http://www.itwire.com/it-industry-news/development/72853-amd-zen-will-favourably-compete-with-intel-skylake.html
>AMD focusing on VR, Gaming, and Datacentres
Looks like /v/fags will ruin this company for good.
Did you know that the CEO from AMD gets pay more than one from Nvidia despite having heavy loses.
Makes you wonder why doesn't it?
>>54590000
Lisa Su herself stated that AMD's new enterprise chips would be competitive in 80% of common server workloads.
AMD's Zen isn't going to be an FPU work horse. The term competitive in itself is ambiguous. Being 10% behind could be argued to still be competitive. Offering lower performance, but more tertiary features could also be said to be competitive.
Also keep in mind that Summit Ridge is competing against the i7 Extreme line of processors, not the higher clocked quad core mainstream i5s and i7s.
And stop posting those fake slides.
>>54590000
>Looks like /v/fags will ruin this company for good.
How so?
Reminder that Zen+ is the arch that will really change things up for AMD.
>>54590768
THANKS BASED KELLER
>>54590768
CERTIFIED
SHIT
WRECKER
>>54590138
Do you have a source for that? I'm curious to read about what is going through their heads if this is true.
>>54590999
January investor call:
Hans Mosesmann
Hey Lisa, a clarification on Zen you said you had some design wins already on the server side of that roadmap. Can you give us a sense on what kind of server wins and what kind of an opportunity you see in terms of market share for the next several years as you come back to the market?
Lisa Su
Yes, so Hans, overall the Zen design win, we have been engaged very early on with you know large OEM and cloud providers on the Zen design point and the platforms that would be useful for Zen. So we have closed our first design win, we are working you know very closely with these OEM partners to make sure that they bring up their platforms concurrently with our own design validation and testing. I think the main message is we are on track with the schedule that we previously discussed in terms of sampling this year. We will introduce first in desktop and so we are having conversations with some of the PC OEMs about getting their platforms ready for desktop and then we will go into enterprise server first full year in 2017.
Hans Mosesmann
Okay and then as a follow up, what kind of performance point should we look at your initial Zen in the server space, is it Zeon E3 class type products or E5s or maybe you can help with some granularity there if you can?
Lisa Su
Yes, we believe that we'll be able to address let's call it you know 80% of the server CPU market with our Zen class of products. So that's a very high end but you know really the meat of the market.
>>54591051
Thanks anon.
I figured AMD would be going for a race to the bottom.
Is it over?
>>54591090
Polaris comes out soon, Zen comes out at the end of the year. Everything is apparently on schedule which is unusual for AMD
>>54591104
Don't tell the shills though. Zen engineering samples have been out in the wild for weeks now, but just a few months ago every intel shill on /g/ was parroting nonsense about the chips not even being taped out yet. All based on a single transcribing error from an interview with an Indian speaking English through an incredibly heavy accent.
Su is the mastermind behind AMD's financial restructuring, and shes a total hardass. Nobody in the company can get away with fucking up any more.
>on schedule
>delayed year after year
>>54590000
>tfw both Zen and Polaris chips are made in my hometown
Should I stop by and ask for a free sample, lads?
>>54591309
You're from NY? By all means try to get a tour of Fab8 and see whats up.
>>54591317
Yeah, I have friends who work in the plant too, one for Global Foundries, another for one of the tool companies.
>tfw a guy got smushed by a tool when they were building one of the plants
>>54590251
Well according to lots of cucks on /g/ supporting games in anyway is the fucking worst thing imaginable. It's a meme but it's tiring as fuck.
>>54591279
Zen has been scheduled for 2016 since it was officially announced. Prior to the shareholder meeting in 2015(?) AMD may have mentioned Zen but never gave an actual date I think.
>>54591129
Do we have any leaked performance info about Zen at all yet?
I think that even if Zen is successful, AMD will be dead in the high performance market. Intel probably has plenty of new designs and engineering samples stored up to easily overshadow Zen. AMD has no secret weapons left.
