Obviously the package is different, but there one has IGP and the other doesn't, or has it disabled, idk. That's the real question here, are they the same die in a different package, or two different dies?
The answer to that question goes back to Ozmodan's comment: Desktops/Servers don't typically need IGP, mobile and lower power parts do. Yes, right now they are all coming from the same bin, but KL-X may prove that they don't necessarily have to - Intel is just doing that for "because".
They could easily continue to make the T/U/Y/HQ packages, that's where they are needed. Those are usually installed as BGA directly onto motherboards as well, so those are already in a different package than the 1151 socket.
Then put the 1151 socket CPUs, and use the additional head room in TDP from the missing IGP into Turbo, you can use the die space for additional cores or cache, and all sorts of other fun stuff.
And Intel does do that, it's just not on 1151 - they do it on X99/X299. Gamers just keep buying the 1151 because it's ~much~ less expensive, and the performance doesn't really suffer for it. If folks jumped on X299 in droves, you can bet that Intel would adjust their marketing accordingly, but so long as X299 motherboards are astronomically priced, it's not going to hit that point.
I'd be surprised if Kaby Lake-X is a different die that lacks a GPU. Intel has long sold salvage parts of their chips with the GPU disabled, mostly as Xeon E3, but occasionally in desktops. AMD has long had desktop versions of APUs with the GPU disabled, too.
If Kaby Lake-X is a separate die from normal Kaby Lake quad cores, then Intel has two high bins and then has to throw the rest of the dies away. That would probably be more expensive than just making it a different bin of the same die.
Not so much the part about 40% better ultrabooks, but midway through the article:
These 8th-generation [possibly U-Series only? from context of the previous part of the article] CPUs are not the upcoming Coffee Lake, as previously expected. Intel characterizes its new chips as a "Kaby Lake refresh," referring to the current Kaby Lake chips as "prior generation" parts. Here’s another surprise: both Coffee Lake and the upcoming 10-nm Cannon Lake chips will all be part of Intel’s eighth-generation Core branding.
Confused? Let’s sum up: Intel’s eighth-generation Core chips will include three separate chip architectures and two process technologies, all under a single brand name. Whew!
Interesting that here, they are saying Cannon Lake will be branded 8th Gen, and that the first iteration of 8xxx series ultra-low power parts are still Kaby cores, not Coffee Lake cores. And that eventually they may be, or other SKUs in the generation may be, superceded by Cannon Lake, but retain the 8th gen branding.
So we knew a bit about Cannon Lake - it would be mobile only. This kind of sheds some light on how Intel may be planning to do that. I was expecting a Haswell/Broadwell type of arrangement, but maybe they are thinking of doing it like the 7xxx-X series, where they have multiple architectures (in that case, Skylake and Kaby Lake) in a same branding generation.
For those not familiar, I still have trouble with Intel's SKUs nomenclature. So far I've more or less decoded:
Y - tablets (~7W)
U - ultra thins (~15W)
E - Embedded (some have ECC RAM support) (~35W)
T - mobile (~35W)
HQ/EQ - high performance mobile (~45W)
No letter - standard desktop (~65W)
K - Unlocked desktop (~90W)
X - Unlocked Enthusiast, no IGP (~110W)
H/HQ used to mean Iris graphics, but now they are intermingled in with other SKUs and HQ has more or less become a laptop chip with a higher TDP cap for Turbo (I think it used to stand for Quad Core Laptop chip, which was high performance, but idk now). EQ is an odd one, I can't find much info on it, I would assume it's a higher performance embedded, but can't really tell.
Intel doesn't appear to be terribly consistent, and there is no clear cut guide I've been able to find from Intel (but not shortage of posts like mine trying to divine out the meaning).
Intel calling Sky Lake, Sky Lake Refresh, and Sky Lake Refresh Refresh three separate generations is just trying to compensate for retroactively deciding that Conroe, Penryn, Bloomfield, Lynnfield, Clarkdale, Gulftown, and some others were all first generation core.
What a questionable benchmark. Personally I doubt there will be much difference between the I7 7700 and the new chips as to gaming and when you start adding up the cost of the new chips and new motherboards, it will be a huge difference. Far as I am concerned the I7 7700 is still king by a large margin.
If you need extra cores, a Ryzen solution would be far cheaper and suffer little in processing power.
I do think it's about time that the average core count went up. It's a welcome change to see 2 core CPUs go away, 4 cores become entry level, and 6/12's come to the main stream.
