Intels Skylake-X and Coffee Lake CPUs Discussion

[kaelleria]

Not really a concern for me. Ive NEVER actually understood how butt-hurt people get over the "ZOMG INTEL MAEKS ME BUY ANNUDER MOBO" shit, PARTICULARLY given how little of an upgrade variouss CPU generations from Intel are anyway. Virtually NO ONE Is jumping from Skylake to Kaby Lake to Coffee Lake. If you have Skylake, Cannon Lake offers no particular upgrade. If you do workloads that need the extra 2 cores and already have Skylake or better....you should be looking at Ryzen or HEDT (from wither manufacturer), not wasting money on Coffee Lake.

---

Future compatability with Ice Lake seems both HIGHLY unlikely (Z390 is still going to be Socket 1151v2, 10nm will ASSUREDLY be on a new socket entirely), and pointless. Youre not going to be seriously jumping from Cannonlake to Ice Lake, or the one after, barring an absolute MIRACLE in IPC advancement that Intel hasnt been able to manage in a decade.

More or less yeah. People should expect 4-5 years out of any machine built these days, even higher quality laptops. AMDs gains seem greater than they are because their chips were garbage for so long. We're approaching the limits of silicon and I don't expect anything miraculous without a fundamentally different manufacturing process.

when can I upgrade from my 2500k? It's very weird to have a CPU this old still handling it's business. That's the only question I have regarding skylake and coffee lake or any future cpu.

Totally depends on your usage. If you're regretting not having more cores like I was when I upgraded from my 2500k then you could get basically anything. If you're looking for more gaming oomph, then you're going to be waiting for a while.

[Lathais]

If you build Intel, you have no compelling reasons to upgrade for 4-5 years anyway... which is just as long as AMD is claiming AM4 will last. Hell, i have no ctually compeling reason to upgrade my Z97 4790K. It in no way bottlenecks any games that wouldnt ALSO be bottlenecked by Coffee Lake.

and Ryzen performs similarly to Haswell for a bit less. So if there's no reason to upgrade to Coffee Lake from Haswell there is also no reason to not build new with a Ryzen if you don't have anything at all or what you are using is really old and outdated.

[mrgreenthump]

I am still convinced their 8700K will not yield any IPC increase and certainly not the 11% Intel claims in a leak.

It won't, it's probably the same deal as with Kabylake, 8700K single core boost is 4.4% higher than 7700k and they tend to count that into the performance upgrade for single core performance.

[Evildeffy]

It won't, it's probably the same deal as with Kabylake, 8700K single core boost is 4.4% higher than 7700k and they tend to count that into the performance upgrade for single core performance.

Assuming the leaks are correct and the Single Core Turbo is indeed 4,7GHz.
There first leak was 4,3GHz Single Core Turbo.

That said they also claimed Kaby Lake was 15% faster than Skylake @ 6700K - 7700K levels.
We know how that turned out as well.

[Life-Binder]

afaik Intel never claimed the 11% is from IPC only

just a flat 11%, frequency included



7700K with +2 cores would be just fine anyway, cores is only thing it lacks

Id take 6 cores @ 4.8+ over 4 cores @ 5.0+ any day, even for current games

[Evildeffy]

afaik Intel never claimed the 11% is from IPC only

just a flat 11%, frequency included

7700K with +2 cores would be just fine anyway, cores is only thing it lacks

Id take 6 cores @ 4.8+ over 4 cores @ 5.0+ any day, even for current games

My apologies I should've worded that differently.
I meant overall performance uplift, IPC included as well as frequency if the leaked clocks remain true.

You know my feelings on those clocks and I'm pretty sure you're in agreement with my assessment on that.

[Life-Binder]

well I know for a fact 7800X can reach 4.8 all cores 100% stable (no delid, no custom loop), so thats the minimum I expect out of 8700K OC

[Evildeffy]

well I know for a fact 7800X can reach 4.8 all cores 100% stable (no delid, no custom loop), so thats the minimum I expect out of 8700K OC

Doesn't necessarily have to reach that number as due to my previously mentioned reasoning as well.

But Intel was counting that as stock clocks, not OC.