>>54591384
Not yet.
There likely won't be any big leaks either. Word is that AMD is super serious about the NDA surrounding Zen. All the "leaked" info that Sweclockers posted a couple years back was info that AMD gave them to publish. Sounds like they actually plan on suing people for breaking NDA.
>>54591404
Intel definitely doesn't have any aces up their sleeves. They're not intentionally holding back some Godmachine architecture. You see them sputtering with 5% IPC uplifts because that is what their engineers can squeeze out. All the low hanging fruit have been picked for serial integer performance. AMD will refine the Zen core, and in a couple years be in the same position.
>>54591434
At least we'll probably get some info about AM4 at Computex. I hope there's some AM4 boards with USB-C support and 12+ SATA slots.
>>54591454
AM4 boards should be on the market by then. Bristol Ridge APUs should launch then, and they're sharing the socket.
>>54591471
>bristol ridge
What's that, the latest revision of the bulldozer arch?
>>54591538
Just Carrizo for desktop
>>54590000
Gamers had always pushed technology forward. We wouldn't have cuda compute cards, unlocked CPUs nor ssd going mainstream if it weren't for gamers buying all the fastest hardware made.
>zen to be out by the end of this year
>retro chinkpad to be out for 2017 anniversary
I think we have our ganoopad fully lined up gentlemen
>>54591685
Summit Ridge sure as shit isn't going in any laptops.
Raven Ridge APUs will probably launch mid or late 2017.
>>54591685
>Tfw retro dankpad might have 3:2 screen
The dick intensifies
>>54591546
Isn't it Carrizo with a like 6 polaris CU on board on a 14nm process? I thought they were using it as a testbed for 14nm APU. I may have heard entirely wrong and gotten bristol and raven ridges mixed up.
>>54591915
Carrizo is 28nm HP, cheap generic bulk as apposed to Kaveri's custom SHP process.
Its two Excavator modules, 8CU, 3rd generation GCN.
Raven Ridge is the first generation Zen based APU that will be 14nm. Nothing is known about it beyond that yet.
>>54590000
Who cares? Intel curbstomps AMDs CPU offerings and AMD can't catch up with Nvidia's 1080.
AMD just needs to die, honestly.
>>54591961
But if it's just Carrizo on desktop why would they bother? Carrizo was mostly about being power efficient in notebooks, makes no sense to bring it to desktop and potentially give AM4 a bad name with it's first real "mainstream" release part.
Don't get me wrong, I'll build a couple of them for light use around the house and put a Zen chip in my desktop. Bristol Ridge should be fine for my gf's normiebook machine and a mini ITX build for the garage that will run sketchup and manage a 3D printer and an OpenCNC machine.
>>54592020
The chips have a dual DDR3/4 memory controller, and they'll be able to make use of DDR4 on socket AM4.
Its all they have as a gap filler. They have no spare cash at the moment.
>>54592020
No OEM showcased the potential of Carrizo desu. Most Carrizo laptop design were limited by the 15w tdp envelope and equip with only ddr3 1600 running in single channel. A desktop carrizo 65w tdp apu, having 10% IPC improvement over Kaveri and gpu color compression in the tonga gcn should put it near an i3 haswell and with 8 CUs run most games at 1366x768 high quality or 1680x1050 med quality.
I hope we'll see Bristol Ridge and Summit Ridge in six months or so. Then I can finally upgrade from my i5 3570K.
>>54592623
>Bristol Ridge
>upgrade
No.
>>54591633
no, ssds going mainstream would happen regardless of gamers due to how massive an upgrade it is on how a computer feels to use.
>>54592667
I don't have number but I like to imagine that gamers made up most of the early adopters and a good portion of consumers of ssds.
>>54592644
Why not? It'll have 4 full Zen cores and an integrated Polaris GPU. Should be faster than my old i5.
>>54592967
No, you retard.
Bristol Ridge is just repackaged Carrizo. It is not a Zen part.
