Intel 4-5 year tick/tock cycles when AMD is done?

[ShintaiDK]

Clock for Clock and Power / Performance AMD hasn't been very competitive since the first generation i7 Nehalem.

However, because both consoles this generation (PS4/XBONE) have gone to AMD 8 Core Jaguar CPU / GCN 7750GPU I believe the AMD FX 8 core cpus will have a long lifetime in terms of PC ports of those console games.

For gaming purposes I don't think an AMD FX 8 core is a bad investment at all, despite the platform limitations (USB 2 / PCI-E 2.0).

Consoles only use 6 cores max for gaming. Not 8. So even if you wish to promote the FX series, it would only be the FX6xxx.

And yes, the FX8xxx is terrible. Also the FX doesnt have anything in common really with the Jaguar cores. So its not like its optimized for the uarch either.

Even AMD doesnt believe in the FX CPUs.

 

[sxr7171]

That's what intel has been doing. SB is basically nehalem. We havent had any true innovations in their architecture since then. A processor from 5 years ago is in the neighborhood of the latest and greatest haswell in terms of raw performance. But in the smartphone space, a processor from 5 years ago can barely even run a modern smartphone OS or the typical app suite.

But even without AMD the market is still punishing them for it. Intel's lack of innovation has allowed smartphones to pop up like weeds, sort of like how weeds grow around a boulder. Billions of em, 99% powered by competitors.

In 5 years the majority of desktop PC users will have entirely replaced their desktop with a smartphone. Expect there to be no more than one actual pc and several smartphones connectable to several external displays (dumb terminals). In other words, expect no more than one intel x86 microprocessor in each home. I'd say the market is delivering intel the justice it deserves.

Well said. This is the new reality. The IPC improvements are minor each generation. The only difference between nehalem and today's CPUs are power use and heat generation.

I never use my desktop anymore. Even if have a small laptop next to me, 99% of my usage is on my smartphone. This very post is from my phone. My only use for a desktop is for a gaming machine. I don't even set up email or any non gaming apps on it. If MS made a gaming optimized Windows I would use that. Just a pared down 2GB OS for gaming only.

I'm saying this as a big PC fan. Think about the new generation. Many will not even use a PC outside work or school. As those purposes become more cloud based the following generation may not ever use anything beyond a terminal.

 

[videogames101]

Well said. This is the new reality. The IPC improvements are minor each generation. The only difference between nehalem and today's CPUs are power use and heat generation.

I never use my desktop anymore. Even if have a small laptop next to me, 99% of my usage is on my smartphone. This very post is from my phone. My only use for a desktop is for a gaming machine. I don't even set up email or any non gaming apps on it. If MS made a gaming optimized Windows I would use that. Just a pared down 2GB OS for gaming only.

I'm saying this as a big PC fan. Think about the new generation. Many will not even use a PC outside work or school. As those purposes become more cloud based the following generation may not ever use anything beyond a terminal.

terminals connect to something, Intel's server market is secure in my view

 

[sxr7171]

terminals connect to something, Intel's server market is secure in my view

Sure but now it's like one Intel CPU for 5-10 terminals whereas before it would have been 5-10 Intel CPUs sold for the same thing. And maybe even ARM based servers would eat into that space.

 

[witeken]

According to the economic people, the PC market bottomed and should see small improvements, so Intel's PC revenues aren't going to collapse any time soon.

 

[dguy6789]

snip

I sincerely hope I didn't just read that you think if Intel was the only x86 company left that prices would still be reasonable or that they would still produce desirable upgrades in a reasonable time frame.

 

[CHADBOGA]

I sincerely hope I didn't just read that you think if Intel was the only x86 company left that prices would still be reasonable or that they would still produce desirable upgrades in a reasonable time frame.

If AMD died, I think Intel would probably limit their price increases to no more than 20% and quite possibly less than 10%.

The reason for this is that they don't want to create any more momentum for ARM to move up the food chain and because AMD is already largely irrelevant to Intel's pricing strategy.

Intel's greatest competition in the x86 market, doesn't come from AMD, it comes from the installed base of Intel CPU's.

Many people upgrade after 3 or 4 years because they can get something notably better. If there isn't something notably better, people will hang on longer to their CPU's and Intel's Fabs will sit idle.

