Spot the AMD shill.

File: HC28.AMD.Mike Clark.final-page-014.jpg (405KB, 3999x2250px) Image search: [Google]

405KB, 3999x2250px

>>56234651

>>56234660
>AMD has money to pay shills

File: HC28.AMD.Mike Clark.final-page-012.jpg (382KB, 3999x2250px) Image search: [Google]

382KB, 3999x2250px

>>56234688

File: HC28.AMD.Mike Clark.final-page-015.jpg (482KB, 3999x2250px) Image search: [Google]

482KB, 3999x2250px

>>56234711

File: HC28.AMD.Mike Clark.final-page-016.jpg (388KB, 3999x2250px) Image search: [Google]

388KB, 3999x2250px

>>56234733

Virtually the whole slide deck leaked yesterday. I rifled through them and posted a bunch here.

Key take aways:

Zen has the same frequency scaling per watt that Excavator does.
That means about 15w per core at 3.5ghz
10w per core at 3.2ghz
and roughly 5.5w per core for 2.45ghz

The L2 is coherent across the CCX
The L3 is coherent across the entire chip, and all chips in an MCM
This should have been obvious, desperate shitposting shills have been claiming otherwise, but the memory channels are coherent as well. It is not one channel per CCX. Each CCX per die has dual channel memory.
The biggest improvements to the core one could point to are cache bandwidth and latency, the front end, and shear volume of execution resources per core compared to the BD family


Single die = dual channel memory
Dual die = quad channel memory
Quad die = octa channel memory - disruptive memory bandwidth per socket for enterprise chips

The 4 die MCM 32c/64t Opteron with 8 channel memory is 180w and is clocked about 2.5ghz, it will out perform the 24c/48t Xeon E7 8890v4 in total throughput.
The Zen core arch will be big in the server market.


Gamers looking for a new platform won't find what they're looking for in Summit Ridge. The chip isn't going to clock and come anywhere near Skylake.

>>56234758
>PTE

INTEL BTFO

HOW WILL THEY EVER RECOVER?????

P O O
O
O

Just beat Sandy or Ivy and I'm switching. I've been itching for more cores on ffmpeg and make.

>>56235000
Dumb shitposter. The poo memes are only with pajeets and rajesh. Lisa and Keller have nothing to do with them, hell not even Raja has anything to do with them. If Raja was a pajeet, then you and I would be the shit from India.

File: Zen team.jpg (75KB, 715x573px) Image search: [Google]

75KB, 715x573px

>>56235092
The entire Zen design team is white people sans one Sikh anyway

>>56235092
>Dumb shitposter
I just want to briefly second this, just so you get a primitive tribal feel of being surrounded by a group that's laughing at you.

>>56235092
This. Fucking idiots can't even meme anymore.

I'm mildly optimistic about the performance.

>>56234985
No talk about thermal? Seems like no talk about APU as well. I'm thinking of building HTPC that can do gaming at low-medium settings as well with Zen's APU once it comes out. I don't want to install giant heatsink with loud fans, or having to deal with water cooling.

I'm currently holding back buying Carrizo, not to mention someone mentioned this in the comment:
http://www.extremetech.com/computing/222662-poor-oem-designs-single-channel-ram-sabotage-amds-carrizo

>>56236062
Derp, nevermind that article, seems like it's for laptops.

File: 1438271982740.jpg (89KB, 768x1024px) Image search: [Google]

89KB, 768x1024px

Literally the only reason to buy an AMD CPU is if you are some kind of fanatical fanboy.

There is no other legit reason.

>>56234660
Dude
Both Intel fanboys
and AMD fanboys
want Zen to succeed

Because then Intel will get off their ass and work harder again.

Competition is good in this case

14LPP being a disaster will kill any chance Zen has outside of the supercomputer space. Maybe if AMD can find a way to weasel out of the WSA before Zen+ is ready they can get it out on 16FF and have a real competitor on their hands.

>>56236173
>Because then Intel will get off their ass and work harder again.

Intel doesn't need to get off their ass. Intel's best selling products right now are xeon-d and gaymer shit, stuff that AMD has absolutely no answer to (and won't until 2018 or whenever K12 is finally ready).

>>56236195
>14LPP being a disaster
Elaborate?

>>56236195
>I don't understand why market competition is good

Intel doesn't *want* to get off their asses. The only reason they're not even better is because AMD fell so far behind after they sabotaged AMD in the pre-built market and business contracts. Market manipulation was easier than actually competing with AMD.

If they'd actually kept up like in the original Athlon days, we'd have much better chips now than we do

>>56236227
Invisible hand blah blah.
Free market blah blah.
[delusions about human nature]
Etc.

