Following the CPU market has lost it's excitement for me

[Hulk]

I used to love studying the latest CPU trends. My first "IBM compatible" had an SX486, then I moved to a Pentium and read all about the massive architectural improvement and saw them in action running software which was seriously laggy with the 486. Then a Celeron 300a with the 450 overclock, more improvements, on die cache... lots of cool stuff. Clocks were ramping up like crazy. Later a 833MHz PIII, a few P4's, a Core2Duo, Sandy Bridge... and now Haswell. But that is where my upgrades and interest has seemed to stall.

The incredible increase in clockspeeds came to a near halt 10 years ago. Disappointing but there were great strides being made in architecture and multicores so things were still exciting. Anand would always have a deep dive into the latest architecture and that would usually take me a month or more of repeated readings and comparing to previous architectures to fully understand. It was fun.

But now the deep dives are all but gone. Intel isn't really releasing the detailed information like they used to. Or perhaps there really isn't much new to report.
Then desktop releases kind of stalled with Broadwell. Then "tick tock" broke.

I really don't see anything Intel CPU on the horizon that excites me. I have no urge to move from my 4770k, which honestly wasn't a huge upgrade from my 2500k.

Seems like this golden age of cpu's is over.

If you been closely following the x86 market for the last 10 or 20 years have you also lost excitement for this pursuit?

 

[swilli89]

We are on the eve of a huge release from AMD, perhaps their biggest in a decade, and you're not excited about CPUs?

 

[zinfamous]

We are on the eve of a huge release from AMD, perhaps their biggest in a decade, and you're not excited about CPUs?

Even if Zen turns out to be God's gift to nerds, the tech isn't all that exciting compared to everything out there today, and the last 5+ years. it won't possibly be anything as cool, new, and revolutionary as the first Pentium or the first AMD64 chips.

OP's point is that design was exciting and had huge momentum back in those days, but things have just gotten boring--mostly tweaks and improved efficiency on relatively identical backbones. But, that's what happens when you hit the barriers inherent in physical laws.

 

[DrMrLordX]

Seems like this golden age of cpu's is over.

<pedantic>

But the golden age of misusing apostrophes is just beginning!

</pedantic>

 

[TheELF]

Hulk...late to party...all guests gone... long ago.

Ever since motorola gave up on their 68xxx design * people knew that things where getting to an end,when big companies like NEC SiS and so many others stoped making x86 CPUs because they where hitting walls people where diying inside.
At the latest when Apple was forced to change CPU platforms for the second time, and with that made clear the limitations of risc/arm,despair was high.
After that even AMD gave up and went the way of many cores trying to convince people that it was something.
Now even intel seems to go this route with kaby lake probabbly having at least one hex core...very sad times indeed when you are seeing your love die over and over...


* hugely spread in *nix servers,home computers,game consoles,(laser printers,audio hardware,etc) you name it,huge market share.

 

[NTMBK]

Have you tried drinking heavily instead? I recommend gin.

 

[AtenRa]

Have you tried drinking heavily instead? I recommend gin.

You're the first Scotch i hear to recommend gin for getting pissed

 

[Hulk]

We are on the eve of a huge release from AMD, perhaps their biggest in a decade, and you're not excited about CPUs?

There have been so many "AMD is back!" products since C2D put the nail in the Athlon's coffin that I honestly stopped paying attention. Seems like AMD just falls further and further back. A success is catching up a bit and only being 2 or 3 generations behind instead of 4 or 5. I hope Zen is all you think it will be and put some heat on Intel. Problem is clockspeeds have topped out and IPC seems to have as well. Right now i'd just be happy if Zen could push Intel to release an 8/16 core mainstream (affordable) part.

 

[NTMBK]

You're the first Scotch i hear to recommend gin for getting pissed

Well remembered, but I'm actually an Englishman living in exile!

 

[AtenRa]

Well remembered, but I'm actually an Englishman living in exile!

ehehe

 

[moonbogg]

CPU's have become boring for me also. The only thing exciting is all the drama and forum fighting about Zen. I also think that consumer applications have stalled out as well regarding how much CPU power they need. I'm guessing that as programming complexity increases, things get more expensive and difficult. If a super fast CPU came out today, I think the potential for more exciting programs and games would be there, but wouldn't those programs and games also be more difficult and expensive to make?
As complexity increases, human relative capability decreases. What we need more than faster CPU's is a better way to make programs. We need AI to take over both software creation and CPU design. So, instead of trying to design a better CPU, people should focus more on reverse engineering the brain and make a proper AI to handle life's complexities for us.

 

[Hulk]

Yeah alcohol definitely helps as the Redds apple ale on my desk right now can attest.
I'll still be following developments in the CPU market but not with the passion I once had.
For example, the forthcoming 7700k. What I want to know it how it compared to it's predecessors at the same clock in the same apps. Clock speed has stalled so IPC is really all that matters at this point to me.
What is the actual IPC improvement in say video encoding from Broadwell to Skylake to Kaby Lake? We don't know.
When I upgraded from my P4 to my dual core C2D I experience a massive boost in performance, same when I went from the C2D to 2500k. A small improvement from 2500k to 4770k. Now an upgrade would barely be noticeable unless I'm measuring some application that uses new instructions only in the new processor.

