Intel Skylake / Kaby Lake

[CHADBOGA]

ComputerBase tested the long awaited Pentium with HT support. Basically the same performance as a Core i3-6100 (97%) for $64 MSRP!

ComputerBase: Intel Pentium G4560 Tested

- Applications Performance (Windows):
15% faster than $90 FX-6300
21% faster than $60 Pentium G4400 (predecessor)
23% faster than $149 A10-7890K

- Gaming Performance (1080P):
16% faster than $89 FX-6300
26% faster than $60 Pentium G4400 (predecessor)
26% faster than $149 A10-7890K

Ps: That's overall, but in games like GTA V it delivered almost 3x Pentium G4400's minimum framerates.

- Fury X Tests: Ahead of all AMD CPUs/APUs in Total War: Attila and The Witcher 3
- GTX 980 Ti SLI Tests: 24-45% faster than FX-6300

- 23W idle/load difference, down to as low as 16W with a 0.15V undervolt



www.computerbase.de/2017-01/intel-pentium-g4560-test-kaby-lake

That is crazy good value.

Even though I have no need of such a CPU, seeing such a bargain like this is almost enough to tempt me to buy it just for the hell of it.

Almost.

 

[Shivansps]

https://www.youtube.com/watch?v=oy7VLV5Knxw

 

[Arachnotronic]

ComputerBase tested the long awaited Pentium with HT support. Basically the same performance as a Core i3-6100 (97%) for $64 MSRP!

ComputerBase: Intel Pentium G4560 Tested

- Applications Performance (Windows):
15% faster than $90 FX-6300
21% faster than $60 Pentium G4400 (predecessor)
23% faster than $149 A10-7890K

- Gaming Performance (1080P):
16% faster than $89 FX-6300
26% faster than $60 Pentium G4400 (predecessor)
26% faster than $149 A10-7890K

Ps: That's overall, but in games like GTA V it delivered almost 3x Pentium G4400's minimum framerates.

- Fury X Tests: Ahead of all AMD CPUs/APUs in Total War: Attila and The Witcher 3
- GTX 980 Ti SLI Tests: 24-45% faster than FX-6300

- 23W idle/load difference, down to as low as 16W with a 0.15V undervolt



www.computerbase.de/2017-01/intel-pentium-g4560-test-kaby-lake

Talk about value for the money. Wonder why nobody is talking about this gem.

 

[frozentundra123456]

Talk about value for the money. Wonder why nobody is talking about this gem.

Well, intel should have done it a couple of generations ago, as well as making more cores/threads through the entire stack, including a mainstream hex core. Now the people who are interested in performance want more than two cores, and the light users probably don't even know the difference.

 

[jpiniero]

Well, intel should have done it a couple of generations ago, as well as making more cores/threads through the entire stack, including a mainstream hex core. Now the people who are interested in performance want more than two cores, and the light users probably don't even know the difference.

Intel was never going to do this until it made sense in mobile... and if you believe the rumors Cannonlake would have brought 8 cores (plus maybe an full ondie PCH?) a year ago. The fab delays plus the sliding PC sales have sure made an impact.

 

[Arachnotronic]

Intel was never going to do this until it made sense in mobile... and if you believe the rumors Cannonlake would have brought 8 cores (plus maybe an full ondie PCH?) a year ago. The fab delays plus the sliding PC sales have sure made an impact.

Full on-die PCH was never going to be in the cards for desktop CPUs.

Or Cannon Lake for that matter.

 

[frozentundra123456]

Intel was never going to do this until it made sense in mobile... and if you believe the rumors Cannonlake would have brought 8 cores (plus maybe an full ondie PCH?) a year agredo. The fab delays plus the sliding PC sales have sure made an impact.

Maybe if intel had offered more value instead of stubbornly segmenting the market to the nth degree to try to maximize margins, they could have mitigated the slide a bit at least. Just look at the demand for Zen. Now admittedly a part of it is from raging AMD fanboys/Intel haters, who would never purchase from intel no matter what the value, but there are also a lot of other posters who are planning to upgrade because it seems Zen will finally offer sufficient value to make it feasible to upgrade from Intel's quad cores. Now whether or not it really does that is still debatable depending on final pricing and a wider range of benchmarks, but it certainly shows the demand is there.

