do you guys think AMD could actually make comeback in CPU market at this point? there market share is so low, I think they going to have hard time even if zen beats expectations. but I'm not economist nor a long time PC gamer, so my opinion isn't worth much.
I think it mainly depends on OEMs using Zen or not. The community of people that build their own desktops is actually pretty low.
Zen will most likely be favoured by OEMs because of their superior iGPU. Instead of spending money on a dedicated GPU and CPU, OEMs can toss in a Zen APU and call it a day. Add in the fact that a 2-core Zen will compete with Haswell i3s and things don't look good for jewtel.
What a bunch of horseshit.
> Zen favoured because of superior iGPU.
The iGPU of current AMD CPUs is already superior to Intel HD Graphics but still AMD is pretty much dead when it comes to CPUs, no matter if OEM or custom built.
It doesn't matter what you or me *think* faggot. What matters are facts and *market shares*. You can fantasize all you want about AMD being favoured in any way, the market shares show a different picture. : ^
If AMDs Zen architecture meets it's performance expectations of forty percent more than excavator, then they have a real chance of being back in the game as long as they can get there pricing right. This year, I truly believe is the year that brings AMD back, or throws them into oblivion. They also have promising products on the graphics card side of it this year.
Family member, one the most sought out features on a laptop is a long battery life. One of the methods used to achieve this is hardware acceleration. Better igpu=better battery life.
They didn't project desktops to be a growth segment for themselves, and were outright clear that enterprise will be their bread and butter. Selling a few million new Opterons in 2017 will bring massive cash flow back into the company, and thats all that really matters.
Go to a store that sells laptops and see their selection. AMD processors are only on shittops and it's usually only the low end chips. 90% of laptops that are $500 and up run on some sort of Intel Processor. Store can't sell what they don't have, and AMD can't gain market share without a presence in stores. Same deal with online retailers. All high end laptops are equipped with Intel procs.
If they come back, it'll be on server processors. Not the gaymen ones you are hoping for, and probably not the laptop ones either.
Perhaps there's something that will be good as media pc / home server again? Who knows.
CPU market is stagnant because Intel does not want to compete seriously against itself for no reason at all.
If AMD threatens any of their lucrative markets, I bet they'll throw one of two-three generations of better designs that they've already planned out there and kick AMD's butts... again.
Only if they can get OEMs on board for low-end and mid-range systems.
They're hitting the limits of silicon, and there is no reason to go hard on performance increases when consumers won't notice them anyway and enterprise customers are using GPGPU for heavy calculation work anyway.
Intel isn't sitting atop a vault of radically faster X86 core architecture just keeping it locked away because they can. They're producing 5-10% performance uplifts per generation because thats all their massive teams of engineers can come up with. X86 is incredibly well fleshed out, billions of dollars, billions of man hours. Every fraction of 1% requires considerable investment. Every dollar buys less and less. Intel primarily competes against themselves, they have for years, and they will continue for years to come. They have to provide enough value added factor to each new line of chips to make a convincing sale, or OEMs won't be able to sell systems though, and no one involved will make a profit.
The state of X86 has nothing to do with AMD failing to provide competition to intel. It is a matter of having already plucked all the low hanging fruit. It is diminishing returns.
Consumers haven't noticed much for a while now, but while there was still competition, Intel did push out the new chipsets faster.
Right now, they're doing what you expect them to do in the absence of serious competition ... proceed at a very leisurely pace, selling small increments at full cost, many times over. No more big steps.
Until someone releases something in a lucrative key market, then I bet they'll do a big step again to one-up them. That's what drove progress before.
> Intel isn't sitting atop a vault of radically faster X86 core architecture just keeping it locked away because they can.
They are perhaps sitting on two generations of miniaturization and better tech, yes, and also a decent potential for adding moar cores.
They're not realizing it now 'cause no competition.
Zen certainly isn't crap.
Per core. Zen cores are quite small compared to Skylake. That means more space for faster cache, and more cores: Intel top out at 18.
Also, 14nm FinFET from Samsung can actually clock higher, with less leakage, than Intel's equivalent process. Meanwhile Intel haven't been able to get 10nm EUV yield up to any kind of reasonable anything, which has given AMD an opportunity to catch up - perhaps not pull ahead significantly, but make Intel sweat.
Yes, I said Samsung. The validation samples I have seen so far are actually fabbed on Samsung's 14nm process, not GlobalFoundries or TSMC. That includes Polaris (which is running silicon). I don't know why, if they're just experimenting with multiple-sourcing, or if it has something to do with Keller moving to Samsung.
Far more importantly, AMD being somewhere in the market and actually competitive means Intel don't own a complete monopoly on the x86 arch, which will hopefully keep down their prices. Everyone wins.
Whether they can execute is a different matter. It'll be 2017 before we get mainstream Zen parts on their current roadmap: AM4 boards will be out in March but they're still making Excavator APUs on GF28A for them until mid-2016. That would be a very disappointing purchase compared to a Zen/
>Meanwhile Intel haven't been able to get 10nm EUV yield up to any kind of reasonable anything, which has given AMD an opportunity to catch up - perhaps not pull ahead significantly, but make Intel sweat.
What are the odds that they just haven't "succeeded" yet to avoid this being a point in anti monopoly lawsuits, but quickly will succeed once a competitor threatens them...?
They likely don't have anything past Kaby lake fully simulated and debugged yet. There is nothing to be done to conventional IC to massively increase IPC and maintain clockspeed. This has been a problem for over a decade so fundamental that its a cornerstone of solid state EE curriculum. We can't feed instructions to the ALUs fast enough to keep them fully utilized, even if you have a front end that never stalled or was hit by branch penalties. A 3 ALU wide X86 core is utilizing that 3rd ALU so little that you can add the load/stores and registers to implement SMT and give yourself a solid 20-30% multicore performance uplift. This happens because the execution resources simply cannot be utilized by one thread. Intel has 4 ALU integer cores, with massive FPUs, and they're consistently maintaining high clock speeds to boot. That is the absolute pinnacle of a high performance core regardless of ISA, and the 4th ALU is being utilized even less than the 3rd. The only way they can justify this in silicon is by adding more fixed function bits to make sure it doesn't end up as wasted dark silicon, though this likely was a consideration for thermal management.
Intel isn't holding anything at all back. If you want an 8 core mainstream desktop chip you're going to pay for it by way of significantly higher TDP, or lower clocks. Thats all there is to it. Everyone across the industry is working their ass off to squeeze a fraction of a percent from this over tapped well, and that extends to MIPS, ARM, and POWER as well. ARM IPC will reach the same plateau, and with every passing generation vendors will attempt to push clocks higher to gain performance since raising serial integer performance is far too costly and nigh impossible by conventional means.
I think they are holding back, but I'm not saying it will be a "massive" increase (just a generational one, as small as those generation may be now). Might even also just be a more power hungry chipset on a huge-ass socket that dissipates more heat...? IDK.
Just whatever they need to win in the benchmarks and tests in those lucrative sectors of the business they basically monopolize. AMD will probably never get in there again.