Intel's 8th Generation Core Family - Coffee Lake (LGA 1151, 6C/12T)

Where do you expect Core i7-8700K's Turbo to land?

  • 3.8/3.9 GHz

    Votes: 0 0.0%

  • 4.0/4.1 GHz

    Votes: 3 23.1%

  • 4.2/4.3 GHz

    Votes: 6 46.2%

  • 4.4/4.5 GHz

    Votes: 3 23.1%

  • 4.6/4.7 GHz

    Votes: 1 7.7%
  • Total voters
    13
  • Poll closed .
juanrga said:
Didn't I say you just above that OC the mesh improves gaming performance a lot of?

Up to 25% higher performance with a mesh OC

500x1000px-LL-91e0cd94_xU4ML9J.png


I find amazing that many reviews tested OC memory on RyZen to reduce latency and improve performance in games compared ot a stock comparison, but it seems no review of SKL-X has tested gaming performance with OC mesh to reduce latency.
Click to expand...

Intel doesn't seem to be coddling reviewers like AMD is, so they conveniently "forget" to bother. ;)
 
Hmm, the 8700k might just get me to upgrade from my 5820k. The IPC + higher clocks should actually get me a decent boost. I'm loving the fact that we're FINALLY getting passed 4 cores in the mainstream CPU realm. I think we're going to finally start seeing more games/apps that can take advantage of the extra threads now.
 
juanrga said:
Many AMD reivews retested with newest BIOS/AGESA and continue retesting today. Most Intel reviews tested with beta BIOS and forgot to retest with latest BIOS fixing turbo 3 and improving performance.
Click to expand...
Which reminds me. Im sure I remember [H] promising a review. Was I dreaming?
 
juanrga said:
Many AMD reivews retested with newest BIOS/AGESA and continue retesting today. Most Intel reviews tested with beta BIOS and forgot to retest with latest BIOS fixing turbo 3 and improving performance.
Click to expand...

While I agree that they should isn't that mostly due to Ryzen being an entirely new (and rushed) architecture versus Intel's much more mature one?
 
^ What about Coffee Lake, I guess that will still use ringbus right? Wouldn't make any logical sense to move to mesh design with 6 core from these results.

I'm actually really baffled about the X299 Intel HEDT platform, seems like so many things doesn't seem optimal about it, yet they charge a fortune (I'm looking at already 8 core @ $599 and looking at its IPC vs 7700K and the pricing difference with not only the CPU but motherboard and RAM included) and Coffee Lake will probably make that platform seem even more depressive from a bang-for-buck perspective which baffles me why Intel keep doing this to themselves (hello?). Doesn't matter for me personally as I never to begin with had my eyes on it (Coffee Lake or whatever NEW generation always seemed more interesting to me especially from the bang-for-buck perspective to get that latest optimized process that logically has better chances for better clocks/IPC etc than something little bit older but tweaked. I know I know it doesn't always work that way, especially with increasingly more difficult CPU manufacturing process but at least the potential is there).
 
Last edited:
RPGWiZaRD said:
^ What about Coffee Lake, I guess that will still use ringbus right? Wouldn't make any logical sense to move to mesh design with 6 core from these results.
Click to expand...

Still ringbus yes. ICL and TGL mainstream will most likely use ringbus too with the same different uncore compared to enterprise/cloud.
 
Got to wonder about that- looks like the mesh essentially drops IPC in multi-threaded scenarios. That easily means that gaming should be limited to the sub-HPC parts (socket 115x) or just Ryzen, simply because Intel HPC won't be faster and AMD will probably always be cheaper.

So long as there isn't a BIOS update or compiler update or something else to address what we're seeing above, as AMD is frantically (and effectively) pushing out for Ryzen.
 
IdiotInCharge said:
Got to wonder about that- looks like the mesh essentially drops IPC in multi-threaded scenarios. That easily means that gaming should be limited to the sub-HPC parts (socket 115x) or just Ryzen, simply because Intel HPC won't be faster and AMD will probably always be cheaper.

So long as there isn't a BIOS update or compiler update or something else to address what we're seeing above, as AMD is frantically (and effectively) pushing out for Ryzen.
Click to expand...

Mesh clock is a notorious issue on the beta BIOSes. Same as turbo clocks not working correctly etc. Its a new thing the mobo makers haven't used before while they are busy trying to auto overclock.

On the server lines with (real) stock settings its an entirely different story.
 
Well, if the experience mirrors AMD's with their new inter-module interconnect, perhaps Intel will get it all in order. Seems to be something they'd want to put a rush on if the issue is pervasive.
 
IdiotInCharge said:
Well, if the experience mirrors AMD's with their new inter-module interconnect, perhaps Intel will get it all in order. Seems to be something they'd want to put a rush on if the issue is pervasive.
Click to expand...

Its absolutely nothing similar. The compare to that is Omnipath/UPI/QPI.

You would have to look inside a single CCX to find something equal. Just as the ringbus on a mainstream part.
 