[Gray_Matter]

If you build Intel, you have no compelling reasons to upgrade for 4-5 years anyway...

It's not just about upgrades. It's also about availability of replacement MB's. I have a i7-3770 and a i7-2600. Have you tried finding MB's for those suckers? Replacing sockets every year isn't great for keeping older equipment working.

[RuneDK]

More or less yeah. People should expect 4-5 years out of any machine built these days, even higher quality laptops. AMDs gains seem greater than they are because their chips were garbage for so long. We're approaching the limits of silicon and I don't expect anything miraculous without a fundamentally different manufacturing process.




Totally depends on your usage. If you're regretting not having more cores like I was when I upgraded from my 2500k then you could get basically anything. If you're looking for more gaming oomph, then you're going to be waiting for a while.

I game and do spreadsheets for work. I'm praying Canon Lake will make me want to upgrade. However I am going to leave my expectations down...it's been nice only having to add storage or upgrade the GPU for seven years.

[Thunderball]

You still can overclock the iGPU if it has it's own Frequency Parameters, but it still is preventing a system overclock if using the BCLK since that's tied into not the actual iGPU frequency but the "architecture" of it which is the CPU's, this causes catastrophic syncing issues among others.
Huge difference... try again and no you still cannot OC Kaby Lake on Z170 with external clock generators.

You seem to have no idea how iGPUs work (or even discrete GPUs without onboard memory), so further discussion is pointless.

PCIe bus is more forgiving than the iGPU's, still not connected in current tech and yet again irrelevant.

Of course it's connected, all buses are tied together on Intel mainstream platforms via DMA, and all of them are clocked via BCLK.

Nope, it was present in iGPU chips only because HEDT chips still could and they offered the same technologies.
Or is the HEDT line some sort of miracle of technology that disobeys all laws of electrical engineering.

PCIe Frequency was never a part of this discussion, it was raised only (by you) as a point of being tied to it and used as something that'll burn out your PC.
In all eventuality the BCLK increase will crash (potentially destroy) your iGPU/CPU before it can reach harmful PCIe speed levels.

There is a reason why HEDT platform chipsets cost more. They actually allow you to clock things separately. Again, harmful PCIe clock are over +5%, it's at least 10-15% for an iGPU and memory controller.

They actually did as performance difference was minimal going lower with speeds/power savings whilst saved power was very close to 1070s/1080s.
Following the same logic an APU can be extremely efficient as long as you don't try to put a 300W TDP chip in there.
Also don't be pedantic, APUs will never reach (with current technology levels) the performance a dGPU does, so stating "Also they unfortunately dont have an option to put as much stream processors to an APU" is repeating what my statement implied twisted into something to further your own argument.

They are unlike to reach 1060-like levels of performance but anything else is completely possible. Intel iGPUs have already surpassed 750 Ti performance and those have been traditionally extremely bad.

And your point would be wrong.
Let me ask you this, because apparently I have to dumb it down so you can understand: "Why is it that TDP is rated in Watts?"

I'll give you a hint:
It has something to do with the power draw of specific CPUs as it's tied to the amount of heat generated by said chip/CPU when it's consuming power.
Power draw is a very large part of the TDP rating, stating it's not makes me wonder if you are just trolling this thread now or not?

Please stop trolling. Just look up TDP definition, it's a rating designed to help you select cooling capacity for any given CPU. Sure, power draw is related to the amount of heat generated by the CPU, but as all measurements we have are done at the wall (or at best on a 12V rail) so it's pretty hard to how much heat CPU is dissipating because we dont know much it actually draws.

I honestly still cannot believe you still aren't seeing the point of why I drew that comparison for you.
I drew a stark contrast between a 65W TDP CPU by AMD and 91W TDP CPU by Intel to show you exactly what the TDP ratings are meant for.
And why I SPECIFICALLY made an example earlier of false ratings and possible consequences including that of Turbo clocks.
But it's fine, I've given up with trying to teach you, you will not (or simply do not have the ability to) comprehend it.

I dont see any contrasts here. Moreover, I dont see 40% heat output difference either.

