news intel warms up coffee lake with eighth gen desktop core details

Intel warms up Coffee Lake with eighth-gen desktop Core details

Thanks to an early leak, Intel is taking the wraps off its full eighth-generation Core desktop CPU lineup now. After a brief and refreshing stop at Kaby Lake for quad-core chips in 15W power envelopes, the next vista to behold is Coffee Lake. As has been widely rumored, the Coffee Lake CPU family will be the first to bring six-core CPUs to Intel's mainstream LGA 1151 socket.

The six-core Coffee Lake die. Source: Intel

The headlining chip of this lineup is the Core i7-8700K, a six-core, 12-thread part with a 3.7 GHz base clock and a 4.7 GHz single-core Turbo speed. Eighth-gen desktop Core i5 CPUs will have six cores and six threads, while eighth-gen Core i3s will have four cores and four threads. Intel will offer an unlocked CPU at every tier of its eighth-gen lineup, too. Here's the full list of what's coming October 5:

Model Base clock Single-core
Turbo Boost 2.0
TDP Total
speed (MT/s)
Core i7-8700K 3.7 GHz 4.7 GHz 6/12 95W 12MB Dual DDR4-2666 $359
Core i7-8700 3.2 GHz 4.6 GHz 65W $303
Core i5-8600K 3.6 GHz 4.3 GHz 6/6 95W 9MB $257
Core i5-8400 2.8 GHz 4 GHz 65W $182
Core i3-8350K 4 GHz N/A 4/4 91W 6MB Dual DDR4-2400 $168
Core i3-8100 3.6 GHz 65W $117

The fundamental cores of Coffee Lake CPUs remain identical to those in Kaby Lake chips. To get better performance in this CPU generation, Intel is tapping further 14-nm process improvements to put more (and faster) cores in nearly the same power envelope as past desktop chips. The Core i7-8700K and Core i5-8600K only get a 4W TDP bump over the Core i7-7700K and Core i5-7600K on the way to their higher single-core boost speeds, although base clocks seem to have fallen significantly to squeeze into that modestly more generous envelope.

The new chips will require Z370 motherboards to function. While the LGA 1151 socket is physically identical to the one on Z170 and Z270 motherboards, Intel says it improved power delivery for its six-core processors, improved package power delivery for better overclocking of those chips, and improved memory routing support to justify the mandatory motherboard upgrade. I'll have more to say about this in my full review of the chips, but I can't imagine anybody with an eight-month-old Z270 motherboard will be terribly happy to hear this news. Aside from the purported minor tweaks to the socket and memory routing, the platform itself is no more resource-rich than Z270. Builders will still get 24 chipset PCIe 3.0 lanes and 16 CPU PCIe 3.0 lanes for a platform total of 40.

Overclockers will find some new levers and knobs on Z370, at least. Coffee Lake chips will offer per-core overclocking, meaning that a single laggard core doesn't have to hold up the entire chip. Eextreme overclockers will also get memory multipliers for speeds up to 8400 MT/s, real-time control of memory latency settings, and better phase locked loop (PLL) controls.

Performance results remain under embargo for the moment, so I'll have more thoughts about eighth-gen desktop Core CPUs when our full review goes live October 5. Stay tuned.

0 responses to “Intel warms up Coffee Lake with eighth-gen desktop Core details

  1. Yes, there’s a verification process for many HPC sites, but it’s not quite as onerous as you make it seem if all that’s required is a recompile without any real code changes.

  2. Ryzen isn’t “good” in Farcry; it won’t even reach my monitor’s 100Hz refresh rate. You see that in other games, too. If you’re settling for 60Hz, then it might all seem the same to you.

    Also, you seem to be putting a lot of weight on one site’s review. TR’s own Ryzen review show the 7700K ahead in Crysis.

  3. Gonna be an unfortunate nope- Intel has said (or was rumored to have said?) that they’d support adaptive sync when they upgrade their graphics core, which like their CPU cores has stayed put through Skylake, Kaby Lake, and now Coffee Lake.

  4. Hmmmmmmm

    I see display post 1.4. I wonder if these will officially support vesa adaptive sync??

  5. Sounds like it might be an echo of Tick Tock…


  6. I agree. Power 5 had SMT about a year after Northwood, but it didn’t suffer the performance regressions that the P4 derived cores did with SMT in some workloads.

