Ryzen 8C/16T F4 3.6/4.0 Engineering Sample Over 95W At The Moment.

French here : "but they are able to pull through their 6C/12T in (barely) 65 W. Which is anyway very respectable".

>>58426496
Duck pc hardware?
Forewing stabilizer pc hardware?
I don't get this humor.

>>58427307
It's not meant to elicit anything in particular. Most likely pairing two unrelated elements for comedic effect, with the added benefit of being easy to remember.

The French love their dadaism.

>>58427401
>>58427307
Nah, it's because of "Canard WC" which is toilet detergent.
Just a bad wordplay, which french people are even more fond of.

>>58427307
>I don't get this humor.

t. Mohamed Lahouaiej Bouhlel

>>58427526
This and "canard" is also (somewhat old fashioned) slang for newspaper.
Regardless, CPC hardware were also the ones that did the first Ryzen tests.

>>58427526
merci d'avoir clarifié

>>58426496
https://forums.anandtech.com/threads/new-zen-microarchitecture-details.2465645/page-158#post-38675220

F4 OPN(B1 stepping) chips are still 95w TDP. CanardPC was saying that their A0 ES hit slightly over 95w at 3.6ghz/4ghz overclocked. They're saying how impressive it is that the chips reach those clocks at that power level.
Google translate is just confusing people.

There are non SMT enabled chips coming as well.
Plenty of SKUs available at launch.

This is absolutely incredible. Considering selling my 6600k but I need to see actual FPS numbers. I play at 144hz which is always CPU/mem frequency bound.

Can you use ddr4 4266 with the enthusiast chipsets? Shit gave me a 25fps boost in my main game

>>58427646
Unknown yet. One vendor said that DDR4000 works. All of MSI's boards say 2667mhz+, and they said that 3200mhz works with OC.

>>58427618
B1 stepping sounds like a base layer respin too. That could easily reduce power loss.

ANOTHER


MASSIVE


DISAPPOINTMENT

>>58427646
>Considering selling my 6600k


that would be a really dumb idea

>>58427845
Why?

>>58427925
AMD

>>58427646
>he thinks zen is for gaymen fags
just buy an apu and stfu faggot. the gpu market is such dogshit gaymen is the last thing anyone cares about currently.

>>58427936
Ohhh...

>>58426496
>même

What did they meme by this?

Hey.. Hey.
Is this thing actually more power efficient than Core?

>>58428109
It better fucking be because AMD does have a lot of good engineers and Intel has been sitting on its ass for 5 years.

>>58428109
It could be depending on how you look at it. Efficiency isn't a flat metric, and due to various process characteristics, not all clock scaling takes place on the same curve.

In certain instructions at say 3.5ghz Zen might equal or exceed Broadwell in efficiency. It might lose out when comparing other instructions though.
It might be an out of the park homerun below 3ghz depending on the scaling curve also.

>>58427920
Not if he wants to do something besides gayming with his machine.
Quad-cores are truly the cancer of our age.

>>58428109
We'll see when the 22+ core server chips show up, that's where you measure efficiency, and where it actually matters.

Kek, I knew it was a housefire, 8 cores at 95W made no sense at all, it's gonna be at least 130W realistically

>>58429025
Try again. See: >>58427618

>snail eater
go home everyone, party's over

>>58429055
>believing AMD's lies

>>58429094
>trusting a snail eater sub human more than AMD

>>58429094
Seems you're too caught up in your own to see reason.

>>58426496
Now give me a good consumer mobo with consumer features that can support 2 of those 16T processors.

File: french2.png (8KB, 400x400px) Image search: [Google]

8KB, 400x400px

>>58426635

Many thanks, ami.

http://semiaccurate.com/forums/showpost.php?p=279511&postcount=5188
Text for those who don't want to give S|A a page hit:

From sweclockers forum we have a hint that IPC jump over XV core is around ~55% (latest Zen ES).
Since XV is around 15% faster than PD as Per Anadtech's Carrizo generational comparison article, it follows that Zen should be around 1.8x faster than PD , core vs core (no SMT). 1C/2T should be around 31% faster than 1M/2T assuming SMT gain on Zen is 25% and CMT penalty on PD is 15%.

Since Skylake is around 60% faster than XV core at the same clock , Zen should be within ~5% or so from Skylike, IPC wise. Also it should be roughly on Broadwell level or slightly below since Broadwell is ~3% slower IPC wise than Skylake.

Interesting year ahead.