So that part, I'm thankful for.
As far as any actual gains: Intel is claiming 11% IPC upgrade with Coffee Lake. Which is about what they have claimed for each generation subsequent to Sandy Bridge. While those gains are yet to be proven (and the benchmarks linked above neither prove or disprove that), it still remains to be seen if any gains are from actual architectural/process improvements, or mainly from a bump in clock speed. Intel has gone both ways between generations since Sandy.
As a gamer, I'm primarily interested in gains in IPC, although even that has been at a "good enough" place since Sandy Bridge - games are largely GPU-limited at resolutions anyone really cares to play them at (1080p and above). In the few cases they are not, IPC tends to have a bigger impact than core count. My hope is that that situation changes, now that 4 core CPUs will become the standard and 6/12's more mainstream, but consoles haven't moved the needle that much and the PS3 and all current generation Sony/MS consoles are all 8C, and the reality today remains that IPC is still very much king if you just look at games on a PC.
Don't get me wrong, core count excites me, but only in the same way that power in a car excites me. If you wanted a ~lot~ of power, at the expense of everything else, you'd just get a diesel truck. It certainly won't accelerate like a Mustang, but that's the tradeoff you make.
After reading everyone's comments my only contribution is that I used to be in the die-hard Intel camp until after really taking a look at how they have stagnated the PC market in the past decade for no reason, but to be greedy and money hungry.
Hence, for me I decided to go with the R7 1700x and I couldn't be happier. I always recommend it whenever I can because I feel that it is more future-proof, but even more than that, it has the potential to grow with you as a computer power user.
For gaming, I understand the 7700k is the FPS king, but by what...7-10FPS? I play all my games on ultra settings and unless I have an FPS counter on my screen I notice nothing. The other thing is I personally run a TON of background applications while gaming, and would like to get into running VMs and server environments. This is what I mean by the CPU being able to grow with you as you get into more multi-core demanding workloads.
If you are JUST gaming, i5 or r5 all the way. Doesn't make sense to do anything else. However, if you think you want to delve deeper into PCs in the future than AMD is king.
Intel, at least to me, needs to get their shit together and figure out what they are doing. To me it feels like they are just alienating their customer base. Therefore, I will take a 10 FPS hit that I cannot even perceive with my own eyes, and go from 144FPS to 134 FPS, if it means supporting the other side.
I know I'm just a drop of water in a bucket against the ocean that is Intel, but if more and more people convert it will send a message that they need to really start pushing boundaries again.
Intel has been playing the game of squeezing the consumer and stagnating the market since at least the 386/486 with SX and DX. All hardware vendors pull this to some extent. Intel is just the leader of the pack.
Well it's high time, in my opinion, that they get knocked down a few pegs.
FYI I'm not a total Intel hater. I have a i7-3770k as wifeysauce rig and i7-860 in my media center PC. I just want to see Intel squirm a bit lol
I also haven't been on the AMD side since the K6 Spitfire core days, so I wanted to try something new. Been Intel way too long...wanted to strike out into some new territory.
But it's not because of AMD or Ryzen, or anything to do with their CPU division. That's really the coattails they have been riding for the better part of 40 years now. That, and their server/enterprise division, has been where Intel's value has been traditionally. Stock markets want to see growth, that's what drives stock prices, and it's how companies like Amazon and Tesla can have their stock prices continue to skyrocket even though they are losing money.
They are squirming because Desktop/Laptop CPU sales are stagnant at best, or declining depending on how you look at it. They totally missed the boat on Mobile - ARM locked that up and that is a business that is utterly booming. IoT is also in there, and as of right now, that's following the same footsteps as mobile. Their wireless division hasn't really taken off as they would have liked (althoiugh they are starting to make inroads with at least Apple). Their interface division hasn't really taken off as much as they would have liked (Thunderbolt, which recently went royalty free in an effort to keep it relevent). Non-Intel GPUs have taken a large chunk of the HPC market and now their enterprise market is being threatened. Their Tick/Tock approach has stalled out, largely due to the astronomical increase of developing new and smaller process nodes (the corollary to Moore's First Law is Rock's Law: as transistor count goes up, the production price also goes up exponentially), and Intel foundries have historically been leader of the pack in that regard.
