Zen Thread
Images are sometimes not shown due to bandwidth/network limitations. Refreshing the page usually helps.
You are currently reading a thread in /g/ - Technology
You are currently reading a thread in /g/ - Technology
Zen Thread
>Zen is supposed to have around 40% better IPC over Excavator, which was probably taped out sometime in 2011 or 2012
>Zen is refocusing its attention on SMT and single-core performance - although Intel only has one extra thread per core with HyperThreading, other SMT chips have had up to EIGHT
>Zen is going to be on a 14nm node
>Zen will be refocused on single-core/single-threaded performance, but retain the high amount of cores seen in Bulldozer, which attempted a shift to "high compute over cores" but most software still doesnt use "all" of them
>Zen will be a unified socket, meaning there is no choice but for AMD Mini ITX boards to support Zen
>Zen will be both CPUs and APUs, and possibly even strictly APUs for the non-server market if they can manage 8 cores and a whopping GPU
>DirectX 12 and Vulkan will be in full swing by then, which can support iGPU even with a dedicated GPU in the system (supposedly)
>HSA also continues to evolve, continuing to marry the CPU and GPU together... offering a lot of potential to future simultaneous GPU-using productivity/OpenCL stuff as well as the aforementioned simultaneous GPU usage for gaming
>Zen might actually be expensive in the high end, over $400 or $500
Did Keller the Stellar save AMD's CPU line?
Will Intel pay companies not to use Zen if it's good enough to outperform every chip they have again like some past AMD architectures?
Latest new on Zen: GloFo is actually hitting decent 14nm yields AND the engineering samples are doing as well as expected. No hiccups. Rumor has it, Apple saw some early engineering samples and have made plans to use Zen exclusively in one lineup of ther iMacs.
>https://finance.yahoo.com/news/amd-zen-cpus-said-meet-185517672.html
>http://www.techspot.com/news/62481-amd-rumored-building-custom-soc-next-imac.html
>http://www.extremetech.com/computing/217664-globalfoundries-announces-14nm-validation-with-amd-silicon
So basically Zen will compare to an i7 920. Nice.
>>51251617
>8 core
>haswell-tier IPC
>14 nm
>SMT
>comparable to fucking Nehalem
>>51251648
You're right, it's not comparable to Nehalem. Maybe we'll get lucky and it will finally outperform Phenom II's though.
>>51251603
>Zen is supposed to have around 40% better IPC over Excavator, which was probably taped out sometime in 2011 or 2012
The Excavator core arch was one of the first things Keller worked on when he rejoined AMD, and the Carrizo APU didn't tape out until 2014.
Core designs do not "tape out." Taping something out means you send it to the fab to get ready for production runs. Complete products get produced, individual pieces of IP do not.
>although Intel only has one extra thread per core with HyperThreading, other SMT chips have had up to EIGHT
This is completely pointless to mention. You do not want an arch that wastes resources and many logical threads. Each thread uses execution and load/store resources while providing successively less and less performance.
Your first logical thread does not provide near the same performance as one of your physical threads. Your second logical thread provides less performance than the first, and so on.
You wouldn't want AMD to release an arch that was host to two logical threads, and Zen isn't going to. Patch notches show its 4 ALUs and 2 load/stores.
>>51251648
Lol, it's retroactive post-purchase-rationalization
Just wow
>>51251787
+;; Decoders unit has 4 decoders and all of them can decode fast path
+;; and vector type instructions.
+;; Integer unit 4 ALU pipes.
+;; 2 AGU pipes.
+;; Floating point unit 4 FP pipes.
+ 32, /* size of l1 cache. */
+ 512, /* size of l2 cache. */
should I wait for zen with my amd 8320? or buy a skylake for christmas?
>>51251603
>Zen might actually be expensive in the high end, over $400 or $500
I hope not. AMD needs to launch Zen so that they're beating Intel at every price point.
3 core Zen vs dual core Pentiums sub $100
4 core Zen vs Intel i3s and low-end i5s $100 - $175
6 core Zen vs k series i5s and low-end i7s $175 - $250
8 core Zen vs higher end i7s. $250-$300
16 core Zen should be the only thing above $300
>>51253186
6 core Zen will be able to match higher end i7's m8
>>51253480
That's the idea. Smash Intel at every price point.
If we get a 16 core Zen, I bet we get a 12 core version as well. Going on the general assumption that they'll fab 4 core, 8 core, and 16 core and then sell cut down versions to deal with problems.
Mark my words: Zen will suck ass when it comes out. AMD kucks will be on suicide watch world wide.
>mfw AMD will finally be finished
i hope AMD shills' tears taste nice
>>51253624
>Keller not delivering
Ayy lmao, cry more Intel shill
The only thing I fear it's that Zen could become the new K8, and AMD will be unable to come with something better in the future
>>51253186
The only 16 core Zen based chip is an Opteron.
The top end of Summit Ridge is an 8 core part.
>>51253689
Is Opteron going to be on a different socket?
>>51253661
Keller ran away m8
>>51253710
Of course.
The package is much larger, has a massively higher pin count, and it is designed to handle the higher power draw.
Consumer hardware is nothing like enterprise hardware.
>>51253723
RIP
Anyone have a clue as to the release date of Zen? I'm really fucking hyped for it but am getting a bit impatient truth be told. Encoding takes forever on an i5 and I was looking at the x99 platform. But I don't want to upgrade prematurely. Really hoping this is everything AMD says it is.
>>51253773
I heard Q1 2016, but I may be wrong. I doubt they could do it much later than that though.
>>51253723
He finished his job, that's how he works
He goes in, finishes a great design and leaves, the same happened at Apple, before that at AMD, and before that at DEC
He was 5 years there too, he did a great deal of work there, specially when you take into account how long he was at Apple and the monsters that came out from his time there
>>51253812
I've heard Q3 2016, but maybe that's general availability and we'll see a 'halo' chip released before then in limited quantities to prime the market.
>>51253773
>Anyone have a clue as to the release date of Zen?
Late 2016. Q3 if you're lucky. Some leaked slides showed their schedule for first silicon, ES, qualification, and in channel availability. I think it listed October 2016 specifically.
>>51253812
>>51253856
Ok so maybe a year from now. That's a bummer honestly. Ugh. Encoding makes me miss my fx9370 that i had. Sucks that the single core was so terrible.
If AMD fucks up Zen, it is pretty much RIP time.
Zen won't be able to compete with Intel's most recent CPUs, but should atleast be at Haswell level, regarding both performance AND power efficiency.