  7. Intel had claim to that numbering scheme actually. AMD were the ones that stepped on Intel’s toes on this one.

  8. From what I’ve read, they shouldn’t have to “offset losses” in the client area… the new 8000 series REQUIRES X370 mobos and cheaper ones won’t be available until early next year. So really only affects maybe high end gamers and to be honest, I’d probably go for an 8-core X-Skylake chip vs. the i7-8700K (even though I stick with AMD) Intel seems to be lowering the CPU prices, but “upping” the motherboard tax if you want to use it before next year. (almost like the game bundle tax on the Vegas)

    [url<][/url<] Hopefully, Intel won't do the same with IceLake next year for the Intel fans, but who knows?

  9. yeah, I’ll be making the move from an i5 to a Ryzen 5-ish part when the time comes, though when that time is, depends on how soon we start to see big enough performance increases going from 4C/4T to 6C/12T to make it worth the ~$400 entry fee (CPU/mainboard/RAM).

    But it’s so, so nice to see actual progress because of competition again.

  10. There was an article in the short bread a couple of days ago that showed similar results using PUBG. The testers used a Skylake X system and disabled access to cores for multiple test runs. In their case they found that at 6 threads they hit the point where the CPU was no longer the bottleneck — upping it to 8 threads had no further benefit.

    A lot of this will vary from game engine to game engine, but in general it appears that having more threads going forward will be beneficial (which is a recent development in the gaming world).

  11. You’re misquoting me. I said [i<]most [b<]modern[/b<] game [b<]engines[/b<].[/i<] That's not the same as [i<]most games[/i<]. Fallout 4, for example, was openly mocked because it was based on an ancient DX9 engine in 2015. Digitalfoundry specifically used farcry primal as an example of a game that liked high single-threaded performance and Crysis 3 as an example of a game that likes more cores. Those are representatives of the two extremes, so other games will fall somewhere in the middle: Ryzen is good in Farcry and good in Crysis. Intel is great in Farcry and bad in Crysis. Only one of these options is never bad, and the games coming out on newer engines are increasingly common, meaning that the intel quad cores are going to look bad more often.

  12. 18 months ago, AMD dumped a huge pile of info on Zen to the public. Intel knew Zen was coming and probably had a good idea of what it would be even before that. Right about now is when we should expect to see hints of Intel’s contingency plans.

  13. SMT was not premature. It was attached to core that didn’t have enough resources and had couple of brutal glass-jaw performance issues. (original trace cache)

    That’s why Intel reintroduced it only with Nehalem.

  14. You don’t need to thank AMD. You need to thank increased difficulty n scaling clocks and getting IPC improvements for that.

  15. Coincidence? Actually fairly likely. It’s simple. Intel needs to sell chips. When sales are stuck in replacements and no upgrades then Intel has problem. And considering clocks and IPCs were not exactly easy to come by, increasing core counts is fastest way to get xx% of improvement for marketing.

    Reminder: AMD cores are still ways behind Intel. AMD needs more cores just to match in most cases Intel’s fewer cores.

  16. Why is it with AMD it’s always “the next one is going to be awesome” ?

    The anticipation building while listening to the sizzling on the grill is nice and all, but somewhere along the line don’t they need to actually bring a steak to your table?

  17. This also explains why intel’s new HEDT chips are so uncompetitive with Threadripper on price. intel knows that the price of the CPU (and indeed, the hardware) is only a small fraction of the invested cost of HPC systems.

    Many HPC clients (nuclear, military, medical, aviation, finance) have very strict certification and validation requirements that would not be transferrable if a new architecture from a new vendor was dropped into existing nodes. Many big iron customers also have a lot of closed source vendor code that may have been compiled with intel compilers instead of GCC, compilers that look for “Genuine Intel(TM) processor” IDs and will downgrade code to unoptimized libraries if a flag is thrown, for, uh, “compatibility” and “safety” reasons (not entirely unwarranted, given recent AMD hotfix, but odds are low).

    (Imagines intel foreman with big hat and union badge walking in to sales meeting between AMD and existing HPC customer)
    “Say, that’s a nice Beowulf cluster ya got there. It’d be a shame if-”
    (Hefts giant wrench menacingly)
    “[i<]something[/i<] were to happen to it..."

  18. [quote<]Coffee Lake is something I think most people in the know saw coming, and they were just scratching their heads and wondering what Intel was waiting for. It's very likely that they simply learned from AMD's blunder and decided to watch how consumer software progresses on multi thread utilization before putting more cores in their mainstream consumer CPUs.[/quote<] Nice theory. Except intel introduced SMT prematurely, back when Hyperthreading was bad enough to cause slowdowns when enabled. More cynical theory: intel counts pennies and does the minimum amount of work to stay ahead. They lucked out with Sandy, when AMD stumbled, and essentially sat on that idea for 9 years.

