL should launch in August.
CFL should launch in August.
Those prices are crazy.
Those prices are crazy. Clearly not very worried.
Intel fan boys have already been going on about how cheap
they are when its the clearly the opposite.
AMD is the way to go I think.
I surely agree that SOME of
I surely agree that SOME of the prices seem crazy, but that $599 8c/16t will be much more compelling.
The 95w 1800x seem to have
The 95w 1800x seem to have settled at $450 (amazon) and seem to have the same base clock as the $600 140w 7820X
I believe in heavy multitasking (specially rendering) they deliver similar performance when equally clocked.
So it would be interesting if the 1800x match the 7820x in cinebench, but at the same time does it using less energy ?
But most likely the 7820x wont be compared with the $450 1800x,
but with a similarly priced 1955X X399 Ryzen (3.6ghz base 10 core)
Skylake generally offers
Skylake generally offers 5-10% more IPC / thread at the same clock as Ryzen (and Broadwell). Combined with the higher clocks (turbo boost 2.0 = 4.3 GHz for the 8 core.. vs. ~ 3.7 ghz all core turbo on Ryzen 7) and possibly even higher IPC from the 1MB L2 cache, the 7820x is almost certainly going to be noticeably faster than 1800X stock vs stock..
That said i wouldn’t be surprised if 1800X was still = or more power efficient per performance..
can anyone confirm how intels
can anyone confirm how intels turbo works? I see people assuming that 4.3 is going to be all cores but I have seen it mentioned as 2 cores and 4.5 as preferred core (1 core). It also doesn’t seem to be a sustained number based on the description.
Some gamers are going to opt for intel just because. Already see a guy calling the 1800x a disappointment and planning to buy $150 more 7820k even tho its going to clock lower than a 7700k and hes calling the 1800x a disappointment because of a completely single threaded game.
edit: saw the information about the 1 or 2 cores for max turbo 3. Seems strange if all core turbo is 4.3. whats with the base clock then? Under what condition does it go to that clock? my experience with haswell was it went to the base clock with all cores active. Seems something to be investigated. If its all cores, how long will it stay at that frequency for all cores?
Still uncompetitively priced.
Still uncompetitively priced. R7 1700 plus motherboard and memory is much cheaper than even the six core.
AMD will still having raw
AMD will still having raw pricing advantage going forward, yes. I think Intel is betting on single threaded performance and platform advantage for now.
Core i7-7800X vs Core
Core i7-7800X vs Core i7-7740X that is an interest position for buyers.
Also the $620 6 core 6850K seem to have no replacement. If you want more then 28pcie lanes you now have to spend $1000 minimum.
So Intel dropped prices by 15% for the 6 core 28pcie model
But also introduced an 8 core 28pcie model at the same price.
It seem to me Intel had to do a lot of pricing gymnastic in this release to have the lineup make sense.
Good News for the Intel
Good News for the Intel crowd, “but” I want to know more about reducing the L3 cache by a third and the increase in latency between the cores! Why is there always a “but” with Intel?
The “but” is always cost to
The “but” is always cost to manufacture.. (i.e. keep die size down).
However, Skylake-X’s per-core L2 cache is now 4x size (1MB vs 1/4MB), and the L3 is ‘non inclusive’ meaning it doesn’t always have the same data as the L2. This may improve IPC a little further over a standard broadwell/skylake cache setup..
I think this is just outright
I think this is just outright proof that Intel had been kneecapping their products just to skim margins… Their 8 core part dropped almost 40% in price in just 1 generation. I doubt that has ever been done before. And on their top product, they added about 10-20% efficiency AND added 50% compute power.
Be sure to vote with your wallet people. If you support this idea, give Intel your money. If you don’t support what they’ve done, then you need to make sure they see this.
You’re foolish to believe
You're foolish to believe that if AMD were in the same position, that they would not have gone for similar margins. The purpose of a publicly traded company is to turn a profit for their shareholders. When a company has a corner on the market, they can and will increase their margins if the market can bear it.
Someone posts a comment like
Someone posts a comment like this every time AMD releases a part that comes anywhere near Intel’s performance. And someone always says “vote with your wallet”, as if people are supposed to buy inferior products just to show “them” that “we don’t like your higher priced parts, even though they perform much better”. Give me a break! Yeah, buddy, it’s all a big conspiracy designed to fool you and make you part with your money. It’s called free enterprise and an efficient business model…
leave it to intel to rip
leave it to intel to rip their consumers a knew one again. you build it, they will stupidly come.
I think its somewhat obvious
I think its somewhat obvious the cache is 1.375 MB per core based on the Skylake X parts.
If so that would mean 24.75 MB of L3 cache for the 18 core.
Makes me wonder if it’s 1.5MB of cache with room for defects to be disabled.
I also think the high end parts won’t be capped at 165W.
There is rumor of the Xeon processors going up to 205W