If not, then RIP
>>51254060
And they know it otherwise they wouldn't have brought Keller certified shit wrecker in on the development. AMD was losing or had already completely lost ground in the mobile, consumer, and and server markets as it is. This is their hail Mary and they know it.
Discounting the differences in z97 vs z170, Haswell competes within 5% of skylake so I doubt anyone would bitch if Zen can meet that level of performance.
>>51254060
If AMD fucks up Zen and goes under, then its pretty much RIP time for x86. It won't die overnight, but the transition of the industry to ARM hardware will be accelerated.
All signs are good though this time. There's none of the concerns about poor support for an unconventional architecture.
>>51254217
"The industry" is not transitioning to ARM. Give it up.
No, Zen hasn't saved anything. If anything you should expect it to fail with the decisions AMD has taken over the last few years even with the wrecker of shit. I want Zen to deliver. I'm not even asking for them to be toe on toe with Haswell, just get as close as possible to them. If they can deliver an 8 core that beats my 3570k in single threaded programs, then I'm all for it and will probably upgrade even if Intel releases a 7930k with 8 cores by 2017.
I'm sure they want to put out a chip that has beefy zen cores and good graphics, 2016-2017 will be interesting. Either they fail or get a nice tailwind.
>>51254258
The PC industry won't accept a total Intel monopoly though, and players like Nvidia are already pushing towards increasingly capable ARM chips.
>>51254217
As if Intel appallingly just five up on x86 overnight and start making ARM chips. With AMD out of the picture, they'd be under fire by government because monopoly. They'd have to subsidize and somehow get themselves out of the monopoly mess. Then they could jack the prices up on their hardware to insane prices. Then when they've officially milked the market dry, then maybe they'll start making a move in ARM.
Even if you don't like AMD, them dying will NOT help the market. You think Intel has no reason to actually try and innovate now? Just wait until there is OFFICIALLY no competition at all.
>>51254350
Nvidia doesn't even have a single phone manufacturer using their chip. The only thing using a Nvidia arm chip is their shield line. The only manufacturer I could see giving Intel a run for their money in the arm market is Qualcomm. Even them though with their housefire 810 is looking too hot.
>>51254369
I don't think AMD dying would help the market at all. Intel monopoly wouldn't be good for anyone. It would turn a 10 year somewhat uneventful transition away from x86, into a 5 year painful transition.
Its not just stuff like them jacking up prices that worries me. Its stuff like not including PCIE slots on motherboards of certain sockets, or only selling chips surface mounted.
>>51254350
The PC industry doesn't have a choice. A mass transition to ARM on PC would be a logistical nightmare.
Nvidia is pushing for ARM chips because GPGPU can theoretically displace CPUs in HPC workloads. Nvidia wants to advance that trend because they're a high performance GPU company with no high performance CPU offerings.
>>51254453
>>51254369
AMD dying would be the best thing to ever happen in this field, once AMD goes under Intel would be forced to split and we would have proper competition, not to mention Intel would be able to expand his research to new technologies and not just CPUs.
>>51254453
>It would turn a 10 year somewhat uneventful transition away from x86
christ the shit you people pull out of your asses
>once AMD goes under Intel would be forced to split
not going to happen, precisely because people like mr. asspull up there think that x86 is going to be dead in 10 years.
>>51254487
>A mass transition to ARM on PC would be a logistical nightmare.
You're right about that. Shifting to ARM would require major new developments in the ARM hardware ecosystem, things like a common firmware platform for systems, and the loss of a lot of legacy software.
>The PC industry doesn't have a choice.
You're also right about that, though not in the way you mean. If Intel becomes the sole provider of x86 hardware (Via doesn't count) PC makers will have little choice but to start exploring other architectures.
Apple and Google are clearly already leaning in that direction. I expect Valve has the same thought as part of their long term strategy for SteamOS.
>>51254447
But imagine an arm desktop processor.
High end mobile chips are trading blows with pentiums. Imagine what arm could do if they made a 30w chip instead of a 3w one.
>>51254447
Samsung isn't doing too bad in the ARM market either. Sure, big.LITTLE is stupid, but they perform okay.
>>51254543
>x86 is going to be dead in 10 years.
Dead? No, Intel and Microsoft are basically stuck with x86 no matter what happens.
Will it be as relevant to average users as it is today? No.
>>51254499
>AMD dying would be the best thing to ever happen in this field
I can't believe there are people as stupid as you.
>Intel would be forced to split
That is never going to happen. If Intel were to be forced to split they would just pull out of the US entirely. Yeah, they would take a hit by doing so, but not as much as they would by splitting. And if the US still wants viable electronics they will still buy from Intel regardless.
>>51254635
>big.LITTLE
I fucking hate this shit.
It's MOAR CORES, ARM marketing version.
>>51254649
Microsoft is trying to transition to ARM. They've got Windows 10 builds that run on ARM.
>>51254597
>PC makers will have little choice but to start exploring other architectures.
Why is this exactly? Intel has always been the dominant supplier of x86 hardware. Apple has gotten along just fine the last ten years exclusively using intel chips for macs. The reason that intel had to worry about antitrust in the past is that x86 was so dominant, now that's not the case.
>>51254688
>what is stagnation
>That is never going to happen.
kek are you new to the world or something?
legally they can't force Intel to split dumbass, but it is on their well interest in this case Intel's to split to avoid any monopoly litigation not to mention that as history has proven with other companies it would actually benefit them in the long run.
>>51254649
It isn't as relevant to average users as it was 10 years ago. So what?
>>51254714
Microsoft had Windows NT builds that ran on PowerPC. Back in the early 90s, people thought that x86 was going to be dead within 10 years and RISC chips were going to become dominant.
Didn't happen. Why should it happen this time?
>>51254350
Yeah they will, they did it for decades.
It would be interesting to see second-source manufacturing of CPUs to pop up in x86 again, though. I doubt OEMs would push for it however.
>>51254795
They don't have to worry about monopoly litigation because they aren't in a position to be a monopoly anymore.
Having total control of a CPU architecture isn't the same as having a monopoly on the CPU market.
>>51254810
Why is everyone forgetting Apple and iOS? Apple is the one who's gonna make the jump to ARM. They want to be in full control, so naturally they want to make their own chips, which they do for phone and tablet market.
That's why they are introducing iPad Pro, they are steadily moving the market into ARM direction. They don't want to depend on Intel, and they see desktop as a dead market. And they have far more control over iOS and it is more profitable for them.