It isn't that Intel isn't trying. They have tried to diversify in the past 10-15 years. They bought McAfee, that didn't pan out so they sold most of it it off again. They bought Havok - same story. They invested a lot into graphics, and it got them still 10 steps behind everyone else with IGP, and Knights Landing, which they just discontinued a significant portion of. They are betting a lot on Optane, but by all early accounts, it looks like a stinker outside of specific server environments. Intel even has a significant VR push, but I won't turn this into a discussion about VR with my opinions about that.
All in all, it makes me wonder if Intel had spent the same amount of resources they had in attempting to diversify-- all that money spent on Havok and McAfee and such, and doubled down on what they do great: their world-leading foundry and x86 CPU technology - where they would be today. If Intel really tried to continue to push the envelope on CPUs even in the absence of competition (not just efficiency or speed or core count, but overall in general), would the landscape look any different than it does today?
It wasn't until Apple slashed their entire product line down to about 4 products that it turned around in the late 90's and focused on what they do great: not hardware or software, but the overall user experience -- and drilled that user experience down into essence of every product they made. Now it flirts with being one of the most valued companies in the world.
Not that I think Intel should parallel Apple, I'm just a general opponent of a conglomerate that tries to diversify so much that it loses focus of it's core, and Apple's story shows at least one company that realized that (if they still do today, remains to be seen). Intel seems like a company running around like a chicken with it's head cut off: they don't know where they are going, so they are just trying a little bit of everything and hoping something takes off, rather than focusing on what they do well and finding ways to apply that.
Intel's annual revenue is nearly 50 times ARM's. I don't think they want to trade places with ARM. ARM may sell more CPU cores than Intel, but Intel makes a ton of money on every CPU core they sell and ARM doesn't.
Intel still has most of the desktop, laptop, and server markets. What more could they have if they had focused solely on that? Fend off AMD Epyc a little better? The primary argument for Raven Ridge is going to be the integrated GPU, and Intel focusing on x86 won't beat that back.
Atom and Xeon Phi are two Intel efforts at focusing on x86 and pushing it into more markets. Unless you deal with it, you probably don't realize what a ridiculous product Xeon Phi is. Now without the PCI Express cards, all it lets you have is an HPC with one device per node, not very much performance on that one device, and an enormous nightmare to code for if you were hoping for it to sometimes be significantly superior to an ordinary Xeon or Epyc CPU server. You get more development pain than from a GPU-based system, and without the performance upside outside of some really weird outlier cases.
Intel's annual revenue is nearly 50 times ARM's. I don't think they want to trade places with ARM. ARM may sell more CPU cores than Intel, but Intel makes a ton of money on every CPU core they sell and ARM doesn't.
No, but many people on Intel are likely wishing they'd done things differently and managed to compete with ARM. Now they're in a stagnant market that's basically limited to situations where ARM doesn't provide enough computing power.
Intel's annual revenue is nearly 50 times ARM's. I don't think they want to trade places with ARM. ARM may sell more CPU cores than Intel, but Intel makes a ton of money on every CPU core they sell and ARM doesn't.
No, but many people on Intel are likely wishing they'd done things differently and managed to compete with ARM. Now they're in a stagnant market that's basically limited to situations where ARM doesn't provide enough computing power.
While ARM is stuck in a market where they can barely make any money per CPU core. IoT especially is going to be so price sensitive that for a CPU architecture to "win" the market won't make you a ton of money.
This, as with mobile devices, is really a case of the only real winner in the competition being consumers. Intel is doing well to still be able to get a ton of money per desktop, laptop, and server chip they sell. Their data center and HPC position is in serious peril, though.
Intel's annual revenue is nearly 50 times ARM's. I don't think they want to trade places with ARM. ARM may sell more CPU cores than Intel, but Intel makes a ton of money on every CPU core they sell and ARM doesn't.
ARM just holds the license. Don't forget about Samsung, Qualcomm, Apple, Toshiba, and whomever else I am forgetting about that actually make and sell mobile CPUs based on ARM. Add all that up, and ~that~ is the boat that Intel missed when they missed mobile.
Intel's annual revenue is nearly 50 times ARM's. I don't think they want to trade places with ARM. ARM may sell more CPU cores than Intel, but Intel makes a ton of money on every CPU core they sell and ARM doesn't.
No, but many people on Intel are likely wishing they'd done things differently and managed to compete with ARM. Now they're in a stagnant market that's basically limited to situations where ARM doesn't provide enough computing power.
