News Intel 10th Gen Comet Lake-H Arrives With Up to Eight Cores at 5.3 GHz

[watzupken]

10nm has been shipping for quite a long time - got my 2 Dell 13 2-in-1s back in October with Ice Lake... So the idea that 10nm is not shipping is false. Process apparently does not matter to the market - AMD has yet to get any traction at all with their supposed "Superior Product" - all those extra useless cores are like fart cans on Hondas - serve no real purpose (outside 1-2% of the market, unless all you do is run benchmarks, rather than actually use the PC).

I will stick with my i9900K (5Ghz on all 8 cores on air) and dual 2080Tis until something much better comes along. Comet Lake is not for me, Rocket Lake will most likely be my next upgrade - would probably coincide with the new Nvidia GPU launch. Enjoy this "Netburst 2020" event, the new Core is coming - and history will once again repeat itself.

Not sure if history will repeat itself this time even with the release of a new CPU architecture. Intel has always been at the forefront when it comes to fab, and that gave them a significant advantage against the competitors in the past. Their advance CPU architect also contributed to their success. But it’s always been these 2 factors progressing at the same time to give Intel the famous Tick Tock cycle. This however fell through because their 10mm is not working out well and broke the cycle. They basically lost at least 3 years of opportunity to AMD as they persisted with trying to get the 10nm working. Even if the long overdue Tock happens next year, they’ve lost the advantage of having a more advance fab. So I am not expecting a sweeping success as they did when they first introduced the Core series in face of the Athlon 64 days. I suspect it will be a hard fight between AMD, ARM and Intel going forward.

 

[jeremyj_83]

The Cascade Lake are as different from the Whisky Lake as the Gen 2 and 3 Ryzens are from the 1st gen Ryzen. Ryzen is nothing but small iterative changes - the move the "7nm" hasn't made much of a difference - little to no OC ability since the manufacturing is so janky that they have to ultra bin everything just to get usable parts... At some point the treasure chest that Jim Keller left is pretty much depleted...

Little to no OC ability is pointless for a majority of people. I could OC my 4770k, and did for a while, but eventually decided I didn't care and removed the OC. When OC anything in the 9000, or soon 10000, series you need some pretty big coolers, especially for the 9700k & 9900k. That adds a good chunk onto cost as well as making sure that you have a case big enough for the coolers and if running on air enough space to fit your RAM. That doesn't even take into account the added electrical costs and cooling costs during summer.

Zen 2 is far from small iterative changes. Yes Zen - Zen+ was small, but it still increase the IPC by about 3%. What has Intel changed about its micro-architecture since Skylake? Absolutely nothing. Take a 9000 series with the same core count and speed as the 6000 series and you will find that they perform almost identically. Just a refresher Skylake was released in August 2015! Going back even further to be beginning of the Core architecture there was a big just from Conroe to Nehalem and then from Nehalem to Sandy Bridge. Since Sandy Bridge, it has been all iterative changes with 3-5% IPC gains each generation. Yes over enough generations those 3-5% gains make a large difference in performance; however, it took from Sandy Bridge to Skylake to get the same jump as we got from Zen to Zen2.

My engineering workstations moved to Xeon when the Xeon Scalable series was released - all dual socket (single CPU installed on most) - each with a RTX6000 and back ended by 2 16GPU Nvidia DGX systems. We had auditioned the 2nd gen Turdripper and it was just that - ZERO ISV optimizations, janky memory access and godawful performance. TBH if the Xeon W with 28 cores had been around would have run with that over the more expensive Xeon multi socket CPUs... but you get what you need when you need it - cost was not even in the top 5 considerations.

2nd Gen Threadripper was a bit of an odd duck in terms of performance and core count. Due to how the memory is mapped, you could have lower performance with the 32 core CPU than the 12 core CPU. 3rd Gen Threadripper is absolutely a monster in the 24 or 32 core version. With the newest Threadripper you can get more IO than your Xeon workstation with 1 CPU installed and a LOT more performance than it as well, a lot of times the 24 core 3960X is faster than the 28 core Xeon W-3275. All of this while the Threadrippers are using the same or less power as well. https://www.servethehome.com/amd-ryzen-threadripper-3970x-review-32-cores-of-madness/ Sure they might not have some of the ISV optimizations, but a fair amount of Intel optimizations help out the Ryzens as well. Obviously AVX512 is something the AMD chips cannot do, but that isn't very common anyways. Any other AVX instructions will run faster on AMD than Intel anyways.

I have personal knowledge of how fast the 2nd Gen Epyc's are. 1st Gen was usually a slower than 1st Gen Xeon Scalable. Now with 2nd Gen vs 2nd Gen the Epyc is almost always faster, while having clock speed disadvantages in chips with same core counts. My work has some proprietary software and it was running on our 1st Gen Epyc 7401's just fine. CPU usage was at about 50% average across 4 vCPUs with spikes to 80%. Fast forward to October 2019 and we got some new hosts with dual Epyc 7502's and I migrated the VMs to those hosts and the CPU usage went down to 20% average and spikes to 50%. Per core performance between the added clock speed and IPC improvements of the 7502 vs 7401 is a good 50-60%.

 

[sonichedgehog360]

It's precisely articles like this that give Tom's Hardware a bad name. Namely, this beauty: "Eight Cores at 5.3 GHz" This is patently false, up there with "just buy it." You will be hard pressed to find an eight-core desktop CPU from Intel that run stable at 5.3 GHz across all eight cores, never mind a notebook SKU.