"Apps" are getting more powerful by the day, so there's no need for traditional "legacy" software used on PC. And corporate market is moving towards iOS. Most profit right now is in corporate software development for iOS if you're a developer.
Apple will be the death of x86. Not AMD dying, not Windows moving to ARM.
>>51254723
>Intel has always been the dominant supplier of x86 hardware.
But never the only supplier. Even if companies chose to stick with Intel, they were always able to use AMD as a threat to keep Intel from pulling too much shit. If Intel becomes a monopoly, they'll be able to use ARM in that same way, but only if they are able to make ARM into a viable threat.
>>51254810
If the bottom is cut out of the x86 market because the mainstream is on mobile ARM hardware the cost of Intel's higher end parts for desktop will go up.
>>51254814
>Yeah they will, they did it for decades.
Except that they didn't through the 80s and 90s there were a number of attempts to build PC platforms around other architectures by companies like Commodore, Be, and Apple.
>>51254924
PC OEMs didn't start considering non-Intel chips until Intel started designing their own chipsets, which pissed off Compaq in particular. Even despite that, AMD and Cyrix remained minorities outside of the low-end and whitebox markets.
>there were a number of attempts to build PC platforms around other architectures by companies like Commodore, Be, and Apple.
There were not. Commodore themselves made a rather prolific range of standard XT/AT clone systems, Apple and Be systems were never "attempts" to migrate users away from an Intel monopoly, they were merely alternate platforms with different strengths and weaknesses.
>>51252001
yes, he bought it before it was even developed.
>>51253624
>i hope AMD shills' tears taste nice
until they go under and intel is cut into more pieces than ma bell
>>51254917
>That's why they are introducing iPad Pro
This has to be one of the easiest way to tell a paid shill apart from a standard retard, nobody drops "the" from a sentence like this but Apple marketing.
>>51252001
Sure, he is doing post-purchase rationalisation even before the product is released!
You're fucking retarded, holy shit.
>>51254217
>but the transition of the industry to ARM hardware will be accelerated
The moment when the incredibly desktop-inefficient and hella proprietary ARM architecture is a viable choice for PC's is also the moment where humanity has progressed to where CPU's are probably our smallest concern and/or Citra is capable of emulation on an Athlon XP
>>51254217
>ARM
>Actually good for desktop
Kill yourself.
>>51254597
>You're right about that. Shifting to ARM would require major new developments in the ARM hardware ecosystem, things like a common firmware platform for systems, and the loss of a lot of legacy software.
and IPCs that aren't years behind x86
>>51254917
Apple is still only a fraction of the x86 market. Losing them would hurt, but it won't kill x86.
That's assuming they want to leave x86, which I don't think is the case. You say they see the desktop as a dead end, but they still introduced a new mac pro 2 years ago. The mac pro surely doesn't make them all that much money, but Apple still considers it important to have a top performance product for the sake of their image.
>>51254924
>But never the only supplier. Even if companies chose to stick with Intel, they were always able to use AMD as a threat to keep Intel from pulling too much shit. If Intel becomes a monopoly, they'll be able to use ARM in that same way, but only if they are able to make ARM into a viable threat.
ARM is already a threat to intel. It's more of a threat than AMD at this point. Also, the OEMs are useless and can't do shit to preempt the supposed dangers of an intel "monopoly".
>>51254924
>If the bottom is cut out of the x86 market because the mainstream is on mobile ARM hardware the cost of Intel's higher end parts for desktop will go up.
The mainstream is on mobile ARM hardware, and the cost of intel's higher end parts for desktop have gone up.
Not all that substantially of course, because intel has always taken a big margin on their part sales, and they can afford to take less.
Fact is, wintel will persist for the foreseeable future, in a diminished, but stable capacity. Intel is doing a better job producing low-power chips than ARM manufacturers are producing high-performance chips.
>>51253186
>>51253480
zen cores can only be disabled in groups of 4, we're going to get:
>4 cores with disabled gpu
>4 cores with 4 cores disabled
>4 cores with gpu
>8 cores
server cpus will probably use a larger die and socket with 16, 24 and 32 core skus
>>51255397
>It's more of a threat than AMD at this point.
Yeah, but AMD is responding to said threat with ARM K12.
There is absolutely no reason that ARM would be less effective on a desktop than an x86 processor, except that almost the entire ARM market mainly operates in places where performance is less important than power efficiency.
Come on, RISC architectures can seriously kick ass. Look at SPARC in its time, it fucking destroyed x86 left and right for years.
It hasn't happened for a while ever since Intel established the x86 desktop monopoly (because IBM compatibles and windows did), especially after Intel moved x86 instructions over to a decoder and built the CPU cores internally as an unstable RISC architecture (stability in this case referring to consistency of instruction set). RISC architectures definitely carries the potential to seriously slap x86's shit provided it is not primarily focussed on low-power embedded platforms.
http://blog.imgtec.com/mips-processors/loongson-mips64-processors-performance-barrier
I mean, just look at MIPS64.
>>51255539
What makes you say this?
>>51255592
>I mean, just look at MIPS64.
the architecture only chinacrap uses?
>>51255613
You mean, because China isn't fucking drowning in x86 and Windows? Yes.
>>51255608
>>51255684
>Multiple units can be combined for even greater performance
What?
Anyway, this doesn't prove anything you said. A shared L3 cache and being part of the same "unit" doesn't mean some clever microcode can't disable a specific core.
>>51253822
To follow up on this anon's post,
Three years is plenty of time.
It's plenty of time to
>design new arch
>design a significant update to said arch
>submit extensive plans for successor
Among other things like new IP and cutting edge ideas.
Keller's work on the famed Athlon64 was finished in 1999
The joint project for x86-64 ISA was finished in 2000
The plans for a dual-core Athlon 64 x2 were in the pipe that same year (the patents actually were filed in Q4 1999, the dual core athlons weren't out until 2005)
Keller left the company in 1999, and was with AMD for hardly more than a year.
From that single year we got hypertransport, 64-bit CPUs, the modern approach of very wide, moderately piped, high IPC processors. All things he had a hand in.
Again, three years is plenty of time.
>>51255723
they only build interconnects between each unit, so only an entire can be fused off
not sure about disabling a single core 'softly' via microcode but i'm pretty sure AMD stopped doing that for a reason
>>51255723
>A shared L3 cache and being part of the same "unit" doesn't mean some clever microcode can't disable a specific core.
true but ever since bulldozer started theyve just been doing it by torching it with a laser; ostensibly to get away from the "clever code" resulting in unlockable cores in days past
>>51255127
>>i hope AMD shills' tears taste nice
>until they go under and intel is cut into more pieces than ma bell
this hasn't happened to the new telecoms like comcast or centurylink, why would it happen to something that isn't location dependent like cpu manufacturing? you can always buy ARM, can't you? hue, what shitty bait.