>>56236227

I never said competition was bad. I said AMD won't be competing.

AMD going under would probably have an end result of several new competitors entering the field because the patents related to x86 and x86-64 would eventually end up in the public domain if Intel keeps being jewish with them.

>>56236360
>patents
Actually I think AMD going under would be a disaster for the market because of patents.

No power on this planet would allow the patents to go into the public domain (For fucks sake, we can barely get 85+ year old movies/books/cartoons into public domain because of this) ---
AND it's more profitable for any single player in the ensuing shitstorm to get any pieces at all even if it breaks up useful patents even if this renders doing anything a useless nightmare, therefore they will try to because it's more profitable for any single player.

Patent trolling is on the up and up, and only going further as technology gets more complex.

>>56236360
AMD would never "go under", it'd either slowly and selectively sell off its IP, or more likely, get gobbled up by someone else. Probably Samsung.

>>56236424

you're confusing copyright and patents. several patents related to x86 have already expired, you can build an x86 cpu right now legally, you just can't build an x86-64 cpu or one with sse2/3/4 and avx support and all the other misc extensions that are locked down by amd and intel.

if amd or intel goes under there would be a good chance that these would be opened up to avoid total market dominance by one company.

>>56236459
If AMD went bankrupt and/or was bought out their patent portfolio would just go to whoever acquired them. IP is frequently bought and sold these days.

>>56236723
inb4 people claim the x86 license wouldn't transfer

>>56236793
Under the original terms it doesn't but there's nothing stopping the company that acquires AMD from negotiating for a new license. Intel might even be obligated to do so by regulators who would likely be overseeing AMD's acquisition.

>>56236836
This is an important aspect that people don't tend to realize. AMD is more or less all that stands in the way of Intel getting into anti-trust troubles, possibly even being broken up by force.

>>56236195
intel fanboys from the core era have made AMD completely forget about the area.
they have consoles, laptops, servers and super computers to make.

gaymen and desktop fags can enjoy the small niche that intel dominates in. while they continue milking retards who buy overpriced garbage year after year, AMD is making bucks where it really counts, while bringing new tech to the table.

when was the last time intel did anything groundbreaking? integrated gpus? what a joke...

>>56236934
>when was the last time intel did anything groundbreaking? integrated gpus? what a joke...
just shows how worthless gaymen are, they don't even allow tech to progress in the right direction.

instead it's just rebrands, and assholes putting more work into locking their processors to charge 40% more on "k" varient faggotry.

>>56236150
>Being better at parallel tasks doesn't count as a reason

But you're wrong, you stupid nigger.

>>56236934
>when was the last time intel did anything groundbreaking?
I tend to have this mindset as well, but you need to bear in mind there's a much bigger picture and a lot behind the scenes you don't directly see in a "product". Like manufacturing tech, different doping materials, etc.

This is the reason why Intel's 14nm process generates slightly smaller transistors with slightly better power and thermal characteristics. Not all of their billions is fruitlessly, visionlessly, and incompetently thrown into a void.

>>56234651
wtf i hate intel now

>>56236995
marketing tactics, like buzzwords. and how to fuck over customers by locking them out of overclock potential unless they pay out the ass..
pretty sure that doesn't cost billions.
>14nm process generates slightly smaller transistors with slightly better power and thermal characteristics.
that will happen for every product shifting down nm size in production, regardless of company.

>>56235122
Sikhs are bro tier though so he gets a pass

>>56237328
>that will happen for every product shifting down nm size in production, regardless of company.
Processes aren't equivalent, even if they target the same general scale.

Look into a bit.

if AMD died today, I don't think anyone would care much about x86 anymore. my guess is that ARM would win even more customers...
why? because intel is a piece of shit

Intel would be truly amazing if the fossils in management didn't hold it down.

>>56237687
And AMD would be truly amazing if it had money and Intel's fabs

>>56237687
>X would be truly amazing if not for Y
That's the human condition and the nature of our species, which is greatly overexpressed by capitalism.

Very difficult to counterbalance.

>>56237706
Maybe.

>>56237687
Intel's problem is that it doesn't take risks, and thus manages to create nothing really exciting (last interesting thing was Xeon-D and that's just a high core xeon with a GPU and plenty of I/O, nothing really we didn't expect)
AMD takes plenty of risks and falls flat on its face the same amount of times, but at least they try.

>>56236150
I buy AMD so the eternal Jew gets no money from me.
You stick with your own people.