  19. Zen was designed for more than the small, low margin, high end desktop sector. It will fare much better in server and mobile.

  20. I’ll pass on 1st gen Ryzen (including TR) every time. Now when they release Ryzen 2.0, fix the bios issues, and are able to provide the best performance I just might turn my head. The original Athlon was a great chip, but AMD made many blunders that allowed Intel and Nvidia to ride rough shot over them for too long. I need to know AMD is serious about what they do.

  21. Hyperthreading increases heat density since the average utilization of the physical core increases. Additionally, hyperthreading causes two threads to compete for the same resources on a physical core; this is why sometimes you get worse performance with it on.

    The 6 core setup is just better overall.

  22. Perhaps. They would like to have known, for sure. It’s interesting to think they might have spies gathering intel on AMD (pun not intended). Regardless, Intel has known for years and years that AMD’s strategy has been to get ahead on multi threaded processing. Problem is, AMD bet on it way too early and it lost them market share in some of the most profitable segments.

    That is not to say that AMD bet wrong, Intel would have been fools not to bet on it sooner or later. Also, it was well known that AMD was going to try and play catch up on IPC. Therefore, Intel took this time to boost their cores while they have a significant lead in both IPC and clock speed. AMD’s strategy may have had influence on Intel’s strategy, but this is not a knee jerk reaction from Intel in order to respond to Ryzen. Yes, competition is great, and it has kept Intel working. However, let’s be honest, if AMD came out with something that looked like Sandy Bridge in 2011, we’d have had far more competition.

    Coffee Lake is something I think most people in the know saw coming, and they were just scratching their heads and wondering what Intel was waiting for. It’s very likely that they simply learned from AMD’s blunder and decided to watch how consumer software progresses on multi thread utilization before putting more cores in their mainstream consumer CPUs.

  23. There’s a good chance that an i7 8700 may be the CPU to finally replace my 2500k eventually. For those wondering, rumors are saying that it will have the following turbo speeds:

    3.2 GHz Base
    4.3 GHz 6-core Turbo
    4.3 GHz 4-core Turbo
    4.5 GHz 2-core Turbo
    4.6 GHz 1-core Turbo

    Can anyone with a non-K Kabylake system explain how the turbo bins work on a board that allows you to mess with the Turbo frequencies? Is it possible to force the CPU to always run at its max turbo for each bin? If you could force this chip to never go below 4.3Ghz under load, no matter what, that would put it extremely close to the 8700K. I’ve done this with a 4790 on a B85 board. It runs between 3.8 and 4Ghz and never drops to 3.6 under load.

    I’m thinking the best deal for penny pinchers like myself would be in harvesting an i7-8700 (non-K) out of an OEM system at some point and reselling the system with a cheap Pentium or something.

    All the time I see standard non-K i7 chips in very reasonably priced OEM systems. Like this HP i7 7700 system, for example:
    [url<][/url<] A system like that with an 8700 is just begging to have it harvested and swapped out for something cheaper, then resold. You could throw a basic entry\mid level gaming GPU into a system like that with a more affordable CPU and have a very robust entry level gaming system that someone would pay a decent amount for (lots of people prefer to buy locally from individuals). I've gotten basically free i5 and i7 CPUs that way recently. Really, the worst part right now is that RAM prices are still horrible.

  24. I think it’s very likely Intel knew what was coming from Ryzen long before it was released.

  25. The Intel 8350 has a “k” by it, that makes it better by a thousand times if my math is correct 😉

  26. This is true. But the pricing and model/feature mix might have been… [i<]adjusted[/i<]... to account for the new reality.

  27. The problem with this kind of thinking is that Coffee Lake has been on Intel road maps for longer than we’ve known Ryzen’s official core counts. I’m fairly sure Coffee Lake was on Intel Road Map leaks for around a year before Ryzen was launched.

    So no. These processors are not a response to Ryzen. They can’t be considering they existed before Ryzen, back when all we knew about AMD’s new chip was that it was the Zen µArch. The subsystem, core counts, and cache hierarchy had to have been finalized over a year before Zen’s launch just to have the masks and submasks created in time for testing, mass production, and product launch in Oct of this year.

  28. Makes sense then, it doesn’t make it any less confusing. I wonder if they will continue using the B*50 nomenclature for mainstream boards seeing as B350 is taken by AMD. Also, X370, Z370… Hmm.

  29. I wouldn’t say that the R5 1600 is comfortably faster than the i5 7600K in most games. In TR’s own reviewing, the 7600K was tied or beat the *R7 1800X* in every game but Crysis3 (rather niche) for quite a bit less cash.