>>51255210
no hes justifying the purchase of an intel product, not an amd. lrn2readbetween the lines.
>>51255755
>i'm pretty sure AMD stopped doing that for a reason
people say it was for market segmentation but honestly i could have seen it being just about avoiding negative publicity from unlocked CPUs "not working" when modified out of spec
>>51255767
>this hasn't happened to the new telecoms like comcast or centurylink
because there are hundreds of telecoms.
in my area alone there's centurylink, comcast, integra, TDS, level3 and sprint.
meanwhile those mass market ARM desktops are going so well hue
>>51254615
low power chips don't scale well to higher tdp
>>51255613
>>51255646
By the way, alternative architectures aren't bad, especially not when they have hardware-assisted emulation of others.
>>51255842
they also aren't anywhere near mainstream adoption in the western world.
x86 is really fucking long in the tooth but i don't see arm supplanting it
>>51251603
>over a year until release
>no new details emerging
JEsus give it a rest your 404ing substantive threads.
>>51255859
Neither do I, but I hope that x86 is replaced (with instruction emulation in hardware perhaps), especially considering the recent discoveries made about the Intel Management Engine.
>>51255817
in my area (a captial of a state and major metro area) there's only two (which can't compete on speed but rather only value for low usage cases) and they have monopolies on their lines, but there's no rush to break them up at all. if there was no anti monopoly pressure to speak of, why aren't there more here?
>>51255923
the only other arch with good x86 emulation is a shady as fuck one designed by the russian government though, it needs to be an open design by the arm committee or someone similar
>>51255952
>and they have monopolies on their lines, but there's no rush to break them up at all.
Because you're only looking at hardline LECs.
>>51255959
But MIPS64 is open and royalty free.
>>51255965
>implying that wireless can compete with wired on a reliability plus performance standpoint.
>>51256004
>reliability
then get a DS3, i guarantee you that you can
I want to believe, but I'm skeptical that the first wave of Zen chips priced at $500 can beat an i7-6700. We need it to happen though, so Intel can stop fucking us in the ass.
>>51256057
>I'm skeptical that the first wave of Zen chips priced at $500 can beat an i7-6700
they probably won't. the hope is that the nigger-build ones can be overclocked to shit and beat intel on price/performance.
>>51256057
zen is a failure if they aren't heavily undercut like current bulldozer cpu prices imo
not gonna pay i7 prices for an amd i7 competitor with lesser ipc
>>51256019
>DS3
only available from a single provider, but that's not a monopoly because of all the other inferior "options" available.
>>51251603
amd aint doing shit on 2d planar transistors
fucking retards
I just want an 8 core that can beat Sandy Bridge and isn't more expensive than a i7-5820k.
muh emulation
muh encoding
muh VMs
>>51256057
New chip, new expectations.
If they are telling the truth, zen will be much more powerful per core than anything intel has to offer.
If they are lying, the cpu division of amd is finished.
I'll probably upgrade to Zen just so I can avoid giving Intel money.
My 2500k has treated me well, but I'm against their Jewish tactics.
>>51256057
>>51256320
I doubt they'd hype 40% higher IPC only to be discovered to be lying.
Seriously, if it really is 40% (or more) IPC gain then these processors are going to overclock like mad, due to the also reduced TDP.
>>51256347
>I doubt they'd hype 40% higher IPC only to be discovered to be lying.
didn't we hear only slightly lower hype for vishera? i'd pretty much tuned out at that point
>>51256367
IIRC nothing like 40% IPC gain was mentioned for Vishera, only that it would perform very well.
Which it did. Vishera processors seriously deliver some performance when using 8 threads or more.
>>51256320
40% over Excavator is somewhere between Ivy and Broadwell, depending on who you ask.
It's the Zen+ design that will truly rival Intel's current (as in today, not on Zen+ release) stuff.
>>51256347
If the design is anything like other high performance modern x86, it will clock juts like Intel's chips. Barring ridiculousness like shitty thermal paste (Hello Ivy)
Expect top chips to crank out >5Ghz, expect the average chip to play with 4.6-4.8
Although we don't really know how a high performance, large die 16nm design behaves.
>>51256409
>Expect top chips to crank out >5Ghz,
more like overclock to 5ghz, spec at 3.8ghz
>>51256614
>>51256347
>overclock like mad
The eight core Zen will be quite similar to the 5960X except for very heavy FP crunching. Cap this if you want.
One thing to note is that (I was wrong about the feature size) the 14nm FinFet from Glofo has BETTER characteristics than Intel's
We all should keep in mind that Intel's advertising of feature size is measured differently than literally every other lithography company in the world.
>>51256659
>except for very heavy FP crunching.
Why do you say this? And who the fuck gives a fuck about FPUs anyway? Anyone who does serious FP crunching does so using OpenCL or something.
>>51256711
Info put together from leaked data showed that Zen's FPU is slightly less capable than Intel's Skylake. I can't recall the exact terms but it is something like Intel can handle a certain amount of multiplies and divides, and fused ops, while the Zen FPU can handle the same amount of multiplies/fused but less divides.
Don't quote me on that, take it as a generalization of the fact.
>Who gives a fuck about FPUs anyway?
Anyone spending $300 to do real work with their computer, faggot
Go back to /v/
>>51256825
>Anyone spending $300 to do real work with their computer, faggot
>Go back to /v/
No. I'm serious. You need some serious number crunching to reach the FPU's limits. In ANY professional context where FP performance is important they're likely using the GPU for that.
FP on CPUs is just for retards and poorfags nowadays.
>>51256825
If you're doing flops on a CPU, you're just a memer. Anyone who does real work on a computer uses a GPU for that.
>>51257037
This. Peak or average GFLOPS is quite simply a pointless number in measuring a CPU's capability for workloads that actually matter.
>>51253067
Skylark isn't going to give you significant performance gains, so just wait for zen and have enough to build a beast
>>51253773
Encoding wise, the high end Fxs (8350, 9350, 9370), which are much cheaper than an x99 setup, demolish everything else. For general use, I have found it to be more than enough. Have 8350
>>51257460
It sucks that they got rid of cmt.
>>51259246
They bet on it WAY too early.