Last edited:
Similar in that they are both new and unknown quantities both for BIOS engineers and software engineers, I'm quite aware that even if they looked similar from even a low-level perspective, they'd be still be incomparable in implementation.

My only point is that there may be some catching up to do on Intel's part. I'll emphasize may.
 
IdiotInCharge said:
Similar in that they are both new and unknown quantities both for BIOS engineers and software engineers, I'm quite aware that even if they looked similar from even a low-level perspective, they'd be still be incomparable in implementation.

My only point is that there may be some catching up to do on Intel's part. I'll emphasize may.
Click to expand...

It adds latency just like the fabric AMD uses, not quite as high as AMD due to shorter paths. This is why the performance is down on chips using the mesh. There is no magic bios update going to fix it as even overclocking the mesh adds a tiny amount of performance. This is why their IPC is down from the last generation due to the added latency. So I doubt any bios update is going to cause much change in benchmarks on Intel's side, however programs may benefit from tuning just like AMD has had some programs tuned and that added performance. Marketing slides wont be changing that issue for Intel, cause I am sure they are inbound.
 
Ok, I know think is been speculated already, but what % of a chance do you guy feel the Z170 chipset will support Coffee?
I ask because I really want to upgrade my 6600K, and have this weekend to myself so I can put time into OC/deliding a 7700K, and also my local PC store has them in sale until the EOD tomorrow..
That being said, if you were all 80% sure Z170 will support a 8700K then I guess I can get a hold of myself and wait another month..
I can't find the post, but a wccftech (yeah, I know, grain of salt) article said there is little chance that Z170 would support Coffee. So if it wont support it, Im heading to the store tonight.
 
I would be surprised (pleasantly) if there is Z170 support for CFL, as we don't know if it will even be technically possible. We're adding two cores on the top end with CFL, so power requirements may be different enough or they may have made other interface changes that prevent support by Z170.

But if they do? I usually don't replace stuff till it breaks or is massively outclassed, but I I'd be willing to replace my 6700K for two more cores.
 
The selfish part of me hopes they aren't backwards compatible so they will be easier to buy at launch. Sorry.
 
Any word on the iGPU for Coffee? I don't really game much but would still like a little more power than what my last SKL had.

Raven Ridge is just too far away.
 
Nihilus1 said:
Any word on the iGPU for Coffee? I don't really game much but would still like a little more power than what my last SKL had.

Raven Ridge is just too far away.
Click to expand...

For desktop parts, at least the six-core parts, it may be even less potent due to switching die space from GPU cores to CPU cores, thus keeping the overall die size within the same range.
 
IdiotInCharge said:
I would be surprised (pleasantly) if there is Z170 support for CFL, as we don't know if it will even be technically possible. We're adding two cores on the top end with CFL, so power requirements may be different enough or they may have made other interface changes that prevent support by Z170.

But if they do? I usually don't replace stuff till it breaks or is massively outclassed, but I I'd be willing to replace my 6700K for two more cores.
Click to expand...

SKL and KBL supports IMVP8. CFL supports both IMVP8 and IMVP9. CNL, ICL and TGL supports IMVP9. So technically it can.

300 series platforms are listed as IMVP8/IMVP9 compatible.
 
Nihilus1 said:
Any word on the iGPU for Coffee? I don't really game much but would still like a little more power than what my last SKL had.

Raven Ridge is just too far away.
Click to expand...

The S and H models got a GT2 as from KBL. The U models are GT3e. Again same GPU as KBL. (Gen9.5).

If there is any changes it will be at clock.
 
IdiotInCharge said:
For desktop parts, at least the six-core parts, it may be even less potent due to switching die space from GPU cores to CPU cores, thus keeping the overall die size within the same range.
Click to expand...

There are no switching die space. A 4+2 is 126mm2 with CFL-S/H and 6+2 is 149mm2. A KBL 4+2 is 123mm2, the 3mm2 increase is due to going to 14nm++. CFL-U is 186mm2 with 4+3e.
 
Shintai said:
There are no switching die space. A 4+2 is 126mm2 with CFL-S/H and 6+2 is 149mm2. A KBL 4+2 is 123mm2, the 3mm2 increase is due to going to 14nm++. CFL-U is 186mm2 with 4+3e.
Click to expand...

So no backtrack on iGPU hardware resources then? That'll be nice!
 
IdiotInCharge said:
For desktop parts, at least the six-core parts, it may be even less potent due to switching die space from GPU cores to CPU cores, thus keeping the overall die size within the same range.
Click to expand...

It just goes in the opposite direction, with the iGPU being more potent due to higher clocks and improved memory BW.
 
juanrga said:
It just goes in the opposite direction, with the iGPU being more potent due to higher clocks and improved memory BW.
Click to expand...

I don't disagree- I've been pleasantly surprised with the progress Intel has made with their iGPU's. One in my 7500U is a gladiator for an ultrabook, and while I understand that the balance of resources is different on the desktop, more can't hurt.
 
Is coffee lake soldered? That is the question. Even if it isn't HEDT Intel needs to stop being all whack and using paste under that heat spreader.
 
You must log in or register to reply here.
Back
Top