>>54590000
Just in time for my intel boycott.
>>54592996
According to the slide in the OP it will be a Zen APU.
Even if it's called something else, I still hope we'll get their Zen APU soon.
>>54593071
The slide in the OP is fake, you shit eating retard.
http://anandtech.com/show/10223/amd-bristol-ridge-in-notebooks
Do not give a shit...but after the shit Intel has pulled lately I will never buy another Intel processor as long as I FUCKING live.
>>54593084
Why are you so butthurt?
>>54592504
>Most Carrizo laptop design were limited by the 15w tdp envelope
There's a 33% performance difference for twice the TDP. It's just not worth it, 35 watt CPUs and APUs are just not worth it.
>>54593071
RavenRidge is the first Zen APU we currently know about, and that's not due until 2017.
>>54593781
Why stop at 35w tdp on deaktop. The carrizo based desktop Athlon X4 845 is 65w. A top of the line Bristol Ridge APU should be clocked above 4ghz and gpu at 900mhz at 95w tdp, similiar to the recent A10-7890K godavari refresh.
>>54593892
Bristol Ridge SKUs stop at 65w.
HP does not scale with voltage like SHP does, so it won't hit the same clocks that Kaveri did. That being said Kaveri couldn't hit the clocks that Trinity and Richland did either.
>>54593892
Because aside from the one CPU only part there is no carrizo desktop part. The sweet spot for that architecture is very obviously 15 watts, why is it not being released outside the sweet spot a problem?
>>54594226
You're gonna have a problem when I will punch to your sweet spot at 95 watts.
>>54591434
Intel's problem is that they refuse to re-balance their architecture. A huge amount of their execution capabilities are running off just 3 ports with a smattering across the other execution ports. Zen on the otherhand has their execution abilities spread across all of the ports, with many duplications amongst the ports if any of the recent leaks and speculation is correct.
>>54590000
>Zen is designed as a high end halo CPU arch
>mad about being designed for gamers and data centers
Are you fucked in the head? Those are the two biggest demographics that buy high end hardware. They'd be stupid to focus a high end CPU anywhere else.
>>54591633
Pretty sure industrial applications would have pushed the need for all of those things if gamers didn't exist. Large corporations buy these things on such a larger scale that it honestly dwarfs the purchasing power of PC gamers. Also I'm pretty sure the Steam surveys showed that most PC gamers don't actually have very high end hardware.
>>54590138
Does Zen have some feature like Hyperthreading? Otherwise I even at 80% performance and maybe $100 cheaper, it would still have half the logical cores.
>>54594814
Zen is an SMT arch, yes.
>>54594814
>Does Zen have some feature like Hyperthreading?
Yes
>>54594814
It is expected to be an SMT architecture, so yes.
>>54591090
topfuckingkek that's me
>>54594835
>>54594833
>>54594824
Thanks, that's what I was hoping for. Sadly I'll be building a pc about two months from now when I get another video editing/rendering job. Probably going for a 5820k. I'm hoping polaris is released by then at least.
>>54594940
Isnt video editing/rendering one of those things that scales really really well with more cores and threads?
If so, I'd suggest that instead of getting a 5820k, you get an E5-2690v3. You can find them for roughly the same price on ebay (maybe a bit more), and you get an additional 6c/12t. Max all core turbo is 3.1ghz, while 1 and 2 cores can get up to 3.6ghz.
>>54590053
NVidia is a much smaller company, Su has a PhD and more experience. Further, AMD had losses before she was hired, so them having losses is hardly an argument that she should get paid less. Or do you think they should have hired her by telling her that she gets a shitty salary since their company is losing money?
>>54595144
>NVidia is a much smaller company
Nvidia has higher revenues and a market cap eight times larger than AMD
im tired of waiting for zen
ill just limp along with my i5 4690 but i couldnt wait for polaris fuck knows when its even coming out as AMD are being vauge as fuck.