If Bulldozer was made a decade later in 2021, it probably would have done much better.
>>51259266
I honestly would have preferred bulldozer with zen cores over zen.
Smt is shit compared to cmt.
Imagine, 4 core zen vs i3, 6 core zen vs i5, 8 core zen vs i7, and 12 core zen because fuck Intel.
And a really cheap dual core for budget desktops.
>>51257460
>Encoding wise, the high end Fxs (8350, 9350, 9370), which are much cheaper than an x99 setup, demolish everything else. For general use, I have found it to be more than enough. Have 8350
Why is bulldozer such garbage at 3d rendering? It's highly threaded like encoding but FX loses to 4 core i5s with no HT.
>>51259662
ok
>>51260059
Try it in any non synthetic. Modo, Blender, renderman all perform badly on my 8120 @ 4.2ghz.
Keller has hinted at Zen not necessarily matching Skylake in terms of raw performance, but quite possibly beating it in terms of performance per Watt.
>>51255539
This is not even remotely true, and Summit Ridge does not have an IGP.
>>51255684
These "leaked" slides are fake.
Stop spreading misinformation.
>>51252001
>retroactive post-purchase rationalization
dat redundancy tho
If you want ivy bridge performance in 2016 whynot just buy ivy bridge now?
Zen and AMD are still dead in the water unless they can sell to OEMs or have a mobile version.
Not that it's Intel's fault, they've cockblocked them in the tablet and laptop market for years.
>>51261118
>mobile version of an enthusiast desktop CPU
lol no
The primary target for the Zen arch is enterprise. The server market is what they project as their strongest growth market, desktops aren't.
There won't be Zen based APUs until 2017 that span desktop and mobile. Summit Ridge enthusiast CPUs do not have a mobile counterpart.
>>51254217
>transition of the industry to anything
Laptops were supposed to kill desktops years ago.
Now they're dying to Tablets.
The point is that you can't predict the market and how it will pan out, especially with tech because you never know what's right around the corner.
>Suddenly apple convinces everyone that desktops are trendy again
>100000% increase in desktop sales AMD is saved due to market growth and Intel buys Samsung.
>>51251603
>Intel only has one extra thread per core with HyperThreading, other SMT chips have had up to EIGHT
Yes, because additional SMT threads scale so poorly it becomes a waste of silicon at that point.
Without free execution units SMT does nothing. Both threads stalling at the same time is too rare to justify the increased complexity of adding more.
>>51251603
>>Zen will be refocused on single-core/single-threaded performance, but retain the high amount of cores seen in Bulldozer
Nice, Zen will have 4 cores / 8 threads like Bulldozer.
Every AMD thread is how the next new product is going to be the absolute best to ever come out and then always falls short
>>51262225
Don't know about before, but there's a lot happening now. New GPU tech, new CPU tech, an actual fantastic engineer has been working with AMD on zen for once, AMD is much better in DX12/vulkan, amdgpu (open-source graphics drivers which are meant to replace catalyst) is happening, etc. All things combined, there's a LOT to be hyped for. Even if everything only half-works, it's very possible that AMD becomes competitive again.
>tfw desktop uses an FX8320 and 2x7970 setup
Three biggest computing mistakes I've ever made to be honest.
>>51262956
Go back to /a/
>>51262980
Sorry, AMD is doomed. Hopefully ARM will close the gap and Windows will die off, then we'll be at a golden age of personal computing.
>>51262987
No that's all fine, but get better images to post with. All anime is shit
>>51262956
at least you don't need to heat your room in winter.
>>51263006
I don't watch anime
>>51251603
>possibly even strictly APUs for the non-server market if they can manage 8 cores and a whopping GPU
>Zen might actually be expensive in the high end, over $400 or $500
Sounds like it will be pretty shitty... Intel monopoly coming up
>>51263013
I really don't. It's incredible how much heat it puts out.
Can Zen even save AMD at this point, even if it's really good? Last I checked they were hemorrhaging cash every quarter.
>>51263195
Yeah it could,, but that's not gonna happen if it's expensive and APU-only.
>>51262956
>tfw workloads that use all eight cores concurrently on an FX-8350
Linux is currently a LOT faster than Windows on Bulldozer because the OS is made of a bunch of smaller, independent processes that can shuffle between cores cleanly. I can run a huge browser session, a multithreaded compiler, and a couple of VMs comfortably at the same time, and the system doesn't bog down at all, or I can use Handbrake with eight threads.
>>51251603
>>51251603
>>Zen will be a unified socket, meaning there is no choice but for AMD Mini ITX boards to support Zen
This is the most retarded comment you could make. Look at Supermicro's EATX sized dual Xeon boards, they dont all support 150W chips, and these are $500+ boards.
>>Zen might actually be expensive in the high end, over $400 or $500
That isnt expensive for something "high end"
>>51263221
Even if it's good AND relatively cost effective, can't intel just continue to undercut them at every turn? Even if intel takes a loss or just breaks even on their hardware, isn't that enough to kill AMD?
>>51263238
My desktop is strictly for gaming. I use it maybe a few times a month.
>>51263262
Intel never undercut. They make same-performance hardware for 3x the cost.
>goyim still fall for it
>>51263262
>can't intel just continue to undercut them at every turn
They probably could, but I don't think they will. They want to make money short term, they will probably start innovating a bit but still staying the "higher quality, higher price, you get what you pay for" product, they just make more money like that. They probably don't want a full monopoly since that will probably result in gov. intervention and actually competent ARM companies moving into desktop. Not that it wouldn't fuck us over as customers...
But, if intel did undercut an already great CPU, we could stock up on some great cheap CPUs, to last us though the monopoly.
>>51252001
Fuck me, you are pretty goddamn stupid.
Will I need to buy a new mobo ? I have a AM3+ socket and it would suck
Will TDP be through the roof like Vishera ?
>>51263238
Agreed
>fx-8350 OC to 4.7ghz
>has trading station open on 4 screens
>firefox/Chrome combo
>using handbrake
>video editing
>not usually open at the same time but has been before
>playing games sometimes while doing all these processes
no issue handing all the processes, the fx-8350 was a good buy for me.
>>51263267
thats cool man.
>people shit posting about the fx-8350 are all from /v/
>>51263716
Most likely you're going to have to, the AM3+ is a dying socket.
>>51263799
most retards that defends jewvidia and jewtel so hardalso comes from either leddit or /v/
>>51263799
You'd be able to do all that on an i5 though. People aren't saying bad things about AMD because they're fanboys. AMD used to be the best choice, but they've stagnated and are in a very bad financial situation (Intel's R&D budget is greater than AMD's entire revenue) and it shows in both benchmarks and real world tests.