While ARM is stuck in a market where they can barely make any money per CPU core. IoT especially is going to be so price sensitive that for a CPU architecture to "win" the market won't make you a ton of money.
Not a lot of money, over an estimated 1.5B units per year (and that's just phones, not including tablets, TV's, refrigerators, cars, coffee makers, etc.), and it really starts to add up quickly.
Compares to PC sales, which have been declining over the past several years, and are only about 20% of what the smart phone market volume is expected to be. Sure, margins may be better, but that's a lot of volume, and today it's not uncommon for a smart device to cost more than a basic PC.
And servers, let's not forget those. Also starting to decline, and only about 5% of the volume of the PC market, let alone trying to compare it to the smart device market. They have nice margins, but the difference in volume here is orders of magnitude.
So yeah... ARM may not be making a ton of cash - they are just an IP holder, they don't really do much other than create the standard. They don't have to build or maintain a fab, they don't have to deal with production and inventory. They don't have to maintain a large production floor staff in a clean room environment. They are just a group of engineers that need to ratify standards, and a few lawyers to parse all the contracts for everyone lining up to pay them to use their technology.
Yes - chips are more expensive than they were before. A Top Tier i7-k runs $360 now, compared to the $310-320 it used to cost. But it also packs 50% more cores, and you can find price competitive options in the SKU lineup that present value options compared to previous price points.
So I have to admit - with respect to pricing, you get more processing power for less money in the Coffee Lake series than you could have with any other series from Intel in recent history.
Intel is claiming "Best Gaming Processor Ever". I don't know about that, waiting for benchmarks. They have tweaked the boost a bit - a lower floor (<3.8 across the i7/i5 lineup), with a higher ceiling; going up to 4.7GHz in single core count boost modes.
The i3 looks nearly identical to previous generation i5. They cost about $60 less, which is fantastic, but they also removed boost, so you get a static, and lower, stock frequency. Given there is a i3-K edition, that probably works out in consumer's favor, but again, benchmarks will tell.
Granted, part of this good news comes with the news that Intel is also requiring the new 370 chipset to use with Coffee Lake. All 8000 series processors will require a 300 series motherboard, just because. Even though socket and functionality-wise, you should be able to use 100-300 series motherboards with 6000 through 8000 series chips.
Z370 supports.... umm, nothing really new from the Z270 that I can find. A faster "supported" memory speed (although I seem to recall the memory controller is on the CPU die), and better power delivery specification (although any "good" motherboard is going to already have power delivery beyond specs for OCing). That's all I could find.
So a good bit of that "CPU doesn't really cost more than I expected it to" is taken away by "But you need to buy into a new chipset just because...."
Granted, part of this good news comes with the news that Intel is also requiring the new 370 chipset to use with Coffee Lake. All 8000 series processors will require a 300 series motherboard, just because. Even though socket and functionality-wise, you should be able to use 100-300 series motherboards with 6000 through 8000 series chips.
Gigabyte has now apparently said that Intel also is preventing Z370 from working with Kaby Lake. Apparently Gigabyte says it would work fine, but Intel is demanding that they artificially disable it in firmware.
No idea what Intel hopes to gain by that, as it seems to be looking for creative reasons to be petty. It's one thing to say, we won't officially support it but motherboard vendors can do so on their own. But to artificially disable it?
AmazingAveryAge of Conan AdvocateMemberUncommonPosts: 7,188
The way I see it is that it sucks that there is not much difference between the z270 and z370 chipsets. The z370 does have better power delivery in design.
However, it looks like that they are drawing a clear line here.
Coffee lake on 300 series boards and it also looks like Cannon Lake too. October 5th launch for Coffee Lake (14nm) and then Cannon Lake (10nm) comes near year end and ramp up in early 2018. Both of these will be compatible on z370 and z390 boards. So that is why the line had to be drawn.
I also noticed that the z370 boards are going to be more expensive. I still think a Ryzen setup is your best option these days. I just feel like I am constantly getting screwed by sticking with Intel. Most of my builds these days are Ryzen 1700x builds, but I will admit a lot of these are not all gaming builds.
Another interesting thing, I was in a IT shop the other day and they had 40 inch UHD TV's on many desks. The manager told me they are much cheaper than a monitor and provide far more work space for their developers. He told me he spends less than $300 for each.
Get an Intel and Nvidia and you'll probably never have to search or post on forums about your poor performance.