Ill get the new round of cards in 2018 everything else is a stopgap solution
>>54596569
Non APU Zen I think is supposed to arrive Q4 2016, APU-Zen Q1 2017.
When is Zen+?
>>54596731
Sometime early-mid 2018 hopefully.
I think its also possible that Zen+ may be on IBM's 7nm node.
>>54596750
holy shit 7nm
wait what how does IBM relate to AMD
you're scaring me
>>54596569
>i5 4690
Zen won't be an upgrade for you except maybe hyperthreading
>>54596766
IBM and GlobalFoundries and Samsung are partners in the common platform alliance. They collaborate and share tech so that they don't have to spend all of the money individually.
>>54596766
IBM is doing the bulk of the R&D on the nodes, wherein its then licensed out to GloFo and Samsung. They've already done a 7nm SiGe test wafer (pic related), so all we have to do now is wait for the bugs to be ironed out and another functional test wafer to be constructed to make sure that yes, 7nm SiGe works as it should and chips can be manufactured on it.
>>54596848
Maybe not out to Samsung, but I definitely know GloFo is supposed to get it.
Lets just hope they dont fucking botch it again.
>>54594781
games are mainstream and the common people, huge industries that make use for that tech are not and are few, more business from big to small don't use PCs for more than sending e-mails and storing some infos and usually still run that shit on pentium IVs.
>>54594837
me too senpai, I think i'm going with zen + polaris on the cheap, or zen + vega if I have the money
>>54591338
>tfw a guy got smushed by a tool when they were building one of the plants
>zen confirmed delayed to 2018
>>54592017
You need to die, idiot
I really hope Jim 'Ill fuck ur shit up' Keller delivers
>>54592504
>Most Carrizo laptop design were limited by the 15w tdp envelope
This is the same problem with their Jaguar and Puma architecture APUs. I have a feeling that if Puma APUs were designed for higher voltages and clock rates, they would give Haswell-Ts a run for its money clock-per-clock.
>>54591051
but you know
byt you know
my eyes burn
did you make it up anon?
>>54590000
Compete with Sky lake, when Kaby lake. AMD literally one generation behind.
>>54596798
>>54596848
IBM as a foundry no longer exists. Their fab business, IP, engineers, and a bunch of staff are all part of Global Foundries now.
>>54590000
>FM3
Wtf I thought Zen was supposed to be on AM4?
>>54603286
OP's picture is fake.
Though FM3 is an internal development name that AMD has used for a while now. The socket's noncode name is AM4.
>>54603286
Fake slide released prior to the 2015 financial analysis day. Most of the information on it is wrong or somewhat inaccurate.
>>54590000
Bristol Ridge wasn't supposed to be Zen I thought
>>54604021
Read the damn thread.
>>54600894
a few percentage point difference in IPC won't matter if they're still offering more cores for less, it'll be certainly better than the current 50% gap
>>54604044
Per core with normalized clocks, Excavator is 50-90% slower than Haswell in a number of FPU heavy ops. Even in X265 encoding. The performance gap between AMD and intel is far larger than 50% if you take all performance metrics into account.
http://forums.anandtech.com/showthread.php?t=2465645&page=40
>>54604192
If they sell a decent midrange 4-6 core product for reasonable monies I'm in. I have an i5-2400 and a Xeon e3-1240 machine. If it keeps up with that but is cheap I'm happy.
>>54590000
Nice quads.
Also how long until their new CPU's? I want to know when to upgrade (can't switch to Intel, not replacing my mobo).
>>54604192
>muh IPC
Can't this meme die already?
It's like measuring the performance of an internal combustion engine by only measuring displacement of a single cylinder.
>>54604308
Q4 2016 for the non-apus, Q1 2017 for the Apus.
I think. I know Zen in some form is supposed to start dropping in Q4 2016.
>>54604347
>l-lets not measure efficiency
>if you i-ignore this metric my product is s-still r-relevant
AMD faggots literally on suicide watch
>>54604308
The first Summit Ridge chips will be out at the end of the year, and you'll have to replace your motherboard.