    [url<][/url<] Edit: Looking at your video only two of the fives games favored Ryzen R7 over i5.

  30. Intel has high core counts for years – in the server space. Ryzen has forced them to bring higher core counts to their consumer line, and reduce their prices. So yes, thanks AMD.

  31. The base clock of the 7700K is still higher. So I don’t think the 8700K completely trounces its predecessor. The 7700K may surprisingly hold its own but of course we should see some real benchmarks.

  32. The mere fact that AMD has chosen to sell chips with 50% of the silicon dark isn’t proof that things are bad for AMD. It could also mean they have a ton of chips, and they wanted to access all price points. If so many people wanted 8 high-margin core models that they were sold out, and then AMD introduced 4 core models, that would suggest problems to me.

    Of course things would be better for AMD if there was insatiable demand for high-margin chips, but thats not the market. There are a certain amount of customers at each price level and in each time period.

    I don’t think I have enough information to make any strong conclusions.

    The financials for next quarter will be interesting, though I don’t see a reason for them to ship a great deal more product than the past quarter.

  33. The substrate is still thinner as well. It’s not just size, it’s size and much thinner substrate. I’m fairly sure that as long as that substrate is in use the die would have to be absolutely massive to safely use solder on it. Hell, they don’t even solder the HEDT parts any more because even the 8 and 10 core parts are still smaller than some of the older generations 4 cores.

  34. The boost clock is not a guarantee that it will be able to hit that clock rate within the specified TDP. It will vary depending on the workload, but there is no chance an 8700 will be able to sustain 4.6 GHz at 65 W when running a heavy multi-threaded workload like Cinebench.

  35. Unlike AMD, Intel has separately-manufactured dual-core and quad-core dice, and different cores for different iGPU configurations, too. So how many i7s get repackaged as Celerons? not many, if any at all.

  36. We could look at Intel selling Celerons and Pentiums as bad for them also. How many Intel dies cover Celeron to i7 desktop range? All AMD did with R3 was disable SMT, so it’s not an issue of those chips having die defects. All those $110 R3 1200s could have been R5 1400s (at least).

  37. Two weeks ago, Fry’s had the i3-7300 for $110. I snagged one online then drove down to pick it up. 4.0 GHz and 4 MB L3 cache. BlueIris monster.

  38. [quote<]The AMD CPUs have basically a hard wall at 4GHz, and they're already not as fast in single-threaded performance-per-clock as Skylake/Kaby Lake. Coffee Lake doesn't extend that, of course, but the gap exists. [/quote<] I agree - no surprise. [quote<]The $180-ish i5 still hits a turbo clock of 4GHz. That's going against a 4-core, 8-thread Ryzen 5, and I have a feeling that 6C6T will beat 4C8T pretty handily.[/quote<] Yeah, SMT is overrated for performance, although extra threads CAN be helpful. Sometimes. Like when virtualizing, which Intel would like to pretend people don't do "on the desktop." So I used Ryzen 5 1600 for my home VM server, and Kaby Lake i5-7600K for my home desktop. [quote<]And I say all that as someone who paid good money for a Ryzen system in March. I'm going to use the crap out of it, for sure, but Coffee Lake takes off some of the shine.[/quote<] It always loses that shine about 6 months in. 🙂 I would really like to replace my main office PC's 4.4GHz Richland APU with an i5-8600K. I would prefer to buy AMD. I just don't need 12 threads that Ryzen 5 would offer. 6 really fast threads (yes, it will be OC'd, it always is) is very much the ideal situation for what I do daily.

  39. I’m still kind of hoping that at some point Intel will release an unlocked 4C/8T version of this (with full-cache). Basically an i7-7700K++. It would be a good use of rejected 8700K chips where only two of the cores are defective and disabled, but the rest work just fine. A chip like that (along with the socket/platform improvements) should be able to reach really high clocks without generating as much heat as the 6-core 8700K would (at the same clock speed).

  40. Agreed. I think it exists because they either have too many faulty dies or they can’t sell them at higher prices. Either way that’s bad news for AMD.

  41. I think Ryzen 3 wouldn’t exist if there was a problem supplying dies for Epyc. They can’t have so many with >2 defective cores, and even if they did, even they can be sold in packs of four at a good margin.

    There must be a limit to how fast they can test the server parts though. I wonder what that testing infrastructure even looks like.

  42. Well, the products thus far available are aimed at large computers, which do not represent the meat of the market. So I guess we need to wait and see.