 

[.vodka]

Well, intel should have done it a couple of generations ago, as well as making more cores/threads through the entire stack, including a mainstream hex core. Now the people who are interested in performance want more than two cores, and the light users probably don't even know the difference.

You've hit the nail right in the head with this. Intel's market segmenting is nonsense lately. This CPU is perfect for those average joe builds, a good step up from cheap G1820/H81/DDR3! I like it. Hell, this might even make me change that old $5 E8400 spare parts backup rig once it is available for sale, a nice performance increase and lots less power consumption.

I wonder how much of the buzz around Zen has to do with this $64 gem being released. Talk about timing... Would Intel'd even bothered with enabling HT on the Pentium line if AMD didn't compete anymore?

At least until AMD releases Raven Ridge, this G4620 that also decodes HEVC is something they can't really touch with Bristol Ridge at least on CPU performance.

 

[jpiniero]

Maybe if intel had offered more value instead of stubbornly segmenting the market to the nth degree to try to maximize margins, they could have mitigated the slide a bit at least. Just look at the demand for Zen. Now admittedly a part of it is from raging AMD fanboys/Intel haters, who would never purchase from intel no matter what the value, but there are also a lot of other posters who are planning to upgrade because it seems Zen will finally offer sufficient value to make it feasible to upgrade from Intel's quad cores. Now whether or not it really does that is still debatable depending on final pricing and a wider range of benchmarks, but it certainly shows the demand is there.

Except you are talking about the enthusiast/gamer market, which is tiny. Growing, but tiny.

 

[Arachnotronic]

Maybe if intel had offered more value instead of stubbornly segmenting the market to the nth degree to try to maximize margins, they could have mitigated the slide a bit at least.

Enthusiast/gaming market is growing, in some cases (like notebooks), 30%+. The parts of the market that are sliding are the ones that aren't performance sensitive to begin with.

Just look at the demand for Zen.

What demand? There is a lot of "enthusiasm" from a few people on internet forums for what their IDEA of Zen products will be ("half the price of Intel counterparts, with equal or better performance!"), but don't confuse the postings of a small but vocal group of anonymous people on the internet with "demand," especially since there are so many unknowns about the Zen CPUs (you know, like pricing...).

If you used sentiment of people on enthusiast forums as gauges of demand, NVIDIA would have 20% dGPU market share and shrinking, AMD would have 80% and growing, and "K" series CPUs from Intel would be declining at a double-digit rate year after year because "my Sandy Bridge is good enough" and "why buy a new CPU/mobo, when used X58 boards and LGA 1366 Xeons are the way to go for perf/$!"

The reality is far different, so using what people say on these boards or similar boards to try to get any sense of the market-wide demand for any product, even one tailored explicitly at enthusiasts, is going to be pretty much a waste of your time.

because it seems Zen will finally offer sufficient value to make it feasible to upgrade from Intel's quad cores. Now whether or not it really does that is still debatable depending on final pricing and a wider range of benchmarks, but it certainly shows the demand is there.

Why didn't these people buy the 5820K for $389 and a cheap X99 board if more cores, more memory bandwidth, and more PCIe lanes (all part of "value") was so important to them? Because the reality is more cores just isn't that useful for gaming and the people who do want more cores would have bit the bullet and picked up x79/x99 systems over the last several years.

Again, refer back to my comments about conflating the comments of vocal internet posters, especially people who have openly admitted that they pretty much loathe the fact that there's anything Intel inside of their systems but AMD's complete neglect of this market led them to go with Intel anyway, with what's going on in the market.

 

[frozentundra123456]

If you read my entire post, instead of selectively quoting it, you would see that I *did* say that what Zen will offer is still debatable depending on independent benchmarks and pricing. And I am really getting weary of the HEDT platform (older architecture/process server rejects on an expensive platform with no upgrade path from mainstream processors) being used as an excuse for not offering a mainstream hex core. Are you really trying to say that Broadwell E is a good substitute for a Kaby Lake hex core with increased ipc and better overclocking on the revised 14nm+ process?