>>54604347
>calling things a meme because you don't understand them
None of AMD's parts clock high enough to make up for the disparity in IPC.
Zen is touted as a 40% uplift in IPC over Excavator, and Excavator can be as much as 90% behind *Haswell* in certain ops. Summit Ridge is going to be competing against Skyake and Kaby Lake. Its not going head to head with Haswell so they're at an even greater disadvantage.
A 40% uplift is still going to be crushed in anything FPUcentric.
>>54604397
Well shit, looks like I'll just get the most recent AM3+ line if muh gaymes need it and then just rebuild completely in 3-5 years.
>>54604397
You're forgetting that AMD is no longer using the fused floating point unit that they were using in the construction cores.
Pic related is what Zen is currently expected to look like architecturally according to the latest leaks.
>>54604391
>spending 50% more die area to get 10% more IPC is efficient
>>54604397
10 years ago you might have had a point. In the era of 8, 16 and 32 core processors you don't.
>>54590000
if I have to buy a new motherboard anyways,
why wouldn't I get a i7-6700k while I'm at it?
>>54604478
>moving goalposts
Like I said, if it's similar in performance but cheaper I'll get it.
>>54604470
That has no bearing on throughput. If they could achieve a greater than 40% uplift in FPU performance they would be touting that figure, but they're not. They would be saying that they could address all areas of the server market, but they're not.
There is literally nothing worse than internet fanboys who hype themselves up into believing that X new product is going to be the second coming. They refuse to listen to reason, rabidly shitpost to defend their delusional position, then act like they never saw it coming when the product doesn't live up to their fantasy. AMD didn't create the Bulldozer media fiasco. Forum fanboys did. All the little retards on /g/ did.
>>54604478
>still proving you don't understand what these words mean
An 8 core 16 thread CPU clocked at 3ghz with an IPC of 1
vs
An 8 core 16 thread CPU clocked at 3ghz with an IPC of .5
Guess which one performs better? AMD doesn't offer more cores than intel does, not in HEDT, nor in enterprise either. They're not winning out in anything by offering more cores and higher clockspeeds. The only exception to this was Vishera vs stock Sandy Bridge and Ivy Bridge i5/i7s in select integer bound workloads. It isn't true any more.
>>54604478
IPC matters because that kind of parallelism isn't available in most tasks
>>54604592
>AMD doesn't offer more cores than intel
[citation needed]
Rumor has it there won't even be any Zen quadcores. It's all 6 and 8 cores against Kaby Lake's 4 cores.
>>54604643
When was the last time anyone who does any serious computing asked for more per-core performance?
Moar cores and if that don't work more GPUs is what matters these days.
>>54590053
/g/ discussing business give me a headache... More pay means more money so a ceo is better? Justified by the company? Are you even aware on how a ceo is remunerated and do you know how much stock mr. Nvidia holds relative to former employer AMD and their segmentation.. Is your life dictated by salary and brand perception relative to financial strength
Intel is shifting from the pc market to higher growth sectors, nvidia has a huge bet on driverless cars, amd has relative growth potential in china striking some relatively high deals, lost some discrete market share but its still up for play, nvidia doing their best to nail the VR coffin literally
Competition is needed or rather an illusion of choice
>>54604592
>AMD doesn't offer more cores than intel does, not in HEDT, nor in enterprise either.
Server Zen is supposed to offer 16C APUs and 32C CPUs.
Even at Haswell level performance, it will beat intel quite handily in core count, as the best intel can do is 24C (if they turn the last two on), because intel has this massive fucking boner for monster monolithic dies that yield like shit.
In the lower consumer segment, AMD will be doing 6 and 8c dies vs fucking 4 from intel. In the upper, AMD can easily release binned Zen server dies as 12c enthusiast chips, and the best Intel can respond with is the 10c 5960x successor.