>>56237362
depends on your definition of equivalent
but it's been proven, shifting down allows more transistors and less heat. again, regardless of company, whether it's intel or AMD or even nvidia

>>56236862
that and also the fact that zillions of other cpu manufacturers and designers exist. the idea that x86 design and manufacture is a legally discrete market in the anti-trust sense is pure speculation.

in any case AMD's going nowhere in the foreseeable future.

>>56237753
mantle, hbm, ddr and gddr
amd is always pushing
while everyone else just buys up the competition then tries to milk long as they can until amd comes out with something new for them to steal and then overprice even when a third party does it (see physx for reference)

>>56238122
it's owned by the same people.
notice how both gpu and cpu have a duopoly atm?

>>56238177
I'm not an Intel fan, I'm just putting it how it really is. The development of different doping agents and lithography substrates, does ultimately yield functional differences between processes.

http://www.fool.com/investing/general/2015/05/18/heres-how-samsungs-14-nanometer-transistors-compar.aspx

As a whole, Intel is garbage as a company. But you must be realistic.

>>56235122
I wonder how he held in there instead of following the rest of the *cat team to Samsung
>>56236210
>Elaborate?
He's a dumb poster, he believed the FUD of the Applel chips that were fabbed by both TSMC and Samsung, and somehow thinks the RX480 should be hitting 2 Ghz when GCN has a short pipeline suited to less than 1 Ghz
>>56236195
>tel doesn't need to get off their ass. Intel's best selling products right now are xeon-d
This is what AMD is targeting, this will be a massive change in the server market, Zen's massive throughput will justify the higher power consumption over Xeon-D parts
>>56237665
>why? because intel is a piece of shit
Intel's now an ARM licensee, we will see ARM SoC's by Intel in a not too distant future
>>56237753
Intel has massive risks more than once, all of them quite promising and spectacular failures, with a few just being major dissapointments
From their early RISC efforts to the i740, then Itanium and then Labarree, then the whole Atom line

>>56235122

They have a team doing R&D in China.

The worrisome thing is the fact that the L3 is going to be some large victim cache instead of a writeback cache as speculated. And I'm not so sure about the L2 size being that big and the benefits in terms of speed and latency. I do agree with the reasoning for a 64 KB instruction L1 cache.

The new AMD specific instructions with the cache and page table is interesting. Maybe they can standardize something there. Now it really comes down to execution.

I'm glad I brought 1k stock for $2.35 average. Easy profits.

>>56234651
When will AMD stop with these shitty slides and bring real benchmarks and comparisons?
This is the exact same shill they did with FX-series and RX-series(gpu).

>it's the year 2152
>AMD pushes ZEN back another year
>you're still waiting and posting colorful graphics on how awesome ZEN will be

>>56239029
Zen has never been delayed senpai, Intel is desperate with this shilling

>>56239088
actually I just bought a 8350 because Zen is taking forever
still AMD sucks major donkey balls
they could hurry up a little bit
they advertise Zen for 1,5 years now, or even longer

>>56239111
>actually I just bought a 8350 because Zen is taking forever
Literally you didn't actually do that, this is textbook crypto.me shilling
>still AMD sucks major donkey balls
>they could hurry up a little bit
Passive-aggresive bullshit
>they advertise Zen for 1,5 years now, or even longer
And as dessert another textbook crypto.me tactic, pulling literal bullshit out of nowhere, AMD states Zen availability for end of 2016, and full volume at Q1 2017
And it has always been like this, only WCCF and other rumor mills have stated otherwise

Literal b8/shill post

what

File: 1470807518493.png (436KB, 532x502px) Image search: [Google]

436KB, 532x502px

>>56236173
>Because then Intel will get off their ass and work harder again.

What will actually happen is that they'll just Jew AMD by paying off OEMs again, get a slap on the wrist from court, and continue giving you a 5% upgrade every year at the same prices like always while AMD falls to bankruptcy.

>>56239250
>>What will actually happen is that they'll just Jew AMD by paying off OEMs again, get a slap on the wrist from court
Last time this happened we got Core 2

>>56238263
>Intel has massive risks more than once, all of them quite promising and spectacular failures, with a few just being major dissapointments
>From their early RISC efforts to the i740, then Itanium and then Labarree, then the whole Atom line
None of these are anything close to recent, their highest continuous risk is their cutting edge fabs, but as for interest IP, nothing.
Unless you count xpoint

File: clearly intel shill.jpg (536KB, 1434x1080px) Image search: [Google]

536KB, 1434x1080px

>>56239236
>implications

>>56239291
Then you're trolling, or an applied shill, or worse an unpaid shill

>>56239303
SHILLING FOR WHAT

Because I don't support the long wait for Zen?
Fucking fanboys holy shit
stop sucking AMDs dick

>>56234651
Finaly some competition.