You buy particular parts, not brand names. I assure you that there have been many people who bought an Intel Atom based part and then wondered why it was so slow, especially in the pre-Silvermont days when they didn't even have out of order execution.
What's funny is that the number of times Nvidia drivers caused serious problems in the past 2 years is more than the number of all the times amd had problems decades ago for which they are still getting called out for, ...
And everybody forgets about it in a weeks time, ...
Neither company is perfect, its just one of them gets a free pass over and over again.
I have been using AMD exclusively since 2006. I only had 1 driver problem since then and it was caused by an Upgrade to Windows on a custom MSI fan profile. I can't really speak for the problems with nVidia recently since I have not been using them recently. I may switch in January, but it really depends on what nVidia comes out with since I need something like the Titan XP or Vega Frontier and nVidia has been rather poor at maintaining the Titan XP.
Comments
If Kaby Lake-X is a separate die from normal Kaby Lake quad cores, then Intel has two high bins and then has to throw the rest of the dies away. That would probably be more expensive than just making it a different bin of the same die.
One thing to note - the pricing in the article is in CAD, not US Dollars.
http://www.pcworld.com/article/3217330/components-processors/intel-8th-gen-core-cpu-price-release-date-specs-features-faq.html
Not so much the part about 40% better ultrabooks, but midway through the article:
Interesting that here, they are saying Cannon Lake will be branded 8th Gen, and that the first iteration of 8xxx series ultra-low power parts are still Kaby cores, not Coffee Lake cores. And that eventually they may be, or other SKUs in the generation may be, superceded by Cannon Lake, but retain the 8th gen branding.
So we knew a bit about Cannon Lake - it would be mobile only. This kind of sheds some light on how Intel may be planning to do that. I was expecting a Haswell/Broadwell type of arrangement, but maybe they are thinking of doing it like the 7xxx-X series, where they have multiple architectures (in that case, Skylake and Kaby Lake) in a same branding generation.
For those not familiar, I still have trouble with Intel's SKUs nomenclature. So far I've more or less decoded:
Y - tablets (~7W)
U - ultra thins (~15W)
E - Embedded (some have ECC RAM support) (~35W)
T - mobile (~35W)
HQ/EQ - high performance mobile (~45W)
No letter - standard desktop (~65W)
K - Unlocked desktop (~90W)
X - Unlocked Enthusiast, no IGP (~110W)
H/HQ used to mean Iris graphics, but now they are intermingled in with other SKUs and HQ has more or less become a laptop chip with a higher TDP cap for Turbo (I think it used to stand for Quad Core Laptop chip, which was high performance, but idk now). EQ is an odd one, I can't find much info on it, I would assume it's a higher performance embedded, but can't really tell.
Intel doesn't appear to be terribly consistent, and there is no clear cut guide I've been able to find from Intel (but not shortage of posts like mine trying to divine out the meaning).
4C/8T Kaby Lake refresh for ultra-low power (U-series)
No Coffee Lake, yet.
https://hothardware.com/news/intel-core-i7-8700k-coffee-lake-i5-8400-cpus-sisoft-sandra-benchmark
Up the core count by 50%, actual performance goes up by 32%.
Sounds about right.
If you need extra cores, a Ryzen solution would be far cheaper and suffer little in processing power.
Intel is tilting at windmills again.
So that part, I'm thankful for.
As far as any actual gains: Intel is claiming 11% IPC upgrade with Coffee Lake. Which is about what they have claimed for each generation subsequent to Sandy Bridge. While those gains are yet to be proven (and the benchmarks linked above neither prove or disprove that), it still remains to be seen if any gains are from actual architectural/process improvements, or mainly from a bump in clock speed. Intel has gone both ways between generations since Sandy.
As a gamer, I'm primarily interested in gains in IPC, although even that has been at a "good enough" place since Sandy Bridge - games are largely GPU-limited at resolutions anyone really cares to play them at (1080p and above). In the few cases they are not, IPC tends to have a bigger impact than core count. My hope is that that situation changes, now that 4 core CPUs will become the standard and 6/12's more mainstream, but consoles haven't moved the needle that much and the PS3 and all current generation Sony/MS consoles are all 8C, and the reality today remains that IPC is still very much king if you just look at games on a PC.