Keep hoping for that magical wish for the i3s to be 100/200 series motherboards compatible, it's never been done in Intel's history and they aren't about to start now.
Even with the information being delivered to you by Intel's Press Release you still think otherwise.

Coffee Lake is 300 series, understand that already as well.

This is confirmed where, by whom? Or are you a soothsayer?

[Kagthul]

It's not just about upgrades. It's also about availability of replacement MB's. I have a i7-3770 and a i7-2600. Have you tried finding MB's for those suckers? Replacing sockets every year isn't great for keeping older equipment working.

I mean, that's fair. Its not a problem i've really had though. On a personal computer, ive never had a Motherboard fail in any computer, including some that i STILL have around that are 15-20 years old or older (I still have an old toaster box mac - two of them, actually. An original 128k and an SE/30).

For stuff i've built (helping my friends' business, and doing low-scale local freelance building and IT) which amounts to ~500 machines in the last 5 years, i think we've had.... four or maybe five motherboard failures, and one or two DoAs. All the failures happened well within the "still current and/or readily available timeline, and ALL of them were on cheaper motherboards (H-series boards or the like). And the.. i think two that died in the "boards aren't in production but are still readily available" timeline weren't full faliures, just a port dying that was inconvenient (bad USB ports, and we dont want to have to have a USB hub for that one rig we deploy at a tournament out of 200).

Thats anecdotal, i know, even at 500+ rigs, but i dont think it's a -massive- problem.

Doesn't mean it isnt one when it happens, though.

[Gray_Matter]

I mean, that's fair. Its not a problem i've really had though. On a personal computer, ive never had a Motherboard fail in any computer, including some that i STILL have around that are 15-20 years old or older (I still have an old toaster box mac - two of them, actually. An original 128k and an SE/30).

For stuff i've built (helping my friends' business, and doing low-scale local freelance building and IT) which amounts to ~500 machines in the last 5 years, i think we've had.... four or maybe five motherboard failures, and one or two DoAs. All the failures happened well within the "still current and/or readily available timeline, and ALL of them were on cheaper motherboards (H-series boards or the like). And the.. i think two that died in the "boards aren't in production but are still readily available" timeline weren't full faliures, just a port dying that was inconvenient (bad USB ports, and we dont want to have to have a USB hub for that one rig we deploy at a tournament out of 200).

Thats anecdotal, i know, even at 500+ rigs, but i dont think it's a -massive- problem.

Doesn't mean it isnt one when it happens, though.

Maybe I'm unlucky but I have seen it happen reasonably often. Most common is ports failing. Sometimes it fails because of outside factors like lightning or someone shooting a hole through the case by accident (I have seen the results of both). I used to work for an ASUS importer and ASUS gave us an extra 2% of boards as replacements for warranty swap outs. So, if you bought 100 boards then they would send you 102 and you would be responsible for all swap outs under warranty. That was pretty close to spot on and that's within the 2 year warranty period. Even if I work with that, we, as a household have 5 computers. At a 1 in 50 chance of a failure we end up with a 10% chance that one of the boards will fail. Not big but not small either. Especially when you consider that's just the odds without taking into account outside factors. You are also limited by memory availability. For example, AM3 supported DDR2 and DDR3. Fortunately RAM has a higher availability but it does start to get pricey as it gets older.

My point is just that Intel's rapid fire socket changing doesn't help consumers and it is something that they should be trying to avoid. At the moment, they are in a fortunate position where they can dictate a lot of things. That could well change with Ryzen and Epyc and that will only benefit consumers. Intel make good products but they have a habit of taking advantage of consumers and their near monopoly status which is not a good thing.

[Thunderball]

My point is just that Intel's rapid fire socket changing doesn't help consumers and it is something that they should be trying to avoid. At the moment, they are in a fortunate position where they can dictate a lot of things. That could well change with Ryzen and Epyc and that will only benefit consumers. Intel make good products but they have a habit of taking advantage of consumers and their near monopoly status which is not a good thing.

Agreed, but they dont have much choice considering their rate of improvement with each generation. It would be pretty hard to sell CPUs otherwise.

[Evildeffy]

You seem to have no idea how iGPUs work (or even discrete GPUs without onboard memory), so further discussion is pointless.