    I wonder how the total revenue compares between their Ryzen and above-Ryzen product lines. Some of the Threadripper parts are (to my ignorant eyes) possibly lower margin than the Ryzen 1800X, depending on how much margin is lost to just getting an item out and sold, and how much is tied up in manufacture and testing.

  43. [quote<]WOW, that's some fully-baked engineering![/quote<] And here we thought Intel's investment in the Boulder design bureau would go up in smoke.

  44. Possibly… But that 10-15% is probably wishful thinking. Didn’t the marketing slide mention “up to 10%” increase in performance compared to current 16nm processes? If accurate, we’re probably looking at closer to 5% increase in performance due to the 12nm GF LP upgrade. But maybe we’ll be pleasantly surprise!

  45. WOW, that’s some fully-baked engineering!

    Same 14nm process with 4.6 GHz (boost) 6/12 for the 8700, and only 65 watts..

  46. They’ll have to really sell them in volume. I mean, they’re excellent performers relative to similarly-priced Intel parts (as STH’s in-depth benchmarks show), but are they really going to move enough to make up for taking losses on mass-market products? They’re better off scrapping the mass market altogether for now and sell every eligible die as an EPYC CPU if there was really that much demand.

  47. If they can move enough expensive Epyc parts to get their “enterprise & embedded” division turning a large enough profit then that could offset losses from noticeable price cuts in the regular client division.

    Then again, offsetting losses doesn’t exactly sound like the strategy you want to have to take over the world.

  48. The rest of that story is that the i7-7700K [url=<]is priced at $279[/url<]. You don't think they'll discount 8th-gen parts, too?

  49. They were [url=<]teetering on the brink of profitability[/url<] during the first full quarter of Ryzen availability. Dropping prices does not seem like an option, at least not as much as Demetri wants.

  50. For the millionth time, Intel isn’t selling CPUs to please overclockers. They’re selling CPUs to please businesses who will buy by the tens of thousands, not twenty-thousand overclockers who will buy one chip.

  51. Meh, AMD will need to reduce their prices, but they can sell at a profit nonetheless. Before too long they can stop trying to sell Zeppelin in the 4C segment in favor of Raven Ridge, presumably the better models of that will clock quite similar to Core i3. I would certainly favor 4C/8T Zen with the AMD GPU over quad i3 (quite possibly including the K version). Intel is not so overpowering here.

  52. But if they added 2 more cores to every CPU, does this mean the dies are big enough for them to use solder again????

    *asking the real questions*

    EDIT: [url=<]wccftech[/url<] says 149mm^2 for 6-core Coffee, which is still smaller than sandy (216 mm^2), so... maybe no?

  53. I wonder if they are going to bother making 2C dies anymore. Maybe from here its all 4C and 6C parts, mobile and desktop.

  54. A 3570K that I have gently running at 4.0 Ghz. It’s fast enough for almost everything I do, but I recently upgraded to a 144 Hz display, so I’m a little bit CPU bottlenecked with some games. I should get a boost not just from the new CPU but also from 16 GB of 3200 Mhz DDR4.

    So far I’ve been building new PCs roughly every 5 years. The only thing keeping me from upgrading is the fact that RAM prices are very high right now. I could wait another year and hope prices go back down, but I have a feeling that RAM will remain elevated for the foreseeable future.

  55. Same IPC and architecture as Kaby, so in terms of performance there is no difference. it’s still going to be tested with the same RAM as Kaby too.

  56. [url<][/url<] Read and understand. Then please stop making this statement. It's not the TIM it's the attachment process for the heat spreader that creates the problems.

  57. I heard they needed more current to feed 2x the number of cores (in some cases), so couldn’t (yea, we all know it’s *wouldn’t*) reuse Z270.

  58. To the best of my knowledge, there are no Cannon Lake desktop chips. They don’t appear on any 2018 road maps. And a lack of desktop Cannon Lake is also consistent with what we know about Intel’s 10nm problems with the basic 10nm process hitting it’s low power and frequency targets but not scaling up properly at all.

    There is however a theoretical 8 core Coffee Lake Refresh chips comming on a new Z390 platform in 2H 2018.

  59. Unless I’m mistaken, AMD is betting on the 12nm GF LP process for their Zen+ launch, which ought to give them between 10-15% extra process releated performance along with whatever minor tweaks they see fit to do to the basic Zen microarchitecture.

    If the chips top out at 4.2 GHz XFR now, that would put them in the 4.6-4.8 GHz range in a best-case scenario for the same TDP.