>>54590000
AMD will fuck up once again just like with >moar cores! because VR is meme tech just like 3D TVs.
Mark my fucking words
>>54604733
>Rumor has it there won't even be any Zen quadcores
They meant no single-module CPU-only Zens at launch. The APUs will all be single Zen modules, or 4 cores, and will come after Q1 2017.
AM4 will only debut with dual module Excavator APUs ("quad-core") and dual module Zen CPUs. (6/8-core)
>>54604733
>I live under a rock
i7e comes in 6,8, and soon to be 10 core variants.
They have 18 and 22 core Xeon dies.
>>54604776
Thats one specific chip, a purpose built APU. It is not their default server platform, and its not a part that has any real timeframe yet. Its just something that AMD kind of always wanted to do, and had noted that such a chip would be technically possible around 18.
It was originally envisioned as a Bulldozer chip. The recent Zeppelin codename hasn't materialized as anything yet.
Such a part would tackle intel based systems using a secondary Xeon Phi card. They're not general purpose.
>>54604966
well, its also expected that AMD will do MCMs like they've done in the past, so "purpose built APU" would be more along the lines of "Stick 16C CPU die, a modified Polaris Die, and some HBM on an interposer, then shove it into the package". Much easier than making a whole new die specifically for it.
With the exception of the monster that was Fiji, AMD doesnt really do big monolithic dies. They take a more modular approach, and actually I wouldnt be surprised if their 32C server CPUs actually end up being a quartet of 8C dies on the same physical package. It'd be notably cheaper to manufacture than a pair of 16C dies
>>54604966
>i7e comes in 6,8, and soon to be 10 core variants.They have 18 and 22 core Xeon dies.
most consumers don't use the X series boards/ chips, too expensive.
even on the Xeon side, most applications don't need that much CPU performance and won't bother due to the costs.
>>54592017
Why do you want monopolies.
>>54605587
Because hes a shitposter.
>>54605120
>It'd be notably cheaper to manufacture than a pair of 16C dies
Then why did Intel stop doing that after the Pentium D?
>>54606615
They didnt. Intel kept up with the MCM scheme with the core 2 quads, which were just a pair of C2D dies on the same package. They stopped because 12MB of L2 cache was becoming a bastard to keep coherent across 2 dies, and FSB was seriously starting to show its age, especially with server systems where as many as 24 individual cores (Intel Dunnington; native hex-core Core 2 die) were fighting over the same memory and front side bus.
Going to the on-die memory controller then giving the dies high speed *bidirectional* interconnects (FSB could only go one way at a time, QPI, DMI, and HyperTransport can go both ways simultaneously) means communication between multiple dies in the same package is much simpler and faster, and it becomes possible to feasibly stuff multiple dies on the same package again.
Please tell me what AMD could possibly release that will favorably compete with my overcooked i3 6100 @ 4.5ghz for $120?
>>54590000
Who cares, let them make what sells. They can start branching out when they're earning more
>>54606779
How about an i3 skylake comparable cpu with gtx 950 comparable gpu baked in for $200
>>54607176
>How about an i3 skylake comparable cpu with gtx 950 comparable gpu baked in for $200
kek
Not happening
"Console-level" APUs Lisa Suu was talking about is about R7 265 tier at best, so that's more like GTX 750 Ti level of performance if she's not talking out of her ass like her predecessors.
I doubt that it'll even come close to that, though. AMD always overpromise and under-deliver. I bet you anything that it'll offer Ivy Bridge CPU performance at its highest end APU and GT 740-R7 250X GPU performance. There's no way that it can go past that without hitting a thermal barrier.
>>54606779
>i3-6100
>4.5 Ghz
That's impossible, the i3-6100 is a low-binned locked multiplier chip. A i3-6300 sounds more realistic, but that's also a locked multiplier chip.
>>54607270
>doesn't know about BASED Skylake BCLK OC
Too bad it puts the actual CPU at risk so that Intel had to disable it.
