Why does intel gimp some of their CPU lines?

[OCGuy]

You should be able to move down to the front row at the stadium if there are unfilled seats, right? I mean, yea you paid for a nose bleed, but the seat is already there. It doesn't cost them any more for you to sit in it, right?

 

[bononos]

Has anyone been able to demonstrate how or why overclocked k chips would affect sales of Xeon processors? That just doesn't make sense to me. After all this time, are people really going to be duped into thinking an overclocked k chip is actually a Xeon? Given the price on the 4770k, would any shady OEM/builder actually make a profit by trying to remark the 4770k as a Xeon? Connect those dots and maybe Intel has an incentive to kill VT-d on k chips; however, I just don't see that being the case.
.........

I wasn't referring to shady remarking schemes. The others have explained market segmentation quite well. But I would also think that Intel might want to protect their good reputation in ensuring that non-overclocked chips be used in non-trivial computing applications. So you can't just consider the impact of the typical gamer/homeuser on Intel's bottom line.

 

[DominionSeraph]

So, I have to pay an extra, what, $700 to gain VT-d and lose AVX2/TSX/etc? What if I want an unlocked multiplier, VT-d, AND AVX2? I guess Haswell-e is going to take care of that at a very high price point, eventually. That doesn't help anyone right now.

Oh noes, someone has to wait for technology?

Where's my 1TB SSD for $1? I want it now!

 

[DrMrLordX]

I wasn't referring to shady remarking schemes. The others have explained market segmentation quite well. But I would also think that Intel might want to protect their good reputation in ensuring that non-overclocked chips be used in non-trivial computing applications. So you can't just consider the impact of the typical gamer/homeuser on Intel's bottom line.

Okay, that sort of clarifies the point a bit, though they should take care to protect their reputation in other areas as well . . .

Apparently, it does. People with actual data and inside knowledge get paid to make these decisions, and obviously their segmentation system isn't an issue at this time.

You'd think so. Sometimes it's just hard to see exactly what is their angle.

It's just market segmentation. If you're running important applications and want stability you really shouldn't be overclocking, but you may be using ecc memory and should be running at validated stock frequencies, so here's a chip that has been validated for x features and purposes. Now intel can target that specific market and optimize the price point for it.

Where consumer chips can affect sales of xeon processors is when shared features exist but the optimal price points in those markets differ...also it can muddy your data on what prices are working in each market. Even if OEM's may not actually use an unlocked part, they can use it's pricing as leverage in price negotiations.

See, that second paragraph? That makes some sense, though it is somewhat lamentable that OEMs would haggle over Xeon pricing based on non-Xeon CPUs that, you know, don't support ECC RAM. But that isn't necessarily Intel's fault, and it certainly isn't at their behest.

I think some of the griping over missing instructions on k chips could have been prevented if they had bothered to explain it to the enthusiast community somehow. Still won't make that small minority of overclocking VM users happy, but whatever.

Oh noes, someone has to wait for technology?

Where's my 1TB SSD for $1? I want it now!

Bad analogy. Intel has been quite able to sell unlocked Haswell chips with VT-d/TSX/etc., they have simply chosen not to do so. Apparently the OP doesn't want to wait for Haswell-e.

edit: added quote/reply

 

[F1shF4t]

Here's one example from a few years ago which fits in the whole "does not make sense" market segmentation. (And something that annoyed me before)
Celeron SU2300 - Has VT-x
Pentium SU4100 - No VT-x
Core 2 SU7300 - Has VT-x

Why is that the more expensive and higher performing model is lacking instruction of the lower model. Same point about the K chips.

What I don't get is why Intel would create new instruction sets, that they presumably want to be widely adopted, and then remove those instruction sets from some, or even most of their chips. What benefit do they get if nobody adopts the instructions because very few people have chips that can use them?

This. No AVX on celeron and pentium haswells is just stupid.

 

[ShintaiDK]

The AVX/AVX2 part I agree on. That one is just plain silly.

And they could have limited AVX2 on Celeron/Pentium to 128bit caches if they wanted some sort of segmentation there.

 

[Nothingness]

The AVX/AVX2 part I agree on. That one is just plain silly.

Yes, and TSX not being on 4770K also is completely stupid.

And they could have limited AVX2 on Celeron/Pentium to 128bit caches if they wanted some sort of segmentation there.

That'd be much harder to implement than just fusing off AVX.

 

[ShintaiDK]

Yes, and TSX not being on 4770K also is completely stupid.


That'd be much harder to implement than just fusing off AVX.

TSX I can understand. Its essentially a serverside feature for now.

To limit AVX/AVX2 to 128bit cache is quite easy and can be done in microcode if not fused. Haswell already switches between 128 and 256bit depending on load. Including +100mV.

 

[seitur]

Yeah no AVX/AVX2 on Celerons&Pentiums is just very silly.

 

[Nothingness]

TSX I can understand. Its essentially a serverside feature for now.