Relevant links:
http://www.sweclockers.com/forum/post/16600771
https://forums.anandtech.com/threads/zen-es-benchmark-from-french-hardware-magazine.2495505/page-25#post-38657740
http://www.anandtech.com/show/9483/intel-skylake-review-6700k-6600k-ddr4-ddr3-ipc-6th-generation/9

File: certified shit wrecker.jpg (13KB, 381x286px) Image search: [Google]

13KB, 381x286px

>>58430128
Zen+ only needs to bring about a 10% IPC uplift and marginally higher clocks at equal power and AMD will be neck and neck competitive with intel.
The Shit Wrecker did it. Its unprecedented for a brand new arch to have such a huge performance uplift in its first iteration.

Never doubt the Shit Wrecker. Hes certified.

>>58430271
well, he make the base design for the Apple A series processors, and those fuckers have been shitting on everything in sight for years.

>>58430271
>>58430306
Makes you wonder what he'll achieve at Tesla.

>>58430339
One could only speculate. As far as I know Tesla has no intention to design their own ASICs, so I'm not exactly sure what Keller would be doing.

I suppose its possible that they'll try to move away from off the shelf hardware to create a more specialized integrated system, but the financial investment would be immense.

>>58427401
>dadaism

I just got flashbacks from my debater roommate and his marxist textbooks and tendencies.

fucking hell

>>58427646
>>58427670
an MSI booth dude told some youtuber (I think it was paul's hardware) that the boards were 4000mhz OC (can't remember what the chip was, but iirc it was a 2666mhz).

I don't think those OC memories on the mobo's ever presented a problem, I've never had a kit that reached the rated OC limitation but I also never heard anyone not being able to reach the stated upper thresholds; on premium boards that would definentely become news worthy...

>>58430816
if not, then i'm fucked.
I got a 3600mhz corsair lpx kit following an anon advice to get memory sooner rather than later in account of price hikes.

>>58427618
>non SMT enabled chips coming
just_jew_my_shit_up.gif

>>58428109
>AMD besting anything
listen to yourself

>>58430271
>. Its unprecedented for a brand new arch to have such a huge performance uplift in its first iteration.

no it isn't, he's done it twice before.

>>58426496
The title is shit... What is this thread about?

>>58430128
so it's basically shit?

they already fucked up by not releasing it after ces along with vega, I was wainting for zen to build a new pc, but I can't wait anymore. already ordered skylake and a gtx.

>>58432583
It literally says that zen is meme.

>>58432594
Good goy.

>>58432642
nice try poojeet
you're late on the market as usual

>>58432642
even if it's better, they fucked up by not releasing it after ces.

>>58432648
Ho я жe cлaвщит. Neck yourself, newshit.

>>58432650
You want another Phenom with bug in final silicon?

>>58430845
I'd do the same,but I have 4ghz modules hope come in time

>>58432660
I want to upgrade. amd apparently, don't want my money

>>58432679
Buy Intel then. Who gives a shit.

>>58430271
>you will never be a demi-god of computing architecture and engineering

I feel like a fucking loser when I see his accomplishments.

>>58432726
Same with Raja. AMD has quality engineers. They just need half decent marketing team.

>>58432740
that reminds me when nvidia marketing top man almost left nvidia for amd
but was pulled back IN, and not a peep form here ever again, leather jacked man dealt with him appropriately in chinese mafia traditions

>>58432594
>so it's basically shit?
With this high of a jump in performance, you'd literally have to be an Intel shill to not congratulate AMD on their new design.

They closed the gap, in one generation, from miles apart to meters.

>>58430128
how far is skylake from broadwell? 7-9%?
5% from skylake for zen
basically it is slightly faster than broadwell but at the same time slightly slower single core than skylake making it best 8 core at a time?

how is this bad?

>>58432681
>Who gives a shit
amd

>>58432759
but it's not faster and not as power efficient.

>>58432813
Skymeme is like 2% better compared to Haswell. It's Haswell that was better than Ivy by like 12%.

>>58432827
They don't, you're a single consumer.

>>58432834
all the facts are against your theory though

>>58432847
>implying
amd fucked up again, just accept it.

>>58432834
It may be faster and it's already as power efficient as current Intel lineup.

>>58432858
>facts
what facts, it's all just theory and rumours from controlled benchmarks.

Canard PC is the Charlie Hebdo of hardware.

>>58432861
(you).

>>58432868
>controlled
Meme.

>>58432834
Expecting it to be faster is absurd. Intel's had 8 years to refine and hone their architecture to an ultrafine point.

It is very very close to being as power efficient, if not as power efficient. 8 cores at 95W at 3.6Ghz/4.0Ghz, when they were expecting 10W per core at 3.15Ghz and 16W per core at 3.5Ghz.

Instead, they're at ~12W per core at 3.6Ghz, and only pushing higher the closer to their release they get.