 

[happy medium]

In 2012 Intel had 85%+ of the cpu market, I'd guess its even more now.
Zen will at best equal Haswell/skylake cpu's.
We have seen very little IPC improvement since Sandy bridge.
Once Zen arrives, my guess is AMD's IPC will come to a slow trickle just like Intel has.

In short I believe cpu's will get even more boring after Zen releases.
In 2018 I think its time for the "more core" race. The new software/games will have better support for it.

 

[WhoBeDaPlaya]

It's already boring when you can build a 16c/32t 128GB monster for < $500 :
http://www.natex.us/mobile/Product.aspx?ProductCode=S2600CP-SR0H8-128GB-12800

Or you snag a used workstation and pop dual 6c/12t Xeons in them (what I did with two HP Z800s running 2x X5675s each).

Hard to match the price/performance of those, with the provisos of older platforms, ST being less important and embarrassingly parallel workloads.

 

[happy medium]

It's already boring when you can build a 16c/32t 128GB monster for < $500 :
http://www.natex.us/mobile/Product.aspx?ProductCode=S2600CP-SR0H8-128GB-12800

Or you snag a used workstation and pop dual 6c/12t Xeons in them (what I did with two HP Z800s running 2x X5675s each).

Hard to match the price/performance of those, with the provisos of older platforms, ST being less important and embarrassingly parallel workloads.

wow that's cheap.

 

[frozentundra123456]

We are on the eve of a huge release from AMD, perhaps their biggest in a decade, and you're not excited about CPUs?

Not really. It is big for AMD, and if I were a stockholder I would be excited or terrified, depending on your point of view, but it really isn't, IMO, going to change the cpu landscape dramatically. It will simply be another option of similar performance to intel. It may offer moderately better value for those who want a lot of cores, but I think those who believe AMD will offer comparable performance at a huge discount are deluding themselves. And I certainly don't see them offering better absolute performance than intel. Now if they were to offer a laptop APU with HBM, 1050 Ti graphics and efficient cpu performance at less than a discrete option, that might be exciting.

 

[WhoBeDaPlaya]

wow that's cheap.

Yup. Since most of my heavy compute workloads scale well with threads (eg. vid encoding, 3D rendering, some optics/RF/EM programs that I wrote), it's more economical to build a large core/thread count box for that, and have a high clock + IPC 4c/8t (or hopefully 6c/12t soon) box for everyday needs / gaming.

 

[Xpage]

I'll be excited when somebody decided to put on die (interposer) HBM or the switch to new materials SiGe for transistors so we can have a Ghz race to 10Ghz

 

[reb0rn]

Its true for desktop consumers CPU`s are boring more and more...
R&D even with a lot of cash is producing just few % per 1-2 years, which for ordinary ppl mean almost nothing, but I would not say that will not change
We just need some big breakthrough, and I am sure one day it will happen!

 

[ultimatebob]

Yeah, I blame the lack of competition from AMD. Intel has no push to make rapid technology advancements, so we'll be stuck at 10% performance improvements a year for awhile.

The ARM chips are starting to encroach on x86-64 on the low end, so maybe one of them will become a third player at some point.

 

[dave_the_nerd]

You grew up. Also, the market got boring.

Get a classic motorcycle*. No, not that kind. That kind is for posers. Get that kind. You know the one. Then work on it, and get together with other people who have the same kind and drink beer and talk about how your kind is better than that other kind.

*Or an old muscle car, or a particular type of ham radio, or... well, you know. It's cheaper than a quad-Titan rig.

 

[Hulk]

I don't so much assign blame to any entity on the slowing CPU advancement but if I had to list reasons...

AMD not being competitive with Intel.
Mobile phones and tablet SoC being a real and serious threat to Intel forcing them to move resources away from higher powered desktop parts to the super low power SoC's.
Desktop/laptop performance reaching a "fast enough" point for 99.9% of the population, web browsing, Word, Excel, Photoshop, video editing, ...
And finally, the laws of physics and/or limitation of current silicon manufacturing processes which have topped out clock speeds and are make die shrinks increasingly challenging.

 

[moonbogg]

You guys are all way off base. We need robots to replace humans. That's the only answer.

 

[frozentundra123456]

Actually, I have been shopping for a recent model used car, and there is a lot of exciting computer based technology in autos these days. Anti-lock brakes, stability control and traction control are pretty much standard. Then you have adaptive cruise control, blind spot warning, parking assist, and of course all the computer assisted engine management technology that manages to extract amazing amounts of horsepower out of clean and relatively small engines. And now we even have apple car play and 4g lte hotspots in cars. I dont really know what companies are providing the computer power for all this though.

 

[happy medium]

I dont really know what companies are providing the computer power for all this though.

Nvidia