Detailed Comet Lake-S CPU lineup appears to have leaked

[QuorTek]

do we have to get a new motherboard too?

[ET3D]

What I'm sure is on everyone's mind is what kind of performance these CPUs will have in real world applications, such as Word and Excel.

Anyway, nice to see the effects of competition.

[Xlucine]

It is - but the 3960X is not really a consumer CPU, its a little unfair to compare it to the 9900k. If we start to bring in CPUs over £1k then the comparisons get very different Not sure what you mean on the boost issue - if you enable MCE the 9900k reportedly hits 5ghz boost on all cores simultaneously, so i'd assume the 10 series will be able to do the same.

You are right about the board, would need replacing either way. I hadn't spotted that

Ah, I though that was what you meant by "high core count". On the consumer socket the 3950X meets the bill then, both highest core counts and the highest single threaded performance (more performance at stock than an overclocked 9900k!).

Intel boost is bad for high core count CPUs as it's too conservative - there's a larger gulf in single threaded performance between intel HEDT chips and intel consumer chips than there is on the AMD side. It also has the arbitrary time limits, but that's a different matter

[MatBailie]

Will these have meltdown/spectre fixes or are we still well off that?

Even the 8series and 9series processors had some mitigations baked in to the silicon, and the 9900-KS had yet more micro-code revisions for yet more mitigations. So, yes, it is very likely the 10series will have even more mitigations baked in.

New vulnerabilities of the same sort are "constantly" being found, however, and once discovered they'll require a mix of micro-code updates (which some UEFI/BIOS manufacturers -may- support for a -while-) or software updates in the OS kernal, which will -likely- have -some- performance impact. That said, most of the vulnerabilities are considered by some to be non-issues for general consumers, we -should- have "the worst" behind us now.

NB: That's even true for AMD, though to a measurably smaller degree.

What will interest me is whether Hyperthreading will continue to be one of the more prominent weak points (such as ZombieLoad).

[MatBailie]

unless your primary focus is gaming

Thing is that something like the r5 3600 is enough to drive pretty much any game to the point that you'll be either GPU limited at high quality settings or >120fps anyway; any CPU more expensive from either company is usually pretty marginal gains. The i5-9400f is a decent gaming chip, but 6c/6t means it can get threadlocked and stutter in some gaming situations. The cheapest Intel part that's either 6c/12t or 8c/8t is >£300, or £130 more than the 3600. We'll see where these new i5s land on price, although the new socket requirement means this lot probably won't be a great net cost at launch.

It's never quite that simple.

While it's true that Ryzen (Zen2) now has superior IPC to iCore (9th Gen), and in some cases has high single thread throughput than iCore (3900X vs 9900K, cinebench), it doesn't necessarily translate in to frames-per-second. At 1080p, in particular, iCore maintains quite a noticable FPS lead over Ryzen.

This appears to be related the behaviours of InfinityFabric, memory controller (speeds, latency interaction with on-chip cache), Windows Scheduler, etc, etc. Which -may- in turn mean that, as software evolves to better utilise Ryzen chips, we could (hopefully) see this discrepancy erode.

While it IS true to say that in many games this only becomes prominent at Very high refresh rates, some games already show the difference at under 120Hz (Assassin's Creed Odyssey, Far Cry 5, etc). The further down the product stack you go the lower the FPS ceiling becomes (the discrepancy is visible at lower FPS on a 3600 than on a 3900X).

This also means that in the years to come more and more games will show this discrepancy up.

https://www.eurogamer.net/articles/digitalfoundry-2019-amd-ryzen-9-3900x-vs-core-i9-9900k-review?page=2

https://www.eurogamer.net/articles/digitalfoundry-2019-amd-ryzen-9-3900x-vs-core-i9-9900k-review?page=3

For those -not- in the market for high FPS (like me, stuck on a 2500K, or those aiming at 4K rather FPS, etc), the argument does become mute. Just like having 8C/16T is overkill for me, because I don't do any highly parrallel work.

I just like to dream about being in that market segment, where the upper end of gaming performance matters

[Tabbykatze]

Would just like to point out mitigations != fixes.

They make it harder to exploit but do not outright fix the issue. Silicon baked mitigations only exist in 9900ks, more recent releases of 9th gen and the hedt lineup.

It seems a lot of people are mistaking a mitigation for a fix...

[DanceswithUnix]

It seems a lot of people are mistaking a mitigation for a fix...

The silicon is still broken, but with a mitigation in place no-one should be able to use the underlying flaw. Otherwise, it isn't a mitigation. For the end user, that particular problem that the mitigation addresses will be as good as fixed, albeit at some penalty.