    A R5 16xx Zen+ @ 4.5 GHz is going to be pretty competitive with the i7-8700K is my guess — especially if you factor in that the $100-150 saved on the CPU+Mobo can be put towards a better GPU.

  60. [quote<]Thank you AMD for forcing Intel to up their core counts.[/quote<] Why does everyone keep saying this? Do you know how long ago a chip like this would have to be designed for it to come to market now? It's a several year old design. Intel just took a smaller risk on multi-threading than AMD has in the past. This is the same reason it took AMD so long to give us Ryzen. After going all-in on multi-threading with Bulldozer, they have known for years that its biggest flaw was IPC, but there was really nothing they could do about it but minor tweaks until a brand new architecture made it past the production phase and out to market. This process takes several years. The only thing AMD can force Intel to do is compete on value by dropping prices, and so far they have really only given Intel a small nudge. Why? Because AMD can't afford to drop prices very low. AMD does beat Intel and Nvidia both on value for certain applications and certain niches of the market, which is how they are still in business. Unfortunately, for mainstream products and gaming they are still the underdog. It will be a few more years before they force competition on performance.

  61. The TDP is only 4W higher, I highly doubt that mandates an all new power delivery system.

    Instead of forcing a new socket to “improve overclocking”, how about they replace their garbage TIM and use solder like real CPU vendors do, drop temps up to 30C, and easily get better OC on lower voltage. Watch the recent GamersNexus video on it.

  62. Or you know the part where they reworked the on chip power delivery and on chip memory subsystem to match the improvements on the new motherboard and chipset.

  63. The marginal cost of another die isn’t the issue. The issue is all the highly-paid R&D and overhead costs associated with running the company, which the product sales have to pay for somehow.

  64. [quote<]Fingers crossed that Cannon Lake or Ice Lake bumps the core count to 8, if not some kind of IPC improvement.[/quote<] Is Cannon Lake using the same core architecture as Sky/Kaby/Coffee Lake? Or is it a new core? I remember the good old tick-tock model when the second generation on a new process was a new architecture, but since it's been tick-tick-tick-tick-tick... I was hoping Cannon Lake might break from the past and have a new architecture? Or do I have to wait for [url=<]Ice Lake[/url<]? Is that with a new core architecture?

  65. Well, 16 lanes of PCIe 4.0 is equivalent to 32 lanes of 3.0.
    And all I hear about lately is how terrible it is to not have enough lanes of PCIe.
    So I figure doubling the throughput and the number of effective lanes would be a good thing.

  66. [quote<]Intel will offer an unlocked CPU at every tier of its eighth-gen lineup, too.[/quote<] Does this mean I can expect an unlocked 2C/4T Pentium G5650K?

  67. Maybe AVX-512 isn’t such a great idea unless you really need it. It seems having to beef everything up for AVX-512 has really hurt in the efficiency department even in non-AVX workloads.

    I’m with you on the core counts but why PCIe4? Is 3.0 limiting you in some way?

  68. Would be easy though to just drop the R5 1600 to $180, and compete on the fact that R5 has hyperthreading, and you can overclock it on the stock cooler on a $70 mobo and only be at the 5-10% IPC disadvantage in single thread. The whole lineup will need to be adjusted that way.

  69. This is why I’m very excited to see how Coffee Lake performs with independent tests such as TR’s. I think the smart money is on these chips. If I was in the market for a new system, this is what I’d get.

    Fingers crossed that Cannon Lake or Ice Lake bumps the core count to 8, if not some kind of IPC improvement.

  70. 3.6GHz is a tad low IMO but I forgot to consider the K series don’t come with a cooler. So yea with that in mind the 8100 definitely makes more sense.

  71. I think I read somewhere that it’s Q1 or H1 2018. AMD’s release dates have historically had ~12-month intervals between them and Ryzen 1.0 came out in March, as we all know. So a Q1 release of Ryzen Episode 2 isn’t all that far-fetched especially if this ‘new’ 12nm node with which it’ll be made is more optimism than reality.

  72. Papermaster was saying they’re going to move both Ryzen and Vega to GF’s “12nm” process next year.

  73. The 8350K isn’t interesting to anyone but overclockers, otherwise it’s simply not worth the more than 40% price increase for the 11% performance improvement over the i3-8100.
    The i3-8100 however should be the new king of budget builds.

  74. If it is actually being made on GloFo’s “12nm” process that was just reported here the other day then I’d expect it to be out in Q3 of 2018 or so (maybe Q4 if an unexpected delay occurs).

    If it is still being made on the same 14nm process as RyZen v1 then it could be out sooner.