 

[Arkaign]

Consoles only use 6 cores max for gaming. Not 8. So even if you wish to promote the FX series, it would only be the FX6xxx.

And yes, the FX8xxx is terrible. Also the FX doesnt have anything in common really with the Jaguar cores. So its not like its optimized for the uarch either.

Even AMD doesnt believe in the FX CPUs.

I dunno, I think the Visheras are okay in many ways (particularly outside of PPW). I always have an i7 in my desktop, but my HTPC actually has a 6350 in it at the moment because I was able to get a mobo/cpu for $79 with USB3/etc. I had no options from Intel at that price, and it's overkill for the job anyway.

What would have been interesting would have been AMD if they hadn't sold off their fabs and bought ATI. If AMD had been truly close to Intel process tech or basically equal, I think they wouldn't have suffered nearly as much.

Even with inflated mining prices and two major console contracts, I can't see that being peanuts compared to what Intel will make by having the lions share of backoffice, desktop, and laptop. Intel has so much cashflow that it's almost ludicrous, and they can hire the best, build the best fabs, and once they start focusing on something like mobile, let's just say I would much rather be them than the competition.

Intel on 10 and 14nm for phone/tablet is going to be bananas.

 

[SiliconWars]

What would have been interesting would have been AMD if they hadn't sold off their fabs and bought ATI. If AMD had been truly close to Intel process tech or basically equal, I think they wouldn't have suffered nearly as much.

The 32nm process AMD has been using for Bulldozer etc is the exact same 32nm process they'd have been using had they not sold the fabs.

Had they not bought ATI they'd have been left without Jaguar, Kabini, Trinity etc. Just Bulldozer/Vishera on that same 32nm.

You're right they wouldn't have "suffered" as much though - but that's because they'd have died 3 years ago.

Intel has so much cashflow that it's almost ludicrous, and they can hire the best, build the best fabs, and once they start focusing on something like mobile, let's just say I would much rather be them than the competition.

Intel's only focus these days is mobile and it's been that way for the past couple of years at least - the problem is they are skating to the puck, not to where it's going.

They are stuck with the base architecture for the next 4+ years anyway. It's not good enough and far too many mistakes were made on Silvermont.

 

[Boze]

The question that I think about often is, "Can AMD turn their situation around?"

Late last year I upgraded from an AMD Athlon II X3 425, 4 GB RAM, GTS 250, to an i7-4820K, X79 MB, 32 GB of RAM, and a GTX 770.

I was really torn about it too. I wanted to pick up an FX-9590 or whatever their flagship chip is nowadays, but at the insane TDP (220 watts) and the enormous performance delta, I really had no choice. What's worse, the i7-4820K was $319.99 from Amazon and the 9590 was $329.99.

I've been using AMD processors since around 2000, so it made me sick to my stomach to abandon them, but at the end of the day, they just can't compete any longer.

What do you folks think it would take for AMD to come out of the gates with a processor that makes Intel's engineers sweat?

 

[Fox5]

A certain Anand Lal Shimpi would disagree with you.



http://www.anandtech.com/show/2702/20

It was Sandy Bridge and Bulldozer when the wheels really fell off for AMD, not Conroe.

Core 2 crushed the Athlon X2s, and Phenom was no competition.
Phenom II was a valiant competitor against Core 2 Quad, but it came out as Nehalem was hitting the scene. Even that wasn't too bad, but Sandy through Haswell have been slow but steady improvements while AMD has barely improved in per core performance.

The processnode is the cheapest part R&D wise. You can simply compare TSMC to Intels R&D budget for example. IC design is where the R&D major expense is. A new uarch like Haswell most likely cost 5-6B$.

I'm pretty sure it's the reverse. See: the size of GloFo vs AMD.
IC design CAN be expensive, but it can also be cheap. $100 - $200 million will design a cpu that can go toe to toe with ARM. A haswell level cpu? I'd guess $1-$2 billion.

 

[ShintaiDK]

I sincerely hope I didn't just read that you think if Intel was the only x86 company left that prices would still be reasonable or that they would still produce desirable upgrades in a reasonable time frame.

You should read again then. Because it seems you assume there is a static demand for CPUs. And thats a false assumption.

Prices will not increase and innovation will not suffer. If anything, prices would actually lower a tiny tad.