I'd love Zen to be AMD's return to greatness, but we need to face reality at some point.
>>51263810
;_;
I will have to buy 300 bucks then.
>>51263838
Yeah seems like, I used to enjoy going to build a pc on reddit. always got downvoted because it wasn't an i5-i7 build and was a reasonable price.
>>51263865
the thing is, I could do that on an i5 but at the time the 8350 was 40% cheaper at the comparable specs of the i5 without an OC from the 8350, It was just the smarter buy. there's not a doubt that AMD has been suffering since 09, look at their stock and has been bought by a few companies during that time, maybe ZEN could be the white rabbit they've been looking for.
>>51263970
>buying money
WOW
>>51264002
you can buy money on the FOREX market.
>>51263401
That's what they are doing now because AMD's products are currently shit compared to theirs in the CPU market. People want performance and they're willing to pay for it. What I'm saying is even if AMD delivers a great CPU, intel, being the larger company by a huge margin, will have the leverage to sustain losses for much longer if they wanted to.
tl;dr - If they wanted to, intel can bleed AMD even if AMD makes something pretty damn good.
>>51263865
>You'd be able to do all that on an i5 though.
No, for eight threads you need a hyperthreaded quad i7 or Xeon E5, which is *significantly* more expensive. I got my FX-8350 for like $120 on sale.
>>51264030
>Believing that money actually exists just like the gold on COMEX
>Not that you're just buying bits in a database somewhere
I expected more of you /g/
>>51264063
>Being proud of paying $120 for a CPU
>>51263013
>at least you don't need to heat your room in winter.
People joke about that but I legitimately forgot to fix the piping in my room for a year and a half straight simply because of my PC.
>>51263799
>>people shit posting about the fx-8350 are all from /v/
Shouldn't surprise you. Not only is their very board founded on shitposting their hobby is largely founded on poser hardware and ignorance.
Intel supports Feminist Frequency.
just saying.
>>51264063
The i5 would still be able to do it since it's so much faster per core.
>>51264077
>using a server grade CPU
>2 sockets
>comparing commercial grade products to consumer grade
>comparing apples to oranges
>/v/ shit posting
Also, you just spouting about gold is telling me you no knowing about investments.
>/pol/ shit posting
>>Being proud of paying $120 for a CPU
>he bought something that works for him that at a price thats reasonable = Poor
/v/ shit posting again.
heres a great book: Stop Acting Rich: And Start Living Like a Real Millionaire.
maybe you'll learn something about savings.
>>51264082
>>Shouldn't surprise you. Not only is their very board founded on shit posting their hobby is largely founded on poser hardware and ignorance.
gotta buy my Gamer Grade products! They can enjoy there 60% markup.
>>51264186
>Seriously guise
>All the gold traded in COMEX is really in a warehouse
>tfw AMD delayed release of ARM Opterons
>8-core zen + Fury X2
>implying this isn't the 4k steambox/HTPC of the future
>>51254206
>Keller certified shit wrecker
>>51261981
Bulldozer 1 thread/core dipshit
Just watch this flop as hard as Fury did.
What about power usage? This isnt going to be another 4000w amd product is it?
I am tying my destiny with AMD.
I will invest all my small little savings into AMD stocks.
If they crash, I am pennyless.
>>51265644
14nm Gloflo, superior to Intel, it'll be better than Skylake in that regard.
>>51265644
nope. New node, more efficient architecture, and then there's all of the power saving and efficiency stuff they cooked up for Excavator chips
>>51265644
28nm bulk is a substantial upgrade over GloFo's 32nm PD-SOI
14nm FinFET is an utterly tremendous upgrade over 28nm buk
>>51265671
Think i should bet my savings of $3000 on them?
>>51265671
>spending money you can't afford
10/10 investing anon.
>>51265847
Their stock is low enough that any uptick whatsoever will turn into a rally, and you could double your money easily. It wouldn't be out of the question for their stocks to hit $8 a share again which would be quadrupling in price. If they regain server market share investors will definitely take note.
Its a gamble. It is not a safe bet.
Only invest money you're comfortable losing.
Hopefully it's not a 500w TDP
>>51266131
AMD's days of power consumption issues are long, long over.
Bulldozer and Piledriver drew a lot of power because the architectural issues were increasing energy needed per op, and the high vcore and current leakage from bugged process meant they were just pissing away power.
28nm Kaveri has a 95w TDP, and sticks to it truthfully.
At 3.7ghz with a 4ghz turbo the A10 7850k will pull a little over 50watts for the CPU alone. Thats with stock vcore which in typical AMD fashion is slightly higher than it needs to be.
The absolute best of the best of 32nm Vishera chips are the high binned FX 8370e chips. Tomshardware overclocked one and got it to 4ghz at 1.17v. It worked out to 90watts measured at the rail.
Very impressive, but if a Kaveri APU were given the same treatment then power consumption per module would be lower still.
At 32nm a Bulldozer/Piledriver module minus the L2 is 18mm2~
At 28nm a Steamroller module is around 17.8mm~
Steamroller is a larger design, so much so that the area scaling going from 32nm to 28nm barely shrunk the module design at all. Still the power consumption per module improved significantly.
Excavator pushed on power consumption even further with new PowerTune IP and advanced voltage adaptive switching.
The Steamroller modules in Kaveri can hit 3.7ghz base clock with a nominal power consumption of 25 watts. Thats a middle of the row production target, not hand picked part running manually set voltages. Tom's Hardware figures validate it as well.
The Excavator modules in Carrizo, still 28nm and a cheaper generic bulk process at that, can hit 3.7ghz with a nominal 22.5w. The high end of their sweet spot on this process is 3.5ghz at just 16w.
The high energy usage per op will still be there due to the flawed nature of the Bulldozer family architecture, but AMD developed tons of power saving IP in the wake of Bulldozer. What they managed to squeeze out of 28nm SHP and 28nm HPP is astonishing.
>>51265671
Thinking of doing the same, albeit with only like $1000.
If it's a good bet, I'd have a really nice chunk to put into index funds. S&P 500 is great, but Nasdaq feels much more /g/...
What to do...
>>51266362
>comment too long
Moving down to 14nm FinFET that displays just a tiny fraction of the leakage current of 28nm HPP is a gigantic boon for them. All of their newly designed power saving IP, a brand new arch with significantly lower required power per op, and a massively more efficient process node will go to making sure Summit Ridge is an extremely power efficient line of CPUs.
