Technology / CPU pricing discussion

[VirtualLarry]

I was just thinking, how the reason that most new CPUs and whatnot, are released at around the same price-points as their predecessors, is because there is still old stock on the market that they have to compete with.

Therefore, there are limits on how much mfgs can charge, even if they want to raise the price-points of some of their goods.

Though, we did see retailer prices on things like the i7-6700K rise to nearly $100 over list price, for probably a majority of their active lifespan on the free market.

And things like spotty (more like rare) availability of the lowest-end Skylake SKUs like the G3900.

But it seems like those price-points that Intel chose back in the Core2 days, have stuck with us all this time.

This is of course changing with Broadwell-E HEDT, with an additional SKU causing a bump in the highest price-point for HEDT (and formerly Extreme Edition) CPUs.

Thoughts on pricing?

Edit: It should probably be noted that volume should probably factor into this discussion as well, and that those price-points determined in the Core2 era were probably designed to maximize ASPs, while maintaining the volume necessary to fill Intel's FABs, and drive Moore's Law forward.

Given the decline of the desktop PC industry (and even laptops), should these price-points change? Because market conditions suggest that they should change, in order to still drive volume enough to keep the FABs running?

Just a hypothetical - if the i5-6600K (and successors) was $150, and the i7-6700K (and successors) was $200-230, would you as a consumer and an enthusiast, upgrade every generation, rather than hold on to the same CPU through 4-5 generations?

IOW, given Intel's slow overall progress in absolute performance, at what price point would you be willing to "jump ship" to the next generation, every generation, or every tick, or every tock? That would certainly drive a lot more volume for Intel's FABs, I would think, if they could count on a much higher percentage of their installed base upgrading sooner.

I mean, if they can sell a low-end quad-core Atom Z3735F CPU, for a tablet or netbook, for $10-15 and make money, then why can't they still make money selling an unlocked Core i5 quad for $150, and with HyperThreading enabled for another $50 or so?

(If Intel does go ahead with their rumored plans to increase mainstream CPU line core counts, we might see exactly this kind of price compression on their quad-core parts.)

I will say, that if the price of mainstream quad Core CPUs drops, then it would also be a good thing for the upgrade (every generation) market, if Intel could plan ahead a little, and stabilize their socket / pin-out, so that it would last more than one tick / tock cycle.

I would favor a change back to the slot-style CPU "cartridge", but with a standardized WC connection. The chassis of the future would have integrated Rads for WC the CPU.

 

[cytg111]

I think there is so many variables involved that i cant even

 

[Roland00Address]

100 units sold at 100 dollars is the same revenue as

125 units sold at 80 dollars

Are you seriously trying to tell me that a 20% price drop on cpu prices is going to yield 25% more total sales, especially when the cpu is just part of the total BOM for a computer.

For example that $100 dollar cpu may be part of a $300 dollar computer, so are you telling me a $280 dollar computer vs a $300 dollar computer is going to yield 25% more sales?

-------

How markets work with maximizing revenue is different when there is 1 seller, vs 2 sellers, vs many sellers. Markets also work differently when you can can create multiple price tiers by changing the quality of the good on both the final product but also change the quality of the good based on skimping on the raw cost (aka make a crappier processor but the processor has a smaller die area so even if the die area to performance is not good you may choose a number not on optimizing total performance but instead optimizing revenue)

 

[jhu]

100 units sold at 100 dollars is the same revenue as

125 units sold at 80 dollars

Are you seriously trying to tell me that a 20% price drop on cpu prices is going to yield 25% more total sales, especially when the cpu is just part of the total BOM for a computer.

For example that $100 dollar cpu may be part of a $300 dollar computer, so are you telling me a $280 dollar computer vs a $300 dollar computer is going to yield 25% more sales?

-------

How markets work with maximizing revenue is different when there is 1 seller, vs 2 sellers, vs many sellers. Markets also work differently when you can change the quality of the good on both the final product but also change the quality of the good based on skimping on the raw cost (aka make a crappier processor but the processor has a smaller die area so even if the die area to performance is not good you may choose a number not on optimizing total performance but instead optimizing revenue)

There's also inflation to take into account. Intel keeps the nominal price the same. But in inflation adjusted dollars, the price has actually gone down.