>>54607265
4 Zen cores and 20~ CU could fit onto a 125mm~ die, and it'd be within a 95w TDP depending on how they're clocked.
A 14nm desktop APU that totally trounces the PS4 in graphics performance is an easy task. The PS4 was a low power cheap part when it was originally designed, and its just 28nm HPP.
The only lynch pin AMD would need for such a desktop chip is on package HBM. Personally I don't expect much from the upcoming Raven Ridge. AMD does have a history of disappointment.
>>54607303
>4 Zen cores and 20~ CU could fit onto a 125mm~ die, and it'd be within a 95w TDP depending on how they're clocked
You don't know that. Rumor has it that Zen's modules are bigger than Excavator, even at "14 nm". And I seriously don't think 20 GCN CUs can fit alongside even the shrunk down Excavator, let alone a hypothetically smaller Zen.
>The only lynch pin AMD would need for such a desktop chip is on package HBM
kek
What makes you think that AMD would put such an expensive, space hungry feature on a low-end chip? Because APUs are only meant for the low-end to make up for the lack of cores.
>>54590768
JIM
I
M
JIM
I
M
JIM
J
I
M
>>54607284
Doesn't BCLK OC work on basically all CPUs?
And how the fug do you get 4.5 Ghz from a bottom bin i3?
>>54607343
>Doesn't BCLK OC work on basically all CPUs?
Technically it's supposed to (back in the early days, that was the only way to overclock CPUs), but modern Intel CPUs tie the memory controller and the DMI ( to the BCLK itself, so increasing the BCLK beyond its stock 100MHz frequency runs the risk of instability with the PCI/e slots, the SATA ports, the ethernet ports, and the memory.
Plus, modern Intel CPUs don't like having their BCLK fiddled with even if it's an unlocked K SKU. There sees to be some sort of safe-guard or kill-switch that prevents the CPU from running at a BCLK higher than 125MHz, even with adequate cooling on everything.
AMD APU/CPUs are a bit more flexible, but the APUs largely have the same BCLK issues as their equivalents Intel if you run them too high.
>>54607335
An Excavator module scaled down to 14nm would be 4.344mm2.
If the entire Carrizo APU were scaled down it would be 75mm2.
If the entire PS4 APU were scaled down it would be 104.4mm2 - This is 8 Jaguar cores, 256bit GDDR5 pads, uncore, and 20 2nd generation GCN CU.
If a single Zen core were 40% larger than an entire Excavator module then it would end up as 6mm2~ in silicon. Thats not a behemoth by any means. My above statement wasn't a guess.
4 Zen cores and 20CU would fit in a 125mm2 die.
I have a 3470, and 970.
Would Zen, and Polaris be a decent upgrade?
>>54607398
Right, but how the fuck is 4.5 Ghz on a Core i3-6100 possible, it's the lowest bin/lowest clocked of the Core series.
>>54607411
Do you know why Excavator is so small compared to Steamroller, despite sharing the same lithography?
If you do know, it's that, but reversed for Zen due to its added features and support.
And don't kid yourself, a 14nm Excavator won't shrink down to half its size. At best, it's a 65-70% shrink.
>>54607428
So I shouldn't?
>>54607265
Wcftech revealed performance of mobile zen and their 950 equivalent ran at 40 watts tdp and their flagship notebook chip was a die shrunk 390x
No reason a desktop apu can't compare to a gtx950 that was what was teased already
>>54607425
>it's the lowest bin/lowest clocked of the Core series.
They're just dual-core i3/Pentium/Celeron dies with faulty L2$. Other than the 1MB smaller L2$, the i3-6100 can run at the same clock speed as the 6320, probably with higher voltages though.
>>54607442
>L2$
Sorry, that should be level 3 cache. Unlike AMD, Intel actually has level 3 caching that doesn't suck shit through a straw.
>>54607442
Neat, I learned something new today, pretty rare for me.