Don't get me wrong, core count excites me, but only in the same way that power in a car excites me. If you wanted a ~lot~ of power, at the expense of everything else, you'd just get a diesel truck. It certainly won't accelerate like a Mustang, but that's the tradeoff you make.
Hence, for me I decided to go with the R7 1700x and I couldn't be happier. I always recommend it whenever I can because I feel that it is more future-proof, but even more than that, it has the potential to grow with you as a computer power user.
For gaming, I understand the 7700k is the FPS king, but by what...7-10FPS? I play all my games on ultra settings and unless I have an FPS counter on my screen I notice nothing. The other thing is I personally run a TON of background applications while gaming, and would like to get into running VMs and server environments. This is what I mean by the CPU being able to grow with you as you get into more multi-core demanding workloads.
If you are JUST gaming, i5 or r5 all the way. Doesn't make sense to do anything else. However, if you think you want to delve deeper into PCs in the future than AMD is king.
Intel, at least to me, needs to get their shit together and figure out what they are doing. To me it feels like they are just alienating their customer base. Therefore, I will take a 10 FPS hit that I cannot even perceive with my own eyes, and go from 144FPS to 134 FPS, if it means supporting the other side.
I know I'm just a drop of water in a bucket against the ocean that is Intel, but if more and more people convert it will send a message that they need to really start pushing boundaries again.
FYI I'm not a total Intel hater. I have a i7-3770k as wifeysauce rig and i7-860 in my media center PC. I just want to see Intel squirm a bit lol
I also haven't been on the AMD side since the K6 Spitfire core days, so I wanted to try something new. Been Intel way too long...wanted to strike out into some new territory.
But it's not because of AMD or Ryzen, or anything to do with their CPU division. That's really the coattails they have been riding for the better part of 40 years now. That, and their server/enterprise division, has been where Intel's value has been traditionally. Stock markets want to see growth, that's what drives stock prices, and it's how companies like Amazon and Tesla can have their stock prices continue to skyrocket even though they are losing money.
They are squirming because Desktop/Laptop CPU sales are stagnant at best, or declining depending on how you look at it. They totally missed the boat on Mobile - ARM locked that up and that is a business that is utterly booming. IoT is also in there, and as of right now, that's following the same footsteps as mobile. Their wireless division hasn't really taken off as they would have liked (althoiugh they are starting to make inroads with at least Apple). Their interface division hasn't really taken off as much as they would have liked (Thunderbolt, which recently went royalty free in an effort to keep it relevent). Non-Intel GPUs have taken a large chunk of the HPC market and now their enterprise market is being threatened. Their Tick/Tock approach has stalled out, largely due to the astronomical increase of developing new and smaller process nodes (the corollary to Moore's First Law is Rock's Law: as transistor count goes up, the production price also goes up exponentially), and Intel foundries have historically been leader of the pack in that regard.
It isn't that Intel isn't trying. They have tried to diversify in the past 10-15 years. They bought McAfee, that didn't pan out so they sold most of it it off again. They bought Havok - same story. They invested a lot into graphics, and it got them still 10 steps behind everyone else with IGP, and Knights Landing, which they just discontinued a significant portion of. They are betting a lot on Optane, but by all early accounts, it looks like a stinker outside of specific server environments. Intel even has a significant VR push, but I won't turn this into a discussion about VR with my opinions about that.
All in all, it makes me wonder if Intel had spent the same amount of resources they had in attempting to diversify-- all that money spent on Havok and McAfee and such, and doubled down on what they do great: their world-leading foundry and x86 CPU technology - where they would be today. If Intel really tried to continue to push the envelope on CPUs even in the absence of competition (not just efficiency or speed or core count, but overall in general), would the landscape look any different than it does today?
It wasn't until Apple slashed their entire product line down to about 4 products that it turned around in the late 90's and focused on what they do great: not hardware or software, but the overall user experience -- and drilled that user experience down into essence of every product they made. Now it flirts with being one of the most valued companies in the world.
Not that I think Intel should parallel Apple, I'm just a general opponent of a conglomerate that tries to diversify so much that it loses focus of it's core, and Apple's story shows at least one company that realized that (if they still do today, remains to be seen). Intel seems like a company running around like a chicken with it's head cut off: they don't know where they are going, so they are just trying a little bit of everything and hoping something takes off, rather than focusing on what they do well and finding ways to apply that.
Intel still has most of the desktop, laptop, and server markets. What more could they have if they had focused solely on that? Fend off AMD Epyc a little better? The primary argument for Raven Ridge is going to be the integrated GPU, and Intel focusing on x86 won't beat that back.