And I think it is obvious from my post history that I am as annoyed as you apparently are at some of the hype for AMD on these forums. But that does not change the fact that there is a market for more cores on the mainstream with a cheaper platform and the most current architecture and process. Now granted, Intel may be moving in the right direction with hyper-threaded Pentiums, overclockable i3s, and a possible mainstream hex core with what, icelake or whenever. (Their roadmap is so muddled that I dont really know what is coming when.) But the point is, they should have done all this earlier. And instead of increasing cores(the obvious way to increase performance since frequency and ipc have pretty much topped out), they are still trying to ring more performance out of dual cores.

 

[CakeMonster]

As I've said before, I have the money to shell out for HEDT, but Intel is not giving me what I want. I game primarily, and being two generations behind with lower clock speed is simply not an option, no matter how many cores they throw in.

 

[DrMrLordX]

Intel should have been selling mainstream hexcores as of Skylake.

 

[Arachnotronic]

Intel should have been selling mainstream hexcores as of Skylake.

Then people would have complained that KBL and CFL aren't big improvements.

The core count increase happens once, then you (Intel) are back to where you were, with entitled enthusiasts complaining that you don't spend any money on R&D and all your engineers are actually just throwing paper airplanes at each other in the rec room.

 

[HilbertSpace]

all your engineers are actually just throwing paper airplanes at each other in the rec room

That's a great quote, it made me laugh.

We need to accept that we are reaching the physical limits (that is, the limits imposed by physics) on silicon transistors (we are getting close to quantum effects dominating at smaller process sizes). The next BIG step in computing will either use different materials (graphene, silicene, spintronic materials) and/or be based on quantum computing. I doubt these advances will come from Intel, Apple, AMD, etc. The fact that D-wave can come up with a quantum computer (albeit severely 'limited' with respect to 'general computation') that can compete with the latest CPU's in very specific test cases, still shows that a start-up can come along and disrupt the whole industry. It has happened in the past and will happen again.

 

[frozentundra123456]

Then people would have complained that KBL and CFL aren't big improvements.

The core count increase happens once, then you (Intel) are back to where you were, with entitled enthusiasts complaining that you don't spend any money on R&D and all your engineers are actually just throwing paper airplanes at each other in the rec room.

Actually, they *are* complaining about that anyway, and it is true, especially on the desktop. I also dont buy your argument that one should withhold progress because it might make future increases smaller. Actually, that is exactly the reason why a lot of people, including myself, who actually are supportive of intel are criticizing them. And seriously, do you really think it is "entitled" to expect a mainstream core count increase, since we have been on quads for what, ten years now, and shrunk from 65 nm to 14nm?

 

[DrMrLordX]

Then people would have complained that KBL and CFL aren't big improvements.

And they'd be correct. Though the small clockspeed bump of Kabylake is okay for people who aren't overclockers.

 

[coercitiv]

The core count increase happens once

Actually... the (marketing) benefit of increasing core count may not even happen once for Intel. There is a possible combo of price/perf from the competition that may look into the eyes of the consumer as the equivalent of a core count increase from Intel. If that happens, CFL will be seen as a mere adjustment (reaction), especially if Intel is forced to change their pricing structure until 6c/12t chips hit the market. (I'm really curious of what will happen here, I admit I have no idea how Intel would react)

Putting the current state of the competition aside, CFL doesn't really look like a well orchestrated plan, more like a last minute addition. I think it may have been timed better if it were in the books to begin with.

 

[Dresdenboy]

ComputerBase tested the long awaited Pentium with HT support. Basically the same performance as a Core i3-6100 (97%) for $64 MSRP!

www.computerbase.de/2017-01/intel-pentium-g4560-test-kaby-lake

How might this affect i3 sales?

 

[coercitiv]

How might this affect i3 sales?

What i3?

PS: the i3 will continue to sell well due to the branding "inertia". I see little effect on the OEM market, only DIY buyers will notice the "disturbance in the force" and act accordingly.

 

[AtenRa]

Then people would have complained that KBL and CFL aren't big improvements.

The core count increase happens once, then you (Intel) are back to where you were, with entitled enthusiasts complaining that you don't spend any money on R&D and all your engineers are actually just throwing paper airplanes at each other in the rec room.