I would be surprised if they even released a 125w rated part. Sweclockers leaked some time ago that the high end would only be 95w, and everything adds up to seem like thats accurate.
Carrizo's total die size is 250mm2. Thats two modules, L2, IGP with 8 CU, IO, media accelerators, FCH, TrustZone processor, and etc. If scaled down to 14nm it would be right around 75mm2.
The modules themselves are 14.48mm2, and scaled down to 14nm they'd be right around 4.3mm2.
Thats an interesting reference point considering what exists inside of the module. 4 ALUs, 4 AGUs, 2 128bit FMACs, MMX unit, dual decoders, 96kb of L1, and all the rest of the good stuff. Some patch notes released by AMD showed a general depiction of Zen's core width. It has 4 ALUs, 2 AGUs, and 4 FPU pipelines. No specifics about the FPU are known so we can't gauge how large they are, but completely doubling the FPU data paths in Excavators would only increase the module to around 5mm2 at 14nm.
I'd guess at Zen cores being right around 5mm to 6nm. With L2 they'd likely be around 8mm2.
As a broad stroke guess I'd put Summit Ridge's die size around 125-150mm2 accounting for L3 and uncore
>>51251603
>Will Intel pay companies not to use Zen if it's good enough to outperform every chip they have again like some past AMD architectures?
I think 3DXPoint exclusivity will be enough for a Zen DOA scenario
>>51251603
>possibly even strictly APUs for the non-server market if they can manage 8 cores and a whopping GPU
Doubt it honestly
I've heard somewhere they're going to offer stand alone cpus
>>51267613
3DXpoint exclusivity will make 3DXpoint DOA.
Nobody wants to use proprietary shit.
>>51267613
The AM4 platform has dedicated M.2 slots specifically for NVMe devices. Extremely high performance SSDs already exist, and continue to get better with every generation. The lifespan of 3D NAND structures makes longevity totally irrelevant, and this is doubly true when you increase drive capacity.
Xpoint memory really isn't anywhere as revolutionary as intel's own marketing says it is. They still have yet to prove they can actually sustain significantly higher IOPS in real workloads instead of hand picked niche scenarios that you will literally never encounter.
>>51267649
Summit Ridge is CPU only.
The APU platform for 2016 is Bristol Ridge, and that is a tweaked Carrizo for socket AM4.
The first Zen based APU is Raven Ridge which is launching in 2017.
>>51253186
>muh coarz
>>51264115
equivalent of paying the mob for "protection" or Al Sharpton's charities to avoid being labeled a racist organization
>>51267718
>non-volatile memory with bandwidth of DRAM, only 10x the latency of DRAM, density of NAND
>not using it
ok guy, you stay away because of muh proprietary
>>51267743
PCI-E won't cut it unless we start getting 80+ lanes in the next 2 years; one of the few things INTC has actually nailed down on 3DXPoint is that it would use current DIMM slots, though not the same protocol
>>51268019
DIMM slots are even less available on current motherboards than PCI-E lanes.
>>51268019
Xpoint is absolutely nowhere near fast enough to compete with, let alone replace DRAM. The dream of a unified storage medium that encompasses permanent storage and system cache is a very long way off.
SSDs are now reaching 2000mbps+ peak reads. 50GB full Blu-ray rips can be moved around in seconds. We're getting to a point where storage is finally outpacing file growth. I'm not worried about M.2 becoming irrelevant in the slightest.
>>51268057
>micro-atx fags need not apply
how many slots do you need when they're going to release 6TB sticks? Do some research instead of talking out of your ass
I wanted to wait for zed, but I couldn't stand having a shit PC anymore and ended up buying a 1231. Why is and waiting till almost 2017?
>>51268159
>>51251603
>Zen might actually be expensive in the high end, over $400 or $500
How the fuck is that going to save amd? They'll have comparable but worse performance compared to intel, and they're also expensive now.
Why?
That doesn't help at all
>>51268255
The high end of Summit Ridge, the 8 core parts, are intended to compete against intel's i7E line. They'd still be priced somewhat cheaper than what they're directly competing against.
Of course the binned chips will be much cheaper. They're not going to be selling a $300 quad Core part, it would be pricing their APUs out of the market since they're still going to be quad cores.
Their bread and butter with the new arch is going to be in servers, not consumer desktops.
>>51268200
>>micro-atx fags need not apply
Which is exactly what I was talking about. Why would I use a DIMM slot that I need for memory when I could use an available PCI-E slot and get the same real-world performance.
>>51268365
how about 1/40th of the latency?
>>51268233
>practically states xpoint is going to be server clusters/scientific only
So the only question is: how much does xpoint cost to manufacture and buy?
Will it ever actually be viable for consumers?
NVMe 3D NAND SSDs are already viable and fast enough for consumers.
They need cheaper storage, not faster, at this point.
>>51268499
Nobody would notice that.
>>51268499
The latency for these devices is already measured in microseconds. Unless this is also going to be significantly cheaper than current SSDs no one is going to buy it, especially not if it's proprietary bullshit.
>>51268499
Translate that to real world performance.
Moving things into and around permanent storage isn't a bottleneck in anything with a high end SSD now. The single exception would be really, really large movies, and with the latest crop of NVMe drives this isn't the case any more.
Xpoint is literally nowhere near competing with DRAM.
The decreased latency in operations for permanent storage doesn't matter in any real world scenario.
>>51268574
And heres the DRAM latency
>>51268574
>what are iops
come on man, use your brain
>>51253624
not an AMD fan but if AMD dies then INTEL will fuck you in the ass with shit CPU's there will be no competition therefore INTEL will do what ever it wants on the market
arm will win in the end because it just has much more ruthless competition.
Capitalism in action dictates that arm designs will competitively improve far faster than x86 designs in a two company market. We've already seen how far arm has come recently.
>>51268611
Again, intel has yet to prove that they can sustain significantly higher IOPS in a real world workload instead of some handpicked niche case that no one will ever encounter.
Samsung tried to do the same thing with claiming 300,000 IOPS, but is actually around 70,000 in most every bench. Though for what its worth some RAID SSDs have been recorded at well above 200,000 IOPS in real common tests.
intel is *claiming* 400,000 IOPS, and for all we know they're using the same bullshit niche synthetic measurement that Samsung used.
Xpoint absolute cannot compete with DRAM.
The increased, if at all, read/write speeds don't matter for permanent storage in the real world.
Latency doesn't matter for storage in the real world.