Atom and Xeon Phi are two Intel efforts at focusing on x86 and pushing it into more markets. Unless you deal with it, you probably don't realize what a ridiculous product Xeon Phi is. Now without the PCI Express cards, all it lets you have is an HPC with one device per node, not very much performance on that one device, and an enormous nightmare to code for if you were hoping for it to sometimes be significantly superior to an ordinary Xeon or Epyc CPU server. You get more development pain than from a GPU-based system, and without the performance upside outside of some really weird outlier cases.
This, as with mobile devices, is really a case of the only real winner in the competition being consumers. Intel is doing well to still be able to get a ton of money per desktop, laptop, and server chip they sell. Their data center and HPC position is in serious peril, though.
Compares to PC sales, which have been declining over the past several years, and are only about 20% of what the smart phone market volume is expected to be. Sure, margins may be better, but that's a lot of volume, and today it's not uncommon for a smart device to cost more than a basic PC.
And servers, let's not forget those. Also starting to decline, and only about 5% of the volume of the PC market, let alone trying to compare it to the smart device market. They have nice margins, but the difference in volume here is orders of magnitude.
So yeah... ARM may not be making a ton of cash - they are just an IP holder, they don't really do much other than create the standard. They don't have to build or maintain a fab, they don't have to deal with production and inventory. They don't have to maintain a large production floor staff in a clean room environment. They are just a group of engineers that need to ratify standards, and a few lawyers to parse all the contracts for everyone lining up to pay them to use their technology.
I admit - Intel surprised me with pricing.
Yes - chips are more expensive than they were before. A Top Tier i7-k runs $360 now, compared to the $310-320 it used to cost. But it also packs 50% more cores, and you can find price competitive options in the SKU lineup that present value options compared to previous price points.
So I have to admit - with respect to pricing, you get more processing power for less money in the Coffee Lake series than you could have with any other series from Intel in recent history.
Intel is claiming "Best Gaming Processor Ever". I don't know about that, waiting for benchmarks. They have tweaked the boost a bit - a lower floor (<3.8 across the i7/i5 lineup), with a higher ceiling; going up to 4.7GHz in single core count boost modes.
The i3 looks nearly identical to previous generation i5. They cost about $60 less, which is fantastic, but they also removed boost, so you get a static, and lower, stock frequency. Given there is a i3-K edition, that probably works out in consumer's favor, but again, benchmarks will tell.
Granted, part of this good news comes with the news that Intel is also requiring the new 370 chipset to use with Coffee Lake. All 8000 series processors will require a 300 series motherboard, just because. Even though socket and functionality-wise, you should be able to use 100-300 series motherboards with 6000 through 8000 series chips.
Z370 supports.... umm, nothing really new from the Z270 that I can find. A faster "supported" memory speed (although I seem to recall the memory controller is on the CPU die), and better power delivery specification (although any "good" motherboard is going to already have power delivery beyond specs for OCing). That's all I could find.
So a good bit of that "CPU doesn't really cost more than I expected it to" is taken away by "But you need to buy into a new chipset just because...."
https://www.hardocp.com/news/2017/09/24/intel_z370_chipset_could_support_kaby_lake_but_will_allow
No idea what Intel hopes to gain by that, as it seems to be looking for creative reasons to be petty. It's one thing to say, we won't officially support it but motherboard vendors can do so on their own. But to artificially disable it?
However, it looks like that they are drawing a clear line here.
Coffee lake on 300 series boards and it also looks like Cannon Lake too.
October 5th launch for Coffee Lake (14nm) and then Cannon Lake (10nm) comes near year end and ramp up in early 2018. Both of these will be compatible on z370 and z390 boards. So that is why the line had to be drawn.
Another interesting thing, I was in a IT shop the other day and they had 40 inch UHD TV's on many desks. The manager told me they are much cheaper than a monitor and provide far more work space for their developers. He told me he spends less than $300 for each.
Get an Intel and Nvidia and you'll probably never have to search or post on forums about your poor performance.
You were saying ?
What's funny is that the number of times Nvidia drivers caused serious problems in the past 2 years is more than the number of all the times amd had problems decades ago for which they are still getting called out for, ...
And everybody forgets about it in a weeks time, ...
Neither company is perfect, its just one of them gets a free pass over and over again.