Do you mean people use non-K 4770 in servers?

Sorry but the TSX removal from K models makes no more sense than the lack of AVX on most Intel CPU.

To limit AVX/AVX2 to 128bit cache is quite easy and can be done in microcode if not fused. Haswell already switches between 128 and 256bit depending on load.

Do you have a link that talks about that 128/256-bit cache switch access (which is what you first claimed)?

Including +100mV.

That increase is due to turning on AVX/AVX2 which is using 256-bit wide registers. This does mean at all that AVX/AVX2 can switch between 128- or 256-bit cache accesses.

 

[schmuckley]

LOL @ you guys defending chip "segmentation"
It's total crap.
Once upon a time,all chips were "unlocked" and you could do whatever you wanted with them if you knew how.Also:AMD and Intel chips ran in the same socket.
PS:AVX on a Celeron would be good (for the consumer)

 

[Nothingness]

LOL @ you guys defending chip "segmentation"
It's total crap.

No, it isn't, it makes economical sense. But Intel went way too far in their segmentation.

 

[Homeles]

LOL @ you guys defending chip "segmentation"
It's total crap.
Once upon a time,all chips were "unlocked" and you could do whatever you wanted with them if you knew how.Also:AMD and Intel chips ran in the same socket.
PS:AVX on a Celeron would be good (for the consumer)

Cowardly hit-and-run post, if I ever saw one.

 

[Elixer]

Apparently the OP doesn't want to wait for Haswell-e.

It all boils down to the waiting game.
There have been tons of posts about 'market segmentation', and a few people trying to defend the segmentation practice.

You can pay more, and get less, and that seems fine for the fan boys.
Win 8 (and on) has Hyper-V 'built in' now, and VMs will become more and more popular, and the missing VT-d on the more expensive K series of chips is disheartening.

This is also why we need more serious competition in the CPU field, AMD seems to be content with their concentration on mobile efforts, so basically, intel can do anything it wants with no real alternative solution.
For now, we just need to wait for haswell-E.

 

[DrMrLordX]

LOL @ you guys defending chip "segmentation"
It's total crap.
Once upon a time,all chips were "unlocked" and you could do whatever you wanted with them if you knew how.Also:AMD and Intel chips ran in the same socket.
PS:AVX on a Celeron would be good (for the consumer)

AMD (and Cyrix, and IDT, etc.) chips running in socket 7 is ancient history. Fully unlocked CPUs up and down the product lineup is also ancient history. You will probably never see anything like the Celeron 300a ever again. No matter how much some of us may personally dislike how Intel (or even AMD) has used market segmentation to improve profits, I think we all have to at least acknowledge that most of those decisions have improved their bottom line (though sometimes it's hard to figure out exactly how). Their decisions to remove certain workstation/server instruction sets from k chips and to remove AVX from the Pentiums/Celerons may make business sense to somebody, somewhere, somehow, but that doesn't mean we have to like those decisions.

It all boils down to the waiting game.
There have been tons of posts about 'market segmentation', and a few people trying to defend the segmentation practice.

You can pay more, and get less, and that seems fine for the fan boys.

There's been worse fanboyism before, and I'm sure there will be worse in the future as well.

Win 8 (and on) has Hyper-V 'built in' now, and VMs will become more and more popular, and the missing VT-d on the more expensive K series of chips is disheartening.

See, that's an interesting angle that hasn't really been discussed here. Many "casual" users are kicking desktops to the curb, leaving "enthusiasts" as a proportionately larger minority among desktop users. If VMs really do catch on fire with your average overclocker, then the omission of VT-d from k chips would be significant gaffe on Intel's part.

This is also why we need more serious competition in the CPU field, AMD seems to be content with their concentration on mobile efforts, so basically, intel can do anything it wants with no real alternative solution.
For now, we just need to wait for haswell-E.

We'll see what AMD can pull off, but it seems pretty obvious that excavator alone won't save them in the desktop/workstation space. HSA to the rescue? That's for another thread.

edit: added quote/response

 

[KingFatty]

If VMs really do catch on fire with your average overclocker, then the omission of VT-d from k chips would be significant gaffe on Intel's part.

Chicken or egg? The presumption seems to be that no, there is not a pressing desire to have both features. In other words, there was no fire in the first place - we are assuming the reason is because there is no desire. But could it be that the reason things haven't caught on fire is precisely because Intel has not offered it, ensuring it doesn't catch on fire?

So is Intel just giving the market what it wants, where segmentation is a natural reflection of what buyers are purchasing?

Or is Intel denying the market and creating pent-up demand/unmet needs for something that Intel is not providing?

I just get the impression that the need for both features on the same chip is initially mentally distressing, but when it comes down to it, it's not so bad to create a problem, and so Intel is fiducially obligated to extract the most profit by segmenting the product and being good capitalists and not giving away all the features in all the chips and being forced to charge too-low rates for chips that businesses are happy to spend more on.