>>58427646
>Can you use ddr4 4266 with the enthusiast chipsets? Shit gave me a 25fps boost in my main game
Bullshit.

>>58433117
Depends on the game.

>>58433125
Nope.

>>58433117
>>58433136
very very fast DDR4 can give you much lower latencies overall which can give your CPU a small to decent performance increase.

Also faster access to anything cached in system memory vs VRAM is always a performance boost.

>>58433148
It wont give you 25 FPS in any kind of video game.

Still falling for the MORE CORES meme like in 2011. Some things never change

>>58433200
But Intel is moar cores company now.

>>58433173
ARMA

>>58433249
Who the fuck still plays Arma?

>>58433209
Only on really high end chips used for workstations. The consumer products are limited at 4c/8t

>>58433270
Not for long. Coffee Lake is going for moar cores to be competetive.

>>58433249
Source.

>>58432945
sense sandybridge, all intel has done was sat on their dick and stop soldering cpus.

amd in one fell swoop apparently not only caught up ipc wise, but also surpassed them in efficiency.

it may not clock as high as an i7, but to the enterprise that doesn't matter, what matters is powerdraw, and considering amd looks to have slightly less power use then a 6900 while matching it... they got that.

intel is about to be caught with their pants down jerking it into the tmi vat.

>>58432726
>>58432740

1) dont compare yourself to the shitwecker, we all come up wanting.

2) there is a difference between keller and raja.
keller gets shit done, while raja has the most fantastic ideas on paper you ever seen, but the actual product is left wanting.

keller put a team together who can help each other and improve, raja has a team that doesn't think of drivers or software implementation.

I mean, gpus are probably harder then cpus to make, and make well, I mean we have a standard called direct x, but apparently every single fucking game needs to have work done on the gpu side to fucking work? the cpu side sure as fuck doesn't work this way. but still... seriously, even if vega is amazing, the showing they are giving currently are not what we expected, not to mention how pissed we were when the countdown was to fucking 5 3 minute youtube videos.

In all honesty amd should have stuck with mantle and not fucking gave it to kronos, followed through with the beta test on their hardware, and then opening it up to everyone like was the original goal. at least more than one fucking game used mantle and every game that used it saw a performance uptick over dx11, not this dx12 shit where it goes backwards or vulcan where there is only 1 game that really implemented it.

for fuck sake they drop the ball do fucking hard and often you think it was soap, in prison, and they were the biggest faggot that ever lived.

>>58432813
>>58432839
from haswell to broadwell was 1% at minimum to 4% at maximum

from bradwell to skylake was negative, as in they lost 2.5% ipc, to positive 4% at best, we leave out the dolphin benchmark as it somehow has generations of intel where it gained nearly 40% ipc.

skylake is not some magical 'so much better the broadwell' jump that everyone seems to think it is.

>>58433991

It already clocks higher than 8-core i7.

>>58434124
yes and no.
96% of intel's 6900k hit 4.2ghz across all cores

stock, yea, its already better, but that's not saying it can hit higher clocks necessarily.

it is promising though, ill give it that.

>>58434070
Countdown was to arch preview, and we got preview, it was misleading tho and AMD needs to gas it's marketing team. Again. Raja is good boy tho.

>>58433173
fallout 4
that game is retarded in how it uses the cpu for rendering shadows, and fps is HEAVILY determined by ram

>>58434146
he is ideas, what he fucking needs is someone to sit there and as him 'will this make the gpu faster now'

look at fucking gcn and tell me that the way it was wasn't a massive mistake? the 7000 line was good ill give them that, but for fuck sake, its taken them 5 fucking years and 4 iterations to get the shit to not be its own fucking bottleneck, and even that's is assuming vega doesnt again bottleneck itself somewhere.

>>58434070
Countdown was to arch preview, and we got preview, it was misleading tho and AMD needs to gas it's marketing team. Again. Raja is good boy tho.

File: are_you_fucking_kidding_me_by_rober_raik-d4clstk.png (46KB, 1055x757px) Image search: [Google]

46KB, 1055x757px

>consumer
>multi socket

>>58434355
use to not be that uncommon.

>>58430339
Well he's proven himself to be multi-talented. He played a large role in creating the 64bit adaptation of intel's x86 instruction set when he worked at AMD.

It was actually hilarious timing because the pentium 4 was such a massive piece of shit with negative gains while AMD was rolling out AMD64 which was better in every single way.

So yeah he can probably pull magic out of his ass however he seems fit.

>>58427646
But 6600 is quite good sempai, and that's coming from an amd fanboy.

>>58434070
Cpu's don't require specific optimizations in the same way as GPU's do