No, 14nm era,
you launch 6-core mainstream 3.5GHz Skylake at $350 in 2015, then 1.5 years later in Q1 2017 you launch a 6-Core mainstream 3.8GHz Kabylake at $299 and 1.5 years later in H2 2018 you launch a 6-Core mainstream 4GHz Coffelake at $250 but also an 8 Core at $399. You sell the first year of the new 14nm process launch at higher prices and as the time goes by and process is getting cheaper you offer same IPC at higher Clocks at LOWER Prices to stimulate Volume sales and the last year you also offer a new 8Core at almost the price of the first 6-Core in 2015.
In this scenario people that bought 6Core Skylake in 2015 will have a reason to upgrade to 8 Core Coffelake after 3 years. Today vast majority will NOT upgrade from 4Core to 4Core with same IPC at the same price 3 years later.

Then 7nm era, you go to CanonLake (or what ever) with a 15-20% IPC boost in late 2019 (you have 3-4 years to increase IPC) and you release a 6-Core mainstream 3.5GHz 7nm at $350 and the same again as with 14nm for 3 launches.

 

[IntelUser2000]

Maybe if intel had offered more value instead of stubbornly segmenting the market to the nth degree to try to maximize margins, they could have mitigated the slide a bit at least.

There really is no answer. Market segmentation is what enabled them to have the high margins and profits they have now. But people eventally see through the veil and they get angry. Long term it's probably better not to segment every part of the market. However the focus is always short term. The problem is probably so numerous that they won't be able to stop it.

 

[Arachnotronic]

There really is no answer. Market segmentation is what enabled them to have the high margins and profits they have now. But people eventally see through the veil and they get angry. Long term it's probably better not to segment every part of the market. However the focus is always short term. The problem is probably so numerous that they won't be able to stop it.

You quoted the wrong individual

 

[frozentundra123456]

No, 14nm era,
you launch 6-core mainstream 3.5GHz Skylake at $350 in 2015, then 1.5 years later in Q1 2017 you launch a 6-Core mainstream 3.8GHz Kabylake at $299 and 1.5 years later in H2 2018 you launch a 6-Core mainstream 4GHz Coffelake at $250 but also an 8 Core at $399. You sell the first year of the new 14nm process launch at higher prices and as the time goes by and process is getting cheaper you offer same IPC at higher Clocks at LOWER Prices to stimulate Volume sales and the last year you also offer a new 8Core at almost the price of the first 6-Core in 2015.
In this scenario people that bought 6Core Skylake in 2015 will have a reason to upgrade to 8 Core Coffelake after 3 years. Today vast majority will NOT upgrade from 4Core to 4Core with same IPC at the same price 3 years later.

Then 7nm era, you go to CanonLake (or what ever) with a 15-20% IPC boost in late 2019 (you have 3-4 years to increase IPC) and you release a 6-Core mainstream 3.5GHz 7nm at $350 and the same again as with 14nm for 3 launches.

Exactly.

 

[Arachnotronic]

But people eventally see through the veil and they get angry.

LOL, no. Maybe a few entitled enthusiasts on tech boards who really are never happy with anything that any company gives them will complain (they always complain), but you really have to understand that the average Joe and business consumers (the two volume customers of mainstream-powered desktops) don't think like this.

Consumers and businesses by and large don't buy processors. They buy computers.

Consumers buy computers generally for the following reasons:

1. The one I have doesn't work well or it broke. I need a new one.
2. I, or the person for whom I am buying it, does not have a computer but wants/needs one.

Those people either go online (using old barely working computer or with their smartphone) and pick out a computer, or they go to Best Buy, Wal-Mart, Staples, or someplace like that and pick out a computer.

They don't know a thing about IPC, clock speed, or anything else. There is sometimes (often?) a helpful infographic in the computer section, almost certainly put there through some marketing agreement between the retailer and Intel, that explains in very simple terms what Celeron, Pentium, Core i3/i5/i7 are, and what you should buy based on your "use case" (these use cases are in very broad terms and would look silly to an enthusiast).

The potential buyer has a certain budget in mind, with perhaps a little bit of wiggle room. Then that person looks at the different computers within his or her price range, and either independently or with the help of a sales rep (who has just enough knowledge to explain in broad terms what the different features/options mean) picks a system that she or he thinks will meet their needs for a while.

That person walks out the door with her or his computer, and probably won't be back to buy another computer for a long time.

If you really want to understand how most people buy computers, just go to the computer section in one of the stores I mentioned above, and ask a sales rep to help you pick out a computer. Tell him/her what you want to use it for, how much you're willing to spend, etc. and watch the process unfold.