 

[CHADBOGA]

I really hate going through the ordeal of installing a new O/S and installing all my applications back on, so that is the biggest obstacle to me updating regularly, but there are other reasons too, like:

1. I haven't ever sold any of my computer equipment to anyone else and couldn't be bothered to start doing it now.

2. Why would I bother for a pissy 20% improvement? I would want a minimum 100% improvement before I update.

3. I'm lazy and have better things to do with my time than update a PC for no good reason.

 

[IntelUser2000]

There's also inflation to take into account. Intel keeps the nominal price the same. But in inflation adjusted dollars, the price has actually gone down.

Inflation adjusted pricing has remained the same since Sandy Bridge for the top chips.

2600K: $317
3770K: $313
4770K: $339
6770K: $350

Also, Intel's PC revenue declined far less than PC volume decrease. Likely due to increasing ASPs.

How? By making things like $1500+ Iris Pro 540 devices. Also on the HEDT platform the 10 core version is at $1500, or 50% more than predecessor.

Pricing for server chips also went up drastically. It went from $2.6k on the top SKU to 4k now. The 4P variants cost $7k, when it used to cost $3.6k. Of course actual customer pricing will be cheaper, but it hints at ASP increase.

Intel also points out they keep having record Core ix sales. ASP increase.

I say they are really being dumb though. Why not increase ASPs in a targeted way? I mean, have GT3e eDRAM on a Core i3 6100 part and price it $50 higher at $160? Lot more people will buy it. That's how you steal markets from discrete cards, not by putting $50 iGPU on a $350 one and making it $400. If they want to stick a iGPU on a halo $350 CPU, heck go all the way and add a $200 one with 216EUs.

 

[NTMBK]

Nah, better off buying a bunch of old Llano dual cores :awe:

 

[Yuriman]

If Intel released a generation with major price cuts, I'd upgrade once, then continue not upgrading for a while again until I could get a significantly better CPU at the same price.

 

[frozentundra123456]

Theoretically, though, from SB to Skylake, there have been 2 die shrinks, so the acutal cost of the cpu should be less. Now granted, you get a bigger igp, and die shrinks are getting more difficult, so I dont know if the actual cost has decreased, but I expect it has.

The problem I have with pricing is not the price itself so much as the segmentation of the best features like edram to extremely expensive, high end devices. There are 3 things intel could do to make cpus a much better value:

1. Make hyperthreading available on more models
2. Make edram more available
3. Mainstream hex core

 

[therealnickdanger]

That's a lot of words to simply say "I wish CPUs were cheaper."

Talking out of my butt, but here's my take:

The slowness in absolute performance growth from Intel is a direct result of them chasing mobility and battery life. We are, however, still getting IPC improvements and more and more useful functions on-chip that far exceed IPC (Quick Sync, AES, etc.), but also some significant power consumption reductions to boot. When taking everything into perspective, I'm not really disappointed by Intel directly. If they produced a CPU with 200% IPC improvements, most applications would be I/O bound in one way or another.

Right now, as I see it, we're going through a bit of a software Renaissance in which the underlying code of operating systems and APIs are being retooled to take advantage of the mainstream application of 64-bits, large memory pools, and 4-core+ (logical) systems. Most people here admit to this very fact by virtue of complaining about it 24/7. We complain about RAM not being utilized and cores not being utilized all the time.

While CPUs may not be very exciting at the moment, there are neat things happening. GPUs are getting more intricate, exotic memory interfaces are emerging, NVMe SSDs, adaptive sync displays... In reality, I see it as the rest of the tech world catching up to CPUs. I think we'll see the CPU pendulum shift back to performance+cores once the application demand gets there.

And that's when I think we'll see quads overtake duals in the mainstream price point and hex/octos overtake quads in their price point. Assuming we're not all forced into the cloud by then...

 

[PingSpike]

Man, I need something. They aren't cheaper. They don't have more PCI-e lanes and other chipset features are doled out at a snail's pace. They don't have more cores. They barely have more performance. The power consumption is well into the diminished returns / who cares? category. And I have to buy a new motherboard (because Intel) and now RAM to even do the upgrade.