 

[ShintaiDK]

The question that I think about often is, "Can AMD turn their situation around?"

They cant. No need to try and fool yourself thinking otherwise. Its simple a numbers game. AMDs CPU division also keep shrinkign rapidly. AMD will be an embedded company with dGPUs, until they run out of steam there too.

And Xbox One sales are already collapsing. January numbers showed an outright disaster. So the cashflow from there already seems to be in trouble. And they depend on that while they restructure the company.

What do you folks think it would take for AMD to come out of the gates with a processor that makes Intel's engineers sweat?

The reborn of christ. Its simply not gonna happen. AMD isnt interested in big x86 cores anymore either.

AMD spends 1.2B$ on R&D. Intel 10.6B$. Even nVidia outspends AMD in R&D. And it shows.

 

[ShintaiDK]

I would be very surprised if this is true.

I'm pretty sure it's the reverse. See: the size of GloFo vs AMD.
IC design CAN be expensive, but it can also be cheap. $100 - $200 million will design a cpu that can go toe to toe with ARM. A haswell level cpu? I'd guess $1-$2 billion.

http://www.zdnet.com/cn/tsmc-leads-taiwans-r-and-d-spending-in-h1-2013-7000019880/

TSMC roughly used 1½B$ on R&D in 2013. Intel used 10.6B$.

Even if we generously double Intels R&D on the processnode compared to TSMC. It still leaves over 7B$ a year for other parts.

 

[Homeles]

I sincerely hope I didn't just read that you think if Intel was the only x86 company left that prices would still be reasonable or that they would still produce desirable upgrades in a reasonable time frame.

They virtually are the only x86 company. The prices you see now are basically monopolistic. If they were to raise prices much higher than they are, they would lose out, especially when you consider that the PC market is struggling to grow as is.

 

[NTMBK]

Core 2 crushed the Athlon X2s, and Phenom was no competition.
Phenom II was a valiant competitor against Core 2 Quad, but it came out as Nehalem was hitting the scene. Even that wasn't too bad, but Sandy through Haswell have been slow but steady improvements while AMD has barely improved in per core performance.

Yeah, I'd agree with that assessment. I just wanted to point out that it's not like Intel could just have sat on their laurels since Conroe, which is what Shintai was implying.

 

[witeken]

I sincerely hope I didn't just read that you think if Intel was the only x86 company left that prices would still be reasonable or that they would still produce desirable upgrades in a reasonable time frame.

Prices will not increase and innovation will not suffer. If anything, prices would actually lower a tiny tad.

Exact-ly.

"We've been following Moore's law for 50 years. How much more competitive can you get?" --Craig Barrett

 

[witeken]

http://www.zdnet.com/cn/tsmc-leads-taiwans-r-and-d-spending-in-h1-2013-7000019880/

TSMC roughly used 1½B$ on R&D in 2013. Intel used 10.6B$.

Even if we generously double Intels R&D on the processnode compared to TSMC. It still leaves over 7B$ a year for other parts.

Intel also has a lot of other costs. Here's Intel's R&D statement for 2012... for some reason I can't copy it, you can look at your own here: 2012 Annual Report.

I wanted to quote the "Research and Development" piece on page 41.

If R&D for new process nodes would be 2B$ and a new microarchitecture would be 7B$, you forget a whole bunch of other things that also cost money (smartphones, tablets, ultrabooks, data center, 450mm wafers/EUV; Intel's investment in ASML, start-up costs new process nodes).

I think that a new microarchitecture costs a lot less than the multibillions you suppose, and that a new process might cost quite a bit more than 2B$.

 

[ShintaiDK]

If R&D for new process nodes would be 2B$ and a new microarchitecture would be 7B$, you forget a whole bunch of other things that also cost money (smartphones, tablets, ultrabooks, data center, 450mm wafers/EUV; Intel's investment in ASML, start-up costs new process nodes).

450mm/EUV is under the node R&D. ASML investment is not under R&D at all.

Everything else goes to different IC designs more or less. Alittle bit gets used on software.

A uarch like Haswell is under R&D for 4-5 years.

Qualcomm is fabless. And they use 3.4B$ just on modems and cheap ARM chips. Its all IC designs for them.