  75. I haven’t heard anything credible. All I have to go on is adding 12 months to Ryzen’s release which is March 2018 but your point about Raven Ridge still pending and thus pushing back Ryzen 2 is probably sound.

  76. The 8600K looks like a decent value, assuming you can get all six cores to a stable overclock of 4.5+ like you can with the previous 4-core X600K parts.

    Thank you AMD for forcing Intel to up their core counts.

  77. Do we have even a rumoured date for Rzyen Mk.2 yet?

    I get the feeling that it’ll be later in 2018 given that we don’t even have Raven Ridge APUs yet, and they’re the mainstream product that AMD needs to sell a lot of.

  78. [quote<]Waiting for the 1st revision or respin of Ryzen then I'll jump in![/quote<] I'm 100% with you. As much of a leap as Zen was over the Construction Machine cores, I've no doubt that AMD had a solid list of meaningful improvements that didn't make the deadline for this iteration. Coffee Lake vs Ryzen 2 should be very interesting!

  79. The AMD CPUs have basically a hard wall at 4GHz, and they’re already not as fast in single-threaded performance-per-clock as Skylake/Kaby Lake. Coffee Lake doesn’t extend that, of course, but the gap exists.

    The $300 i7 tops out at 4.6 GHz, something Ryzen can only dream of. Ryzen 7 has a couple more cores and threads to go with it, but the much-lower single-threaded performance of Ryzen will only break even on heavily-threaded tasks.

    The $180-ish i5 still hits a turbo clock of 4GHz. That’s going against a 4-core, 8-thread Ryzen 5, and I have a feeling that 6C6T will beat 4C8T pretty handily.

    And I say all that as someone who paid good money for a Ryzen system in March. I’m going to use the crap out of it, for sure, but Coffee Lake takes off some of the shine.

  80. I suspect I’ll be grabbing Ryzen’s respin myself (I’m on Devil’s Canyon atm) but it’s still okay to jump on current Ryzen right now:

    Given how the Ryzen 5 1600 is comfortably faster than a 7600K in most modern game engines, it’s an indication that four-thread CPUs are dead, and even an eight-thread 4.8GHz 7700K’s framerates dip below that of a stock Ryzen 5 1600X in some heavier, thread-hungry games ([url=<]look here[/url<] at Crysis 3 as just one example; Even though the 4.8GHz overclocked i7 is getting similar fps numbers to the Ryzen 5 in the toughest scenes of that segment, look at the frame time graph where the i7 is a mess and the Ryzen 5 is a smooth, stable line; That's enough cores vs not enough cores). I think the average framerates are still going to favour Intel's strong IPC and boost clocks but their decision to lock the products with 8 or more threads behind a $303 pricetag is shortsighted, given that most current and practically all future titles are multi-core optimised and that the current console generation has 8 cores (not 6). People care about games NOT stuttering during heavy scenes. They don't care if they get 130 or 180 fps when they're staring at an empty skybox, they just don't want their framerate to tank under pressure. 12 threads is definitely better than 4, comparing to the 7600K. 6 cores of the i5-8400 may be enough to match Ryzen 5's 6 cores, it may not - We get to find out if Intel disabling HT on the i5 is a mistake or not and that's why I can't wait for the reviews! If reaching 8 threads is the deciding factor, Intel won't give you that for anything less than $303, whilst AMD's cheapest Ryzen 5 is $165. As always, the top-end i7 K-series chip will be the one to buy if money is no object, but the sweet-spot price range in Intel's product stack looks to have been ruined through silly locked-multiplier/castrated-cache/missing-SMT product segmentation yet again. I'm looking forward to the full TR frame-time analysis as always. If Ryzen 5 still looks good, I'm happy. If Intel have made huge leaps forward at the popular $200 mark, I'm also happy. Isn't competition great? 🙂

  81. AMD will have to drop prices, but I wouldn’t say Coffee Lake blows them away. The lower half of the lineup tops out at 4.0ghz; it’s the top couple parts where they’re at a big clock speed disadvantage that could hurt ’em the most. R7 will have to drop quite a bit.

  82. And it will be another year before we start seeing b350 itx boards. Also, yeey, 3 generations of cpu’s on the same socket with different chipsets that don’t mix and match…

  83. I guess that any review including a previous-gen i7 is basically going to tell everything you need.

  84. Thanks AMD indeed. The “Intel has to compete against themselves” monopoly defense still has many adherents, but after all these years, it’s funny we were stuck at 4 cores in the consumer range until suddenly AMD got back in the game. Pure coincidence, I’m sure they’ll tell us.