>>54607451
>suck shit through a straw
Top kek
>>54595144
that was true 15 years ago. NVIDIA is much larger than AMD is now, AMD can't even afford to hire whites any more. They only hire H1B indians.
>>54607431
you're using a decent system
wait until Zen+ (or intel chip of the comparable generation) and Vega (or later GPU)
>>54607470
Yea that is true. I built this PC on Nov of 2015, so it is still fairly new.
Thought that it would be cool though, but good advice anyways.
>>54607430
>I have no idea what I'm talking about
Thats just adorable. Excavator had a significant area drop over Steamroller because its a fully automated design and it uses a thin transistor library typically used in GPU logic. This is totally irrelevant when talking about 14nm because you can't compare a planar structure to a Fin. By default all Fins are thin structures where not all planar gates are.
The area scaling factor for 14nm vs 28nm is just over 70%.
Put plainly a 100mm2 28nm die would be 30mm2 in 14nm.
Stop trying to talk out of your ass. I know far more than you ever will.
>>54607462
What are you, 12?
>>54607455
Don't listen to him. 6100 and 6320 are binned for frequency and only the best chips become the 6320. The guy is probably over-volting his i3-6100 to death with that overclock. You should only buy 6320 if you even want to consider BCLK overclocking. Only buy the best.
>>54607479
>By default all Fins are thin structures where not all planar gates are
>The area scaling factor for 14nm vs 28nm is just over 70%.
Looks like you're the one who has no idea what they're talking about.
>>54607466
If you've ever worked at a manufacturing plant in America, you'll know Indians are alot more expensive than whites, senpai.
>>54607520
>tech illiterate babby realized hes out of his league
Absolutely precious.
Tell me, what goes through your mind when you type out a post? Why do you think you can just talk out of your ass all day without being called out? Are you so autismal and desperate to feel superior that you just go into threads and lie about being an expert on things?
Every time I find a little retard like you I ask the same questions, yet none of you can ever give me an answer.
>>54607503
Makes more sense, although I already fell for the i3-6100 meme, I'm not planning on overclocking it.
>>54607550
>i can read a few marketing slides, so that makes me an instant expert on the topic
I should be asking you the same question, you mindless sheep
>>54607561
How dare you challenge me with your somewhat primitive skills?
>>54607561
You didn't answer, little autist.
That isn't a marketing slide, either. Good job trying to deflect.
Its always easy to read retards like you and figure out how shit your life must be for you to do this every day.
>>54607552
No, he's wrong. Literally the only difference between the 6100 and the 6320/6300 is the L3$. Intel just pulls a Jew with the chips clock frequencies by binning a few that can handle higher frequencies with lower voltages. Given the right amount of voltage, they should both be able to hit the same frequencies. I've had a 2600K overclock higher than 2700K, despite the 2700K being rated and sold higher because the 2700K simply would not go over 4.4 GHz. It's the work of the Jews at Intel.
>>54607590
>glofo presentation slides
>not marketing
Guess who the slides are meant to appeal to?
You think consumers are the only folks that need marketing towards? Have you ever worked in the corporate world?
Oh wait
>shitposts and fanboys on 4chan
>uses a single slide from an official looking presentation to base his knowledge and claim superiority over his autistic brethen
Clearly you have not. I wonder why I even bother with you idiots.
>>54607653
>even more shitposting from the autist
>trying to deflect constantly
>trying to discredit empirical information because it proves him wrong
I bet you're a NEET too.
A technical presentation given by a PhD on the nuances of a new process design flows to the engineering departments of their customers is not marketing. They are not courting new clients, they are giving a primer to their current customers.
Continually talking out of your ass and digging yourself a deeper hole isn't going to make you any less pathetic.
https://www.youtube.com/watch?v=SPQvgzL_DSc
>>54607708
>getting rused this hard
I saw this coming from a mile away. You could have not bothered with the cunt after reading this >>54607520
That's a bait post if I ever saw one
>>54607439
>die shrunk 390x
Gonna need sauce on that. Core count and memory controllers say otherwise.