Now if they price their quad cores with HT competitively, intel will be forced to lower their prices for their i7's too and I might finaly upgrade my 4.4 ghz I7 2600k

>>56239316
>SHILLING FOR WHAT
Pure disinformation is shilling
>Because I don't support the long wait for Zen?
What long wait? Zen has always been on track, you're retarded for buying a dead socket and a 4 year old CPU 6 months before there's availability for a massive revision
>Fucking fanboys holy shit
>stop sucking AMDs dick
Textbook shilling
>>56239329
Intel prices will go down across the board bar a few Extreme Edition chips, but the slides point that Zen might not compete with the latest Intel HEDT offerings well enough

>>56239264
Its not like it can happen again because Intel already has all the marketshare

>>56239371
had the socket already, only bought the CPU

if anyone here is shilling, it's you

>>56239405
You bought a 8350 6 months before Zen has availability, let that sink in

>>56239405
I hope you bought it second hand,
AMD FX second hand cpu's are everywhere these days

>>56239291

>8350
>23C
Fake and gay

>>56239492
Speccy reports AMD CPU's temps wrong, either way too hot or way too cold

>>56239492
325W TDP cooler
>>56239458
>>56239481
I might be an AMD user, but I'm not too cheap for a 150€ something CPU
your crying won't make me regret my choice

You honestly want to make me feel bad that I bought AMD
what is your goal?

>>56239616
>You honestly want to make me feel bad that I bought AMD
>what is your goal?
Nice, completely moving the goalposts
Well, at least you won't be spreading FUD about Zen's release date

>>56239616
>You honestly want to make me feel bad that I bought AMD

there's nothing wrong with buying AMD, there's something wrong with buying a sub par cpu for full price when the second hand market is flooded with them, for 150 euros I could get a mobo+8 gb ram+fx cpu where I live

thats how desperate people are to get rid of it

>what is your goal?

to teach you a life long lesson kid

>>56239658
It's ok if you want to buy second hand
I don't and I will never, you won't teach me anything, kid

File: lol15.jpg (33KB, 640x480px) Image search: [Google]

33KB, 640x480px

>>56239695
Not that guy but I'm dying here. Either shill of hardcore amdrone. Thanks for the laugh though.

>>56236062
The APU is called Raven Ridge.
Both Raven Ridge and Summit Ridge top out at 95w SKUs.

>>56236085
>Summit Ridge
>laptops
stay in /v/

>>56236210
Hes just regurgitating bullshit he doesn't understand. One shitposting NEET here has been on a tirade trying to claim theres a problem with the 14nm LPP process because the RX 480 doesn't compete with the GTX 980.
I could go into detail explaining the clocking characteristics and voltage scaling displayed by the process at length, but its pointless. I've done so a handful of times now and its wasted on this board.

The Zen core in Summit Ridge and Raven Ridge parts won't be a speed demon. About 4ghz will be their max OC. This isn't from the process though, its from the architecture. At 3.2ghz a Zen core is drawing 10w. Thats not bad for a server part, in fact its pretty amazing. That just isn't enough frequency to be competitive in gaming benches for a mainstream desktop part.
AMD's new Zen core is the biggest thing they've ever design, the single core is larger than an entire BD family module on ISO process. They've done a tremendous amount of work over the past few years developing power saving IP to decrease energy per op, it just hasn't been enough. They weren't working with a big enough design.

For reference about 10w per core is what 4ghz Skylake allots for itself.
The Raven Ridge APUs however will use a different variant of 14nm LPP from Summit Ridge, but its not going to be drastic. They might hit around 3.5ghz at 10w instead of 3.2ghz.

>>56239763
>the autistic amdrone is here
Damn half your post is pasta.

>>56239744
first I get called intel shill, then troll, after that AMD shill, then you try to make me regret my AMD buy
and I'm clearly none of those
the only problem here is that you can't comprehend a reasonable human being posting on the chans
and the only thing I regret is trying to argue with you autistic little children

>>56239773
I'm not an AMD fanboy in anyway and literally nothing in my post is copypasta.

Go hang yourself, child.

>>56239797
>calls someone a kid
>expects to be takes seriously
Listen here, moron. CPUs are one of the things you can safely buy second hand. You are buying several years old cpu at full price. The only kid who can't manage money here is you.

Some concrete opinions on Zen are up on Realworldtech forums, go read.

>>56239835
Nice reading comprehension. I was called kid first btw.
Why does it trigger you so hard though?

>>56239773
You were referring to me, actually, and I was going to say that anon's post was nice and informative if rather winded.