 

[Atreidin]

So I guess Toyota crapped on me, because there's a place on my car where fog lights go, but I didn't get them.

Or, it could be that I didn't pay for them.

Wow that's a terrible analogy. I mean... wow.

It would be more accurate if Toyota installed the fog lights in your car but then cut the connection to your electronics system so they won't turn on, or maybe covered them up so the light doesn't shine on anything.

Comparing the disabling of fully-constructed and functioning features to missing car parts that never existed in the first place just doesn't work. Lots of people find it weird to buy something that the manufacturer produced at a certain level of quality, but before selling it to them, purposely made worse in order to sell it for a lower price. The end result for the consumer isn't much different from just making two products, but that's not the issue. It feels weird. It feels like you are buying something that was intentionally broken. Where else does this happen besides certain electronics? Bakers don't make a lot of cakes, then intentionally burn some of them so they can put them on the discount rack. Hanes doesn't intentionally rip the elastic in a few pairs of underwear so they can have a super-cheap value line. A carpenter doesn't make a bunch of nice tables then cut an inch off of a few legs so they have some wobbly ones to sell to poor people. Toyota doesn't install fog headlights in a car then break them if you don't ask for them.

It's just a weird concept. If you think it is ultimately a good thing, by all means, everyone is all ears for the reasons why. However I don't think anybody here is so monumentally stupid as to act like the fictitious consumer in your horrible car analogy and it's rather insulting to everyone's intelligence to even suggest it, unless you're just trolling.

 

[Atreidin]

Here's one example from a few years ago which fits in the whole "does not make sense" market segmentation. (And something that annoyed me before)
Celeron SU2300 - Has VT-x
Pentium SU4100 - No VT-x
Core 2 SU7300 - Has VT-x

Why is that the more expensive and higher performing model is lacking instruction of the lower model. Same point about the K chips.



This. No AVX on celeron and pentium haswells is just stupid.

I also find it annoying that some i3s support ECC, but not the higher-end stuff. I want ECC dang it! (Yes I know there's server/workstation processors but I don't get those offered to me for free very often.)

 

[MongGrel]

Why would you really want ECC to be honest the majority of the time, have a X5650 in the main rig and it can run it and do not.

 

[Elixer]

See, that's an interesting angle that hasn't really been discussed here. Many "casual" users are kicking desktops to the curb, leaving "enthusiasts" as a proportionately larger minority among desktop users. If VMs really do catch on fire with your average overclocker, then the omission of VT-d from k chips would be significant gaffe on Intel's part.

Perhaps people don't know that they can actually take a image of an OS, and dump it to a VM, using P2V or disk2VHD.

For those of you that don't know, this means you can basically clone your XP (or whatever) machine to a VM, that you can run within your primary OS.
Win 8+ includes Hyper-V, or you can use vmware or virtualbox and having VT-d enabled on all CPUs would translate to much faster performance, and more reliability, and higher security.

When browsers (or...) really take security seriously, I can see them have a browser OS that is run in a VM. This is just the tip of the iceberg.

 

[zir_blazer]

The problem there is getting native-like functionality and compatibility in a virtualized enviroment. If you want to make a VM out of your old main OS installation, chances are you still want to play games, and that requires a working GPU inside the VM. I don't know if in Hyper-V you can do VGA Passthrough like you can on some Linux solutions (Mainly Xen). It was actually a feature that was hard to get working and required VT-d (And proper BIOS support, and back when I did this Thread it was hard to get a Motherboard with proper support).

 

[Elixer]

For games, the newer VMs do GPU passthrough, and while still not running at native speed, it is becoming closer.
Yeah, that is another aspect of this--though, I bet if VT-d came enabled on all of their CPUs, then mobo makers would have BIOS support for that as well.

 

[Matt1970]

Microsoft does the same exact thing. If you download a copy of 7 from Digital river or buy one that cop has every version on it from Starter to Ultimate with just a simple file in there to tell it which one to install.

 

[mrmt]

Just for the sake of curiosity: I've been people whining about not getting this or that feature, of how Intel is evil about cutting down features, bla bla bla. But everything is about business, so let's analyze the impact of Intel business decision on their bottom line:

- How many of you complaining did not buy a given Intel processor because of lack of these features?

- How many people do you know that didn't buy an Intel processor because lack of features?

- Of those people, what's the next step, buy another but more expensive Intel processor, or buy an AMD processor?

- If you chose an AMD processor, how much did you spend in the AMD processor?

My feeling is that not many people would go AMD because of lack of VT-d, because that choice brings another set of problems (higher power consumption, lack of ST performance, etc etc etc), that choice would bring VT-d but not AVX2, let alone TSX. In the end, that is a business decision that makes sense for Intel.

 

[Nothingness]

I certainly regret buying a 4770K. In hindsight I would have picked a 4771, which has TSX with which I would like to play.

That being said, I certainly like my 4770K