There is no reason to take intel's word for anything regarding IOPS.
The shit doesn't matter, its not revolutionary. Stop eating up and mindlessly regurgitating the utter feces directly from intel's marketing anus.
>>51268698
>Latency doesn't matter for storage in the real world
You are literally retarded.
>>51264186
>gotta buy my Gamer Grade products! They can enjoy there 60% markup.
Reminder that Alienware and Razer are still things.
>>51268872
>tech illiterate retard has no argument whatsoever
Waiting 1/10th of a second or 1/100000000000000000th of a second to move a file doesn't matter. It does not make a difference. Its entirely imperceptible. My home server doesn't even see enough traffic for this to matter if you added up all the time to complete operations over a year of constant use.
Your mindless shilling doesn't work here, child.
>>51268934
>b-but I wouldn't see any benefit so there's no need for it
>>51268972
the point is, since you're too thick in the head, that nobody would see a difference. The only people getting enough traffic would be those hosting enormous servers.
>>51268972
Thats right, you tech illiterate retard.
Storage operations are not intensive, more so they are not constant in anything but a data center's server rack. Even in that one usage case the latency reduction would not translate into real world performance.
>>51268544
>Will it ever actually be viable for consumers?
Isn't the answer to that yes by default? Also, man RAM speeds seem way underwhelming in terms of growth.
>>51268607
That chart doesn't make any sense to me.
Here's a Westmere with (dual rank) DRAM at 1421 and 1600Mhz.
Your picture suggests that it takes considerably better chips 30% faster DRAM clocks to achieve the same latency.
>>51269087
3000mhz - 3400mhz DDR4 is already becoming mainstream.
Some 4266mhz kits already exist, but being they're such a rare binning at the moment they're selling for $500. In a couple years the price will come crashing down.
Thats a substantial benefit over DDR3 at present.
>>51269092
The timings on your kit are much tighter than what was used in testing here.
http://www.corsair.com/en-us/blog/2014/september/ddr3_vs_ddr4_synthetic
The 1600mhz test kit in the bench is 10-10-10-30. If they would be set the timings manually to what the DIMM could actually function at then the latency would have been lower. They could have tightened everything up across the board if they really wanted to.
>>51269092
they might have tested with higher timings.
>>51269087
>Isn't the answer to that yes by default?
Not necessarily. If a tech is amazing for the server industry but unfeasible to port over to consumer-grade tech, then a company may not offer it up there.
When will be the day when you can get a mediocre APU capable of running new games at bearable framerates and settings for like, 4 years for about 300 bucks
You know like, getting a PC with literally a console attached to it
>>51269435
Literally impossible.
The best we can hope from APUs would be embedded HBM type memory alleviating many of the problems present today (which is apparently scheduled for 2017). I would expect 960/7870 levels of performance in a 95w package with 2-4GB of VRAM.
In any case, due to the lower cost nature of APUs I doubt there will ever be one that can handle all new games, all at "medium" settings, for four years after release.
>>51269435
That's never going to happen because PC games always target the high end. Just get a console.
>>51269553
Yeah, was thinking about the PS4. That way I don't really need Windows anymore.
Old leaks.
Summit Ridge starts shipping into Q3 2016. October is a solid bet.
Bristol Ridge Q2 2016
Stoney Ridge Q4 2016
Rave Ridge Q1 2017
So the delay between Zen based desktop CPUs and Zen based APUs won't be huge.
>>51270063
Could have sworn I attached the pic
>>51257399
>Skylark isn't going to give you significant performance gains, so just wait for zen and have enough to build a beast
Yes he'll have significant performance gains cause he's using a fucking 8320.
Damn a fucking pentium is faster on st performance than that shitty processor.
>>51269596
Modern consoles really soured me on keeping current with gaming thanks to rapacious DRM+DLC policies, half-baked releases, the need to manage patches and drive space like with a desktop PC. I have a PS2 and some emulators, and a Linux install of Steam (with an FX-8350 appropriately enough for this thread) that covers the rest of my gaming needs. If I ever get another console it'll probably be a steambox.
>>51270948
>Modern consoles really soured me on keeping current with gaming thanks to rapacious DRM+DLC policies, half-baked releases, the need to manage patches and drive space like with a desktop PC.
This isn't going to change, and PC gaming is becoming the exact same way. You get "season passes" that don't even include all the DLC being released for a game, you get locked content on disc, bonuses for pre ordering at shitty retailers, microtransactions, games that spy on you while you play them and report data back to the publisher for marketing analytics, platforms like Steam trying to force you to give out a phone number, the platform itself being a form of DRM that reserves the right to lock you out of your account and disable all of your games.
You don't own anything any more. This is what happens when normies with no standards blindly give money to massive corporations.
> If I ever get another console it'll probably be a steambox.
A run of the mill PC with Linux isn't a console.
>>51271178
My personal favorite is physical discs that force you to download and install the game through Steam when inserted.
Aside from GOG the days of just installing a game are over. Everyone is trying to lock you into something so they can make even more money off of you.
>>51271178
>A run of the mill PC with Linux isn't a console.
A tiny ARM box with Linux isn't a VCR either, but my Roku still performs the same basic function.
>>51251603
>>51251603
Will Zen have 8 real cores instead of the 4 module shit of the current FX line?
>>51271235
Zen is an entirely new architecture not related to the Bulldozer line.
It does not share the module core layout. Each core is independent, has data pathways as wide as an entire module, and supports one logical thread per core.
>>51254350
>The PC industry won't accept a total Intel monopoly though,
The PC industry will accept whatever Intel wants them to as long as they are paying the industry to do so.
Athlon64 outperformed Intel CPUs two to one on the same clockspeed, and yet Intel didn't get completely wrecked. EU later put a 1 billion € penalty on them for anti-competitive practices because of that. It was chump change as far as they are concerned.
>>51254487
>Nvidia is pushing for ARM chips because
Because they don't have an x86 license, therefore they don't have a product in the CPU market. So they are pushing for ARM, where they do have a product.
If Windows releases an ARM version of their OS and allows users to install it just like they do their x86/64 OS then ARM could become the future arch of the PC market.
>>51270178
>>51270063
FM3 is going to be deprecated. AM4, that Zen uses, is a unified socket. It wont be backwards compatible with FM3 because of DDR4.
>>51274147
Microsoft have proven themselves incapable of such a feat, see Windows RT. Even if they did manage it, the change would be gradual because most people are fine with what they have and ARM doesn't offer anything they want over x86 except maybe cost.
