[Luis Dias]

Very dead.

I still can't wrap my head around it. Why then go all the way down to 10 nm in the first place? Why all the talks about 7 and 5 nm? If it's all about just optimization, then why even bother.

[Phantom Hoover]

DISREGARD THAT I SUCK COCKS

[Spoon]

Yeah no idea either, maybe smaller nm numbers are just kind of akin to blast processing level of marketing talk nowadays.
"Look guys, we have the smallest numbers, buy our latest generation of processors that runs on 10 nm!"

[FrikgFeek]

Maybe. Intel switched from 22nm to 14nm architecture in Q2 2015, more than 1.5 years ago. And we haven't really seen noticeable performance increases from that switch. Or any increases at all, the 5th and 6th gen of intel i7s are weaker than the 4th, a bit weird since they're on 14nm and 4th gen is on 22nm.

[Phantom Hoover]

Process nanometre numbers are essentially completely made up these days, they don't correspond to the size of any feature on the actual chip.

[Mito [PL]]

The fact is that smaller transistors need less power to operate and more of them can be present in the same volume of space. And since it seems that current technology has bumped into a wall regarding (reasonable) clock speeds, the way to improve the raw processing power of a CPU is to place more data processing units in it and coordinate them properly via architecture and software. Using smaller transistors allows for placing more of them in the same volume of space and having the same power usage as before.

And I also think that this lack of performance improvements comes directly from Intel's monopoly in high-performance and high-price part of CPU market. Why would you modernise antything when what you've got now is still only a dream for about 90% of your customers?
Oh, and 22-core hyperthreaded Xeons for proffesional usage...

[FrikgFeek]

And I also think that this lack of performance improvements comes directly from Intel's monopoly in high-performance and high-price part of CPU market. Why would you modernise antything when what you've got now is still only a dream for about 90% of your customers?

You could say the same thing about the GPU market, most people can't afford a GTX 1080 or 1070. Most people couldn't afford a 980 2 years ago yet modern GPUs have advanced since then. Something that was bleeding edge 3 years ago should be mid-end today and that's how it is with GPUs. The new RX480 is 200$ and has almost the same power as the 550$ 290X did 3 years ago.

If AMD was pressing Intel to actually improve their tech year by year a 4790K wouldn't be "a dream". You'd be able to buy a processor with similar power for a reasonable price and the top-end 350$ i7 would walk all over it in benchmarks.

[Phantom Hoover]

You're comparing apples to oranges, GPUs are substantially different from CPUs, haven't hit physical diminishing returns nearly as hard and are able to mix up their architecture much more freely than Intel are.

[Mito [PL]]

Well, I'd say that GPU's are much further away from hitting the maximum frequency limitations for silicon than CPU's. I didn't ever see a 3GHz clocked GPU.

Plus, it would seem that most of new game engines give better results for medicore CPU & monster GPU than monster CPU & medicore GPU. Of course, there are exceptions (Minecraft?). Investing more in GPU than CPU seems to be simply a better option if you have to choose because you can't have both for cash reasons.

[FrikgFeek]

Oh, sure, there are limitations that prevent rapid advancement of performance. But if AMD was actually competitive Intel would probably find a way to improve performance a bit more. Or if they couldn't they'd have to start lowering the prices to match AMD.

Both of those are much better options than the current intel monopoly where they resell the same performance for the same price for 3 years in a row.

[Phantom Hoover]

Plus, it would seem that most of new game engines give better results for medicore CPU & monster GPU than monster CPU & medicore GPU. Of course, there are exceptions (Minecraft?). Investing more in GPU than CPU seems to be simply a better option if you have to choose because you can't have both for cash reasons.

If you hit a GPU bottleneck in a game you can just turn the graphics down. If you hit a CPU bottleneck there's generally nothing you can do about it without compromising the gameplay.

[Gortef]

All I can say that I am honestly really hyped for Ryzen. All the bits and pieces so far seems to indicate that AMD does have something good coming.
Only problem so far seems to be people who either accidentally or intentionally over hype things for whatever reason. Kind of like what happened with RX480.

But anyway, I am hopeful that around the end of February I have a full AMD rig after a decade.

[LaineyBugsDaddy]

Sadly for me, at this point with my resources, I'd be glad just to a socket AM3 board and 8GB DDR3 so that I could take full advantage of my Phenom II X4. This would also let me massively upgrade my wife's rig, as she'd inherit my Phenom X3, the current board, and six gigs of DDR2 800.

[jr2]

Can probably get that pretty cheap on eBay especially after Ryzen comes out, no?

[Dragon]

Very dead.

I still can't wrap my head around it. Why then go all the way down to 10 nm in the first place? Why all the talks about 7 and 5 nm? If it's all about just optimization, then why even bother.

As far as I can tell, we're hitting the point at which Moore's law is running head-first into physics constraints. It might be less due to Intel getting complacent and more about the fact that that it may not even be possible to squeeze more performance out of a single chip. Intel has simply hit the limits for silicon electronics and is just incrementing numbers for marketing purposes (along with some minimal architecture optimizations).

Photonics seem like the only way to progress in that case. Using light instead of electrons could allow higher speeds and even smaller sizes, but this field is relatively recent, so we don't know if it could actually be a practical solution for a desktop computer.

BTW, would there be any benefits to using multiple CPUs? IIRC, there are "processor cards" on market, but they seem to be meant for professional applications such as data analysis and scientific calculations. I'd imagine they'd have the same problems as multicore processors, only worse.

[jr2]

IIRC multiple CPUs back when dual CPU mobos were a thing would only allow for ~30% more speed?  Of course this may have changed since the dual PIII days.

[LaineyBugsDaddy]

Photonics? That name is so meh to me. But then, being a Trekkie, I would want to see it called optronics, and have actual ODN relays and such. That's just the kind of nerd I am.

BTW, I thought they were using copper transistors now, not silicon. Hence, IIRC, the code name of the initial generation of chips that did it being "Coppermine."

[Dragon]

As long as they're using electrons for doing the logic, they're going up against the same limits imposed by electrons' charge and wavelength. Copper might help with heat dissipation, but not by much, I think.

And yeah, optronics sounds nice, but I'm afraid people would think that it's a short for optoelectronics (photodiodes and such) which is another thing entirely.

[Axem]

Copper transistors? That wouldn't work at all (unless you're reworking the laws of physics (in which case give me your secrets!)). Silicon is semiconductor, which means you can basically control if its conducting or not. That's what makes the transistor into a switch and makes computers tick. Copper is a conductor, a very very good one at that, so making copper transistors wouldn't do much other than generate heat. (But it does carry heat away very nicely!)

[LaineyBugsDaddy]

It might have been that copper was being used for the doping material. I'm a little fuzzy on the details.