Intel is talking up its new high-performance (and high-price) X-Series chips at Computex. But while those chips should offer some serious compute power, they’re still based on the same architecture as existing Intel 6th-gen Core “Skylake” processors and 7th-gen Core “Kaby Lake” chips.
But the company is starting to talk a bit more about the upcoming 8th-gen Core series of chips, code-named “Coffee Lake.”
While Intel hasn’t provided many details yet, here’s what the company is willing to say: they offer “more than 30 percent performance improvement” over 7th-gen chips.
You should always take claims like that with a grain of salt, since Intel ran the benchmarks (including “performance tests, sych as SYSmark and MobileMark”) itself and may have performed optimizations ahead of time.
But while a 30-percent performance boost from one chip gen to the next is a pretty big deal, there’s at least one reason to believe Intel’s numbers: the company seems to be comparing a dual-core 7th-gen chip with a quad-core 8th-gen chip with support for higher clock speeds.
The comparison is fair, because both chips have 15 watt TDPs, suggesting that the new quad-core processor will occupy the same space in Intel’s Coffee Lake Lineup that the Core i7-7500U does in the Kaby Lake family.
Of course, doubling the number of CPU cores and threads only helps in tasks that are optimized to take advantage of the multicore architecture. While I suspect single-core performance will get a boost from the higher clock speed, the performance gains will likely be more modest.
FYI, here are some details for the two chip configurations Intel compared to get its 30-percent figure:
- Core i7-7500U – 15 watt, 2-core/4-thread with Turbo speeds up to 3.5 GHz and 8GB DDR4-2133 memory
- 8th-gen Core i7 – 15 watt, 4-core/8-thread with Turbo speeds up to 4 GHz and 8GB DDR4-2400 memory
15W cpu benchmarks… I guess that means laptop chips. The laptop had better have an awesome thermal dissipation solution to keep the Turbo mode on as long as possible. I could see older chips out performing the newer ones if the cooling solution is superior. Just because the new laptop has a newer chip does not mean that it is faster in cpu intensive benchmarks.
Wow, no-one should believe a word of this. It’s really sketchy.
Firstly, going from Intel’s 6th-gen to 7th-gen processors nets you 0.1% improvement.
They’re basically the same. And if they say the improvement from 6th to 7th is 15%, then follow-up by saying the jump from the 7th to 8th gen is 15%… well it makes it clear the actual improvement is again 0.1% faster.
Now secondly, the comparison between the 7th gen and the 8th gen is also between different memory speeds: 2133MHz vs 2400MHz. For all we know the 15% improvement in their cherry picked benchmark came from the increased memory bandwidth. Or should we say the 0.1% improvement. And did anyone else miss the +/- 7% error margin in the footnotes?
So we can actually infer the 8th gen chips are SLOWER than the 7th gen chips, because they NEED faster memory just to KEEP UP. Worse there is a clock speed difference, 4GHz vs 3.5GHz, a difference of 14%. So far we can say the 8th gen chips are only 78% speed of the 7th gen chips. And it gets even more ridiculous since they’re also comparing a 2C/4T chipset to a 4C/8T. Which is exactly double the resources. So the performance of a single 8th gen processor core is ONLY roughly 39% of the performance of a single 7th gen processor core.
What Intel should be saying instead is that in the same low power envelope, they will reduce their IPC performance by 10%, and keep the same clock speeds… but double the core and thread count. That’s an improvement everyone would want, and we’d be looking at an increase in Total Performance of around +80%. However, it seems more likely that Intel will improve IPC by 0.1% like they did the last “jump”, and increase frequencies slightly like they did the last “jump”, but actually manage to do double the core count in the same TDP –basically a doubling of efficiency.
What we don’t know is, what lithography is this “8th gen” built upon:
– Are they still using their +14nm Finfet lithography?
(if it is, it looks like a major rework of their microarchitecture, and a massive leap like from the Core 2 Duo to Sandy Bridge).
– Or are they using their stated 7nm lithography?
My bet, they haven’t made much (4%) improvements to the microarchitecture from the 7th gen to the 8th gen… however, the jump from 14nm to 7nm, which is projected for 2019-2020, has allowed them to double the core/thread count in to the same power/thermal envelope.
Comments are closed.