Intel Linux Developers Continue Work On Coffee Lake Bring-Up

[phoronix]

Phoronix: Intel Linux Developers Continue Work On Coffee Lake Bring-Up

In addition to Intel's Linux open-source kernel developers working on Cannonlake support, they have also been working on the hardware bring-up for Coffee Lake...

http://www.phoronix.com/scan.php?page=news_item&px=Intel-Coffee-Lake-DRM

[starshipeleven]

Coffee Lake is expected to be close to Kabylake but with up to 3% performance gains

Fixed. 30% is lolwtf-astic.

[Mystro256]

Fixed. 30% is lolwtf-astic.

lol marketing departments are always like "1 MILLION PERCENT FASTER*"


*may not be 1 million percent faster, ask your doctor if this product is right for you /s

[Luke_Wolf]

Fixed. 30% is lolwtf-astic.

As Seen On AnandTech(TM)
Basically the 30% claim is comparing a dual core 15w laptop part with a 15w quad core part with presumably lower clocking using SysMark ... So yeah...

[chuckula]

Fixed. 30% is lolwtf-astic.

If going from 4 cores to 6 cores and only claiming a 30% performance boost is unrealistic, then Threadripper is already DOA.

[starshipeleven]

If going from 4 cores to 6 cores and only claiming a 30% performance boost is unrealistic, then Threadripper is already DOA.

If assuming that all software give a fuck about 6 cores and only claiming a 30% performance boost (no ifs no buts), then it's plain pulling numbers out of their backsides.

[chuckula]

If assuming that all software give a fuck about 6 cores and only claiming a 30% performance boost (no ifs no buts), then it's plain pulling numbers out of their backsides.

Care to show me the Intel marketing slide that says literally every piece of software in existence will get an automatic 30% boost? Because I guarantee nobody at Intel wrote that and you are just inventing your own standard. A 30% boost in many (not all) multi-threaded benchmarks while probably having somewhat lower clockspeeds seems quite reasonable to me.

Trust me, if a 16 core Threadripper is slower in a some single-threaded benchmark compared to an 1800X and you are on here in 3 months calling it a miracle just because it does better in multi-threaded benchmarks compared to the same AMD part with half the cores, I'll remember.

[starshipeleven]

Care to show me the Intel marketing slide that says literally every piece of software in existence will get an automatic 30% boost?

1. I didn't blame Intel so you can get down of your white horse, I actually don't even know where this info comes from, I just state that number is unrealistic in its current form.
2. Oh come on you know marketing too, it's all about taking favorable numbers from specific benchmarks out of context and hope people bite the bait.

Trust me, if a 16 core Threadripper is slower in a some single-threaded benchmark compared to an 1800X and you are on here in 3 months calling it a miracle just because it does better in multi-threaded benchmarks compared to the same AMD part with half the cores, I'll remember.

Threadripper is an obvious high-end part for serious CPU work (which is usually multithreaded), so it's main purpose is doing multithreaded jobs.

Coffee Lake is the name of parts aimed at consumer stuff where IPC is king, plain and simple. Hence why I talked of 3%, because that's the average IPC increase between Intel generations, and the most likely "performance increase" an actual consumer will notice (or not notice).

[caligula]

1. I didn't blame Intel so you can get down of your white horse, I actually don't even know where this info comes from, I just state that number is unrealistic in its current form.
2. Oh come on you know marketing too, it's all about taking favorable numbers from specific benchmarks out of context and hope people bite the bait.

Threadripper is an obvious high-end part for serious CPU work (which is usually multithreaded), so it's main purpose is doing multithreaded jobs.

Coffee Lake is the name of parts aimed at consumer stuff where IPC is king, plain and simple. Hence why I talked of 3%, because that's the average IPC increase between Intel generations, and the most likely "performance increase" an actual consumer will notice (or not notice).

The development of low core count desktop CPUs has been quite disappointing in the recent years. My old overclocked 2700k tower PC could still probably compete quite well with the latest gen 7700k. Biggest advantages since year 2011: 32GB of RAM on small mobos, AVX2 (helps even with AES), NVMe SSD, faster USB & faster PCIe. Nothing really fancy in the CPU area per se. Also, I've basically had about the same CPU perf on laptops since 2011. Back then I had a really fast Core i7, now I use i5. Longer battery life is nice to have, for example when watching videos on a plane.

I'm really surprised how long Intel can continue with this. The *bridge/*well/*lake generations are all pretty minimal and incremental changes. After Coffee Lake is out, it seems Intel is pretty much done with the IPC and frequency improvements. No further speedups available there, ever. Well, maybe some improvements if they replaced the tooth paste between the chip and the heatspreader, but many have already been there done that. Why would anyone upgrade both mobo (incompatible socket) and CPU for < 0.5% perf gains? Unfortunately, the CPU is becoming the least interesting component in a system. The saddest part is, there's no way to gain any significant scalability even with software optimizations. I really hope that AMD will succeed this time. Their new stuff would pave the way for properly threaded new software.

[cj.wijtmans]

When is Covfefe lake?