:^)

>>56239797
>>56239801
Im with
>>56239835
This guy, the fuckin child insult is weak in a discussion, as is the an hero suggestion.

Good bread though, ive been wondering what amds aim for zen was considering pc famings a kinda shit niche, i figured cpus for consoles would be a big aim but it seems servers are where they want to make their money

>>56239855
>called first by some other guy
>I'll call everyone kid
Ok, kid.

>>56234651
>Bought Shares in AMD at about $1.88
>Shares now worth about $7.60
>Feels good
>Not sure how long to hold onto them
>what to do with them

>>56239889
sell before AMD Zen is released,unless you want to take the risk

>>56239906
Zen will be a boost. It won't beat Intel, but it'll match it quite well.

>>56239840
Just people going back and forth about the L3 cache using a ring buffer, something AMD amply demonstrated in the PS4 APU's handling of GDDR5 for CPU logic.
The statement about cores accessing any L3 slice is more to do with enterprise Opterons. Any core in the 4 die MCM will have average latency to any L3.

>>56239906
>sell before release
>when the company is going to regain enterprise market share and have positive cash flow

AMD's stock is staging for a bubble.

>>56239925
Any new release of anything always dips shares a bit, it's probably safe to sell a week or two before launch, and buy back in the dip

>>56239936
Proof?

>>56239932
>Just people going back and forth about the L3 cache using a ring buffer, something AMD amply demonstrated in the PS4 APU's handling of GDDR5 for CPU logic.
>The statement about cores accessing any L3 slice is more to do with enterprise Opterons. Any core in the 4 die MCM will have average latency to any L3.
Not like there's much else to talk about since Zen looks very Haswell-like

>>56239873
kid

>>56240012
I disagree with the sentiment that Zen looks like any of intel's Core i arch. It really doesn't at any micro or macro level.
It looks like a larger Jaguar core, and Jaguar looks like a tiny refined K10 targeting low clocks.

>>56234985
The important performance enabler that AMD didn't cover here was how they'll handle turbo core.
If a single thread can operate north of 3.7ghz, consuming 20w~ then they'll be far more competitive in less threaded workloads.
Using half of each CCX to create a highly clocked 4c chip depending on the workload would be incredibly clever.

>>56240955
>3.7GHz

LOL

Intel Kaby lake is 4.5GHz!

>>56240960
Summit Ridge is not competing against mainstream Kaby Lake. Its competing against Haswell-E, and Broadwell-E which only clocks up to 4ghz on a single thread in the top most SKU.

>>56239932
Wont we be looking at, absolutely nominally, 25GB/s over the interconnect fabric for each 8 core unit? Is that enough, latency talk aside.
Intel L3 runs much faster than that.

>>56236934
>>56236934
>when was the last time intel did anything groundbreaking? integrated gpus? what a joke...
3D xpoint and consumer SSDs is all I can think off now

nobody in this thread read or understood the article

I just want a Skylake-like cpu which isn't coming from Jewtel.

>>56241351
25GB/s is the alleged bandwidth per GMI link, but I don't trust random Fudzilla leaks when they make their own graphics.
AMD made the claim of nearly 5X the bandwidth for their L3 cache, and that all cores can access any L3 slice with average latency.
Theres far more to the band end fabric they're using than anyone ha pick up on yet.

>>56241371
Xpoint is literally Micron's internally developed phase change memory with intel branding put on it.

>>56241395
Says you, dipshit.

>>56241395
>Ị'm dumb
>so everyone else must be

How do you function in society.

>>56241430
Intel is working on optoelectronics.

>>56239763
I just want to build a general purpose PC with the form factor of a PS4 desu. I'm thinking of waiting for Raven Ridge, while also waiting for DDR4 RAM prices to drop.

>>56241500

/g/ finds a way.

>>56234651
the fuck is that supposed to show? that graph is apple-tier

>>56240960
>Intel Kaby lake is 4.5GHz!

And still only 5% faster than Skylake at best.

>>56240960
I want summer /g/ to end

>>56234985
>The Zen core arch will be big in the server market.

It won't, because any company buying Zen servers will end up getting a huge price markup on everything else they buy from Intel. So no company can switch because AMD does not offer enough products in every single category, or at the volume Intel does.

>>56241941
shows how Zen scales on a perf/watt basis compared to AMD's other architectures

everything is clearly labeled and the graph explicitly states that it's an estimation

you are apple consumer tier stupid

>>56242034
intel has ample competition as far as SSDs and networking goes. this is a non-issue.

>>56234651
lmao +40% higher IPC is much less impressive if excavator is what's being compared with

does amd have a hardware video enconder on their cpus yet?
the only thing that really keeps me on intel is quick sync.