 

[William Gaatjes]

FarCry 2 destroy both AMD processors and video cards and its made with a so outdated tech... It should not count as a bench for a test...

http://www.lostcircuits.com/mambo//...sk=view&id=64&Itemid=42&limit=1&limitstart=17

Looking at the upper picture with FARCRY 2 benchmark results, it does not look that bad.
I would rather go for average fps than just use the maximum fps.
When taking average fps, the AMD cpu's can keep up.

 

[jpiniero]

IIRC I number I remember is that Intel was spending 4-5x the competition on Fab tech.

They virtually are the only x86 company. The prices you see now are basically monopolistic. If they were to raise prices much higher than they are, they would lose out, especially when you consider that the PC market is struggling to grow as is.

Just because AMD isn't competitive at the high end doesn't mean they aren't at the mid range and low end. Intel has to wait until AMD is completely gone before they can change things to maximize profitability on people who are tied to Windows. The Windows part is key since people who aren't tied would just switch to ARM if Intel got too obnoxious.

 

[witeken]

450mm/EUV is under the node R&D. ASML investment is not under R&D at all.

Everything else goes to different IC designs more or less. Alittle bit gets used on software.

A uarch like Haswell is under R&D for 4-5 years.

Qualcomm is fabless. And they use 3.4B$ just on modems and cheap ARM chips. Its all IC designs for them.

Interesting, I really wish we had more detailed numbers of Intel's R&D spending. BTW, Intel spends another ~12B$ on capital expenditures. I guess I falsely assumed that was included in the R&D budget: 53B$*60% profit margin - 18B$ (R&D and MG&A) - 25% tax rate - 4B$ dividend - 4B$ stock repurchase - acquisitions already gives a pretty low profit, so it seemed unlikely to me that there was place for another 10B$ capex. I'm clearly not a financial expert. Fortunately, this graph makes it clear:

 

[el etro]

Looking at the upper picture with FARCRY 2 benchmark results, it does not look that bad.
I would rather go for average fps than just use the maximum fps.
When taking average fps, the AMD cpu's can keep up.

Yes, is not so bad the CPU difference, but test has made at GPU limited(1080p 8xMSAA) settings.

 

[mrmt]

Interesting, I really wish we had more detailed numbers of Intel's R&D spending. BTW, Intel spends another ~12B$ on capital expenditures. I guess I falsely assumed that was included in the R&D budget: 53B$*60% profit margin - 18B$ (R&D and MG&A) - 25% tax rate - 4B$ dividend - 4B$ stock repurchase - acquisitions already gives a pretty low profit, so it seemed unlikely to me that there was place for another 10B$ capex. I'm clearly not a financial expert. Fortunately, this graph makes it clear:

You are mixing profits with cash flows.

 

[pablo87]

The question that I think about often is, "Can AMD turn their situation around?"

Late last year I upgraded from an AMD Athlon II X3 425, 4 GB RAM, GTS 250, to an i7-4820K, X79 MB, 32 GB of RAM, and a GTX 770.

I was really torn about it too. I wanted to pick up an FX-9590 or whatever their flagship chip is nowadays, but at the insane TDP (220 watts) and the enormous performance delta, I really had no choice. What's worse, the i7-4820K was $319.99 from Amazon and the 9590 was $329.99.

I've been using AMD processors since around 2000, so it made me sick to my stomach to abandon them, but at the end of the day, they just can't compete any longer.

What do you folks think it would take for AMD to come out of the gates with a processor that makes Intel's engineers sweat?

Don't think they can. As a GPU company, they should be abandoning big CPU cores and focusing on a single architecture, small x86 core. A, it probably has more room to improve performance and power mgmt. B, co. Overhead would come down*, profits would be the new normal, customer confidence would grow. C, they've hardly milked the small core architecture permutations wise: 2/4/8 cores * 3 GPUs * DDR or GDDR * big L2 vs small L2 = 36 flavors on paper (personally I'd like a 2core, big GPU, gddr, big L2), or more importantly, application specific. Right now, its spray and pray.

You as a power user might not buy their desktop, but you might consider an AMD laptop if the APU had a big GPU core running off GDDR5.

By the way, SSD really devalues CPUs. Its shocking that most retail. PC's are still Winchester.

* pre AMD, ATI had $170M quarterly SG&A on $635MM revenue IIRC. And they were doing chipsets, TV chips and the ADRENO business (sold to Qualcomm).