    Never mind AMD brought 6 Thuban cores to us all the way back in 2010, so it’s clearly not been an engineering issue that Intel’s engineers couldn’t overcome.

    Waiting for the 1st revision or respin of Ryzen then I’ll jump in!

  85. If you strike down AMD’s 8 core parts now they will become more powerful than you can possibly imagine!

  86. The lower end of each line is going to make Ryzen look pretty bad. The clocks are fast enough and the single-threaded performance is good enough that the $303 Core i7-8700 should soundly beat the $300-ish Ryzen 7 1700, and when that AMD chip is OC’d to its 4GHz wall it’ll probably just break even.

    So it’s been fun, AMD, but it’s back to the nine-digit losses with you.

  87. Man, that’s not just recursion, that’s borderline fractal. Also, we’re lucky that the plural of fish is fish.

  88. Looks like a nice lineup with a surprisingly small price bump over the previous generation. The i3 8100 is certainly going to be included in a lot of budget buying guides. A real quad-core with a set 3.6Ghz clock is quite a chip for $117.

    Looking at this list makes me wish BCLK overclocking was feasible.

  89. What’s really going to bake your noodle later on is, would you still have commented if I hadn’t said anything?

  90. Indeed. Best case scenario, it doesn’t work and you have incompatible parts. Worst case scenario, it doesn’t work and you have incompatible parts, at least one of which is now fried.

  91. For today’s lesson in recursion, drfish will wield the ban hammer on drfish for insulting drfish about an article written by drfish.

  92. Closer to making me want to upgrade.

    Still needs AVX-512 support + PCIe 4.0 at the consumer level. 6-8 cores sounds about right for most uses that I have right now, so the core counts are getting there.

  93. Hey! No name-calling or insults around here! Even if it’s against yourself! Heed this warning or face the Mighty Ban Hammer!

  94. Man, I totally missed the core count revolution when I built off a 7700K. Not upset at all; this thing is still blazing fast for everything that I do.

    Enjoy your copious cores, you CPU crusaders!

  95. Kind of strange not seeing a 4 core with HT.
    Maybe that’ll come later…………….
    Would be interesting to see a 4 core with HT benched against a 6 without……………….

  96. The 8350K should be terrific for those on a budget.

    The 8700K will be the ultimate gaming CPU. I wonder how well this new 14nm process overclocks.

  97. I better hope these CPUs are keyed different from their older Socket 1151 stuff or we are going hear ignorant people trying to get these chips running on older motherboards or vice versa.

  98. I addressed this in the piece, but Intel is now aggregating both CPU and chipset PCIe lanes when it describes those resources. The truth is that PCIe lane count remains unchanged between Kaby Lake and Z270 and Coffee Lake and Z370.

  99. Only $10 more than the 6700K launched for. Looks like the rumours of a price hike were thankfully unfounded! Great to see core counts finally going up.

  100. “This new family introduces the first-ever 6-core Intel Core i5 desktop processor and first-ever 4-core Intel Core i3 desktop processor.”

    And we can all thank AMD for that.

  101. Things are about to get a tiny bit confusing in the CPU space, as Intel’s entire range is shaken up and they’re revising core counts after an eternity of hiatus.

    Thanks, AMD!! And for that I’m gonna get Ryzen!!

  102. Intel doesn’t need to respond to Ryzen more than they need to. These 6-core models take the fight against the 1600X and 1600. The 8700K, for example, is priced about 50% higher than the 1600X and probably won’t deliver 50% more performance [u<]but[/u<] they know they have more brand recognition and they know customers will be willing to pay a bit more for their CPUs. They're weighing Ryzen very carefully and they're responding only as much as they need to.

  103. The only things Intel did on so-called 8th cpu are:
    1.Add one more “+” below the 14nm+ process;
    2.Add more cores into the cpu;
    3.Increase Clock.

  104. The thread count is listed as 12 for i5s in the table.

    I can only imagine how well this I’ll do in 3D Benchmarks. I’m can’t wait to see what this will do under LN2.

  105. Not likely. It is just saying that platform supports “up to” 40 lanes. I very much doubt you’re going to see that many lanes on mainstream.

  106. From Intel’s announcement:

    “This new family introduces the first-ever 6-core Intel Core i5 desktop processor and first-ever 4-core Intel Core i3 desktop processor. The family offers a wide range of performance options for consumers with unlocked1 “K” processors that deliver maximum tuning flexibility at each brand level and up to 40 platform PCIe 3.0 lanes for system expandability on graphics, storage and I/O.”

    Huh. 40 PCIe lanes on mainstream platform?

About Renee Johnson