CHADBOGA is right. Upgrading is a bit of a pain. And the new platform might have bugs or not overclock as well if that your thing. And what do you get for your trouble? As it stands now upgrading basically gives you all the negatives and barely any positives, while costing money. Better to take that money and blow it on a video card or a bigger SSD or a better monitor or something where you'll get an improvement of some kind you can notice.

 

[sxr7171]

It's not really the money, it's just not worth even getting up for these piddling increases. Even if they offer a trade in upgrade for $100 you still have disassemble your system/cooling and put the new CPU in. Not even worth it IMHO. Not even at $50.

 

[StrangerGuy]

It's not really the money, it's just not worth even getting up for these piddling increases. Even if they offer a trade in upgrade for $100 you still have disassemble your system/cooling and put the new CPU in. Not even worth it IMHO. Not even at $50.

I have no need for M2, NVMe, USB3.1 or all SATA3 ports on Skylake either. Those are features worthless in the real world as far as gaming is concerned. I guess that why people are getting increasingly jaded about tech in general: lots of nice sounding bullet points, but with next to no marginal utility increases.

 

[escrow4]

I was thinking of building of a Kaby Lake mini-ITX system but this 5930K tower has so much balls it would strangle me in my sleep. :sneaky:

 

[NTMBK]

Seems like the pattern in AT users is ~5 years between CPUs, ~2-3 years between GPUs.

 

[Denithor]

I say they are really being dumb though. Why not increase ASPs in a targeted way? I mean, have GT3e eDRAM on a Core i3 6100 part and price it $50 higher at $160? Lot more people will buy it. That's how you steal markets from discrete cards, not by putting $50 iGPU on a $350 one and making it $400. If they want to stick a iGPU on a halo $350 CPU, heck go all the way and add a $200 one with 216EUs.

This. Honestly, they are pretty stupid not doing exactly this. Except that it would bankrupt AMD and get them into trouble with regulatory bodies around the entire planet. But, yeah, putting the top end iGPU only on the flagship quad, where almost nobody will use it anyway, is about as dumb as dumb can be.

 

[Denithor]

I was thinking of building of a Kaby Lake mini-ITX system but this 5930K tower has so much balls it would strangle me in my sleep. :sneaky:

I LOL'ed.

 

[Dahak]

Seems like the pattern in AT users is ~5 years between CPUs, ~2-3 years between GPUs.

Thats generally how I upgrade, usually 5-6 years, and a gpu upgrade mid way through

The reason being as others have stated, the performance increase in each upgrade is not worth the time and hassle most of the time.

 

[dave_the_nerd]

Just a hypothetical - if the i5-6600K (and successors) was $150, and the i7-6700K (and successors) was $200-230, would you as a consumer and an enthusiast, upgrade every generation, rather than hold on to the same CPU through 4-5 generations?

No. Because I'm not you.

But seriously, no. Because the CPU cost is only a part of the platform cost (RAM, Motherboard, whatever other parts need to be replaced because they're incompatible) and the bottom line is what I'm looking at, not just the CPU cost. Saving $50 on a CPU is nice, but Microcenter already basically makes that happen, and I'm not upgrading every gen.

Anyway, the old joke applies:

...

So the bartender says, "Well, when beers were $1, all the tourists came and drank 6 beers and left. When beers were $3, all the tourists came and drank 6 beers and left. Now that beers are $5, all the tourists come and drink 5 beers and leave. So now I work less and make more money.

 

[Yuriman]

Desktops are an afterthought. When it comes to notebooks, cutting power consumption is a big deal. Moving from 2.5" SATA to M.2 is a big deal. A single port that can be used for charging as well as connecting peripherals is a big deal. A bigger iGPU is a big deal for some, and Intel is increasingly including hardware accelerated functions.

Consider when landlines were first being replaced by mobile phones. For a long time, people had both, until a mobile phone became small enough, useful enough, and had enough battery life so as to make a conventional phone practically useless. That's where we're headed in computing.