SoylentNews is people
SoylentNews
Refresh Done Right? Intel Comet Lake Packs Up to 10 Cores
[coreboot], an open source project to replace the BIOS and UEFI, has some vital information on Comet Lakes. According to the Github page, Comet Lake-U (CML-U) processors, which are primarily aimed at laptops, carry up to six cores, while the Comet Lake-H (CFL-H) and Comet Lake-S (CMT-S) chips feature up to 10 cores.
Rumors on the street are that AMD's forthcoming Ryzen 3000-series desktop processors could purportedly pack a whopping 16 cores on a single chip. During AMD's presentation at the CES 2019 tech show in January, an eight-core, 16-thread Ryzen 3000-series chip was trading blows with Intel's Core i9-9900K, which could have pressured the Santa Clara chipmaker to cranking Comet Lake's core count to 10 cores for safe measure.
Intel is expected to launch its Comet Lake processors around the middle of the year. It's possible Intel could announce the chips at Computex 2019, which starts May 28.
Also at PCGamesN.
Related: AMD Announces Radeon VII GPU, Teases Third-Generation Ryzen CPU
(Score: 0) by Anonymous Coward on Thursday March 14 2019, @05:21PM (1 child)
I remember this game. https://www.corewars.org/ [corewars.org]
There is a secret to winning, but it ain't gonna be more cores. It's going to be enhancing the efficiency of each core.
(Score: 0) by Anonymous Coward on Thursday March 14 2019, @07:05PM
Are they all connected?
(Score: 2) by DannyB on Thursday March 14 2019, @05:24PM (3 children)
It's finally happening!
I've been wanting to see more and more cores for at least ten years.
I had expected that clock speed would quit increasing, and mostly it has. The next adolescent bragging rights would be "well my system has 16 cores!", "oh yeah, well my system has 24 cores, so there! And with lights and orange juice cooling!"
We are at a point where developers in higher level languages are able to effectively use more cores as long as they can structure their problems in certain ways. The more you think about restructuring problems to run on multiple cores, the better at it you get -- I think. I had completely re-thought one problem. (previously described on SN once)
The lower I set my standards the more accomplishments I have.
(Score: 0) by Anonymous Coward on Thursday March 14 2019, @05:27PM
Golang with goroutines seems to be the easiest way IMHO.
(Score: 1, Insightful) by Anonymous Coward on Thursday March 14 2019, @06:34PM (1 child)
16 vs 24? My system has 32 cores already...
(Score: 2) by takyon on Thursday March 14 2019, @07:07PM
Threadripper 3 also comes out this year. Threadripper 2 raised max core count to 32, matching the first Epyc generation. Now Epyc is up to 64 and we'll see if T3 stays at 32 or increases to 48 or 64 cores.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by Azuma Hazuki on Thursday March 14 2019, @06:04PM (3 children)
Anyone remember all the Intel shills making fun of AMD with the "moar coarz!" meme? Hmmm...
I am "that girl" your mother warned you about...
(Score: 2) by takyon on Thursday March 14 2019, @06:59PM (2 children)
If that was said during the Bulldozer "module" era, then it would have been appropriate. If that was 2015-2016 pre-Zen, then sure.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by Azuma Hazuki on Friday March 15 2019, @01:02AM (1 child)
What goes around, comes around...
I am "that girl" your mother warned you about...
(Score: 2) by takyon on Friday March 15 2019, @03:17AM
I'm eager to see 8-16 cores become mainstream, but there will be cases when the core count is overkill or most cores are idling most of the time.
What was great about Ryzen (the original) is that it delivered up to 8 real cores (not 4x 2 "core" modules, which led to a lawsuit), along with a huge ~52% IPC gain to bridge the gap with Intel. This even beat their marketing, which had only been promising about a 40% improvement. It looks like Ryzen 3000-series (Zen 2) will boost IPC and clock speeds by just enough to almost match or slightly exceed Intel's latest 14nm++++++ yawner.
So Intel fanbois (disgusting) were correct to shit on AMD's Bulldozer modules. And in response, AMD did what had become unexpected and became competitive with Intel in multiple dimensions (it's unclear to me whether Bulldozer and successors were ever broadly competitive with Intel on $/performance, when ignoring things like integrated graphics).
Single-threaded performance is still important. It's a reason why I'm interested in Intel's big.LITTLE-inspired x86 core configuration that was revealed in December.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 0) by Anonymous Coward on Thursday March 14 2019, @06:06PM (1 child)
While it's great that they're continuing to push hardware advancement I was wondering if any of the new CPU's from Intel or AMD solved the flaws involving Spectre and Meltdown? I realize the companies themselves might not want to bring them up again as they still affect the older chips but I haven't found anything confirming they fixed it on newer ones either.
(Score: 2, Interesting) by Anonymous Coward on Thursday March 14 2019, @06:35PM
Meltdown was a major oversight on Intel's part and should be rectifiable (I don't know if they fixed it or not though). Spectre, though, attacks systems that are fundamental to modern processors (note that all of the major players are susceptible to Spectre) and I don't know if there is any fix that wouldn't severely hurt instruction throughput. Nobody is getting rid of speculative execution and branch prediction.
(Score: 0) by Anonymous Coward on Friday March 15 2019, @03:25AM (1 child)
keep adding cores. and vector instructions. features >> correctness.
(Score: 2) by takyon on Friday March 15 2019, @03:47AM
One man's bug is the NSA's feature.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 1) by LAV8.ORg on Monday March 18 2019, @12:35AM
Look at a list of processor FLOPS per core and you'll see that the average mark has barely budged over the past decade. Only very new high end CPUs and the odd few elderly server CPUs are significantly above the baseline. 10 cores is 1x better than 4 for the vast majority of software that remains bounded to a single process.
I mean I'll gladly accept more cores... but I'd yield max_n cores for significantly more OPS per core for a general use desktop.