Keep trying there slick... you'll get it eventually.

Of course it's connected, all buses are tied together on Intel mainstream platforms via DMA, and all of them are clocked via BCLK.

The clocks are separated in current tech but sure go ahead, overclocking the BCLK according to you increase PCIexpress as well then right?

There is a reason why HEDT platform chipsets cost more. They actually allow you to clock things separately. Again, harmful PCIe clock are over +5%, it's at least 10-15% for an iGPU and memory controller.

..... Alright, apparently every Architecture from Bloomfield -> Westmere and Lynnfield and prior is HEDT, good to know.
Like I said before (several times now), in terms of forgiveness it's IMC -> PCIe -> iGPU but of course to you it's not true, sure that's fine... I mean your own example of LGA775 points to an overclock of between 120 - 140MHz being harmful which on LGA775 is already overclocked by 20MHz before you start

And like your examples 20MHz was actually doable if you wanted to stabilize above ~235 - 240MHz base clocks on the first Gen i7 and "technically but not really" 2nd gen i3/i5/i7 (Lynnfield) so that's already 20%, pretty much over your own estimate .. so yeah continue on, it seems you truly do know how it all works. /sarcasm.

They are unlike to reach 1060-like levels of performance but anything else is completely possible. Intel iGPUs have already surpassed 750 Ti performance and those have been traditionally extremely bad.

Relevancy of the statement? I never stated "ZOMG! GTX 1060 LEVELS BRUV!" I stated good efficiency when not pushed to all it's got.
And Intel's iGPUs are still weaker than the crappy APUs from AMD (Kaveri/Godavari etc.) ... do you really think they'll make a worse APU than they already have?
Good luck with that.

Please stop trolling. Just look up TDP definition, it's a rating designed to help you select cooling capacity for any given CPU. Sure, power draw is related to the amount of heat generated by the CPU, but as all measurements we have are done at the wall (or at best on a 12V rail) so it's pretty hard to how much heat CPU is dissipating because we dont know much it actually draws.

Power measurements are read at the EPS socket for specific CPU draw, not wall. (Here's such a tool: Click me!
You just made every Electrical Engineer turn in their grave yet again with that statement as you just stated every electrical device in the world doesn't work like that.
Hell Nikola Tesla and Thomas Edison should rise from the grave just to correct you on that.

I dont see any contrasts here. Moreover, I dont see 40% heat output difference either.

Like I said ... you either will not or actually CANNOT comprehend it.

This is confirmed where, by whom? Or are you a soothsayer?

Previously stated... history and added stupidity.
What is cheaper? Holding and maintaining 2 different uArchs and silicons or doing so with 1 being cut down?
I'll give you a hint... it's not the former!

It's an entirely new series, Core i3 is not called Kaby Lake R but Coffee Lake-S, why would they shame their own marketing and PR team by not making the Core i5/i7 not compatible with 100/200 series but actually allow their Core i3 to compatible?
It's Coffee Lake ... not Kaby Lake, it's simple logic and logic that has remained true with Intel for literally almost 30 years.

But no worries, they'll change just for you now.

[Lathais]

Id take 6 cores @ 4.8+ over 4 cores @ 5.0+ any day, even for current games

And that's about as much as the difference is between a R7 1600 and a 7700k, but you shit all over Ryzen. Ok, makes sense.

[Life-Binder]

And that's about as much as the difference is between a R7 1600 and a 7700k

wrong, but keep trying

maybe with Ryzen 2 your post will actually come close to reality



P.S. http://www.gamersnexus.net/guides/30...k-144hz-gaming

https://youtu.be/2_fAzBB_oAQ?t=7m19s

[Xjev]

So do we know when is it-8600k coming?

[Life-Binder]

october at earliest

[Zenfoldor]

It's not just about upgrades. It's also about availability of replacement MB's. I have a i7-3770 and a i7-2600. Have you tried finding MB's for those suckers? Replacing sockets every year isn't great for keeping older equipment working.

High end players stay in any game by:

1. Buying quality
2. Selling their old stuff

Resale of 2700k Intel chip+mobo vs old AMD chip, for example.

Quality is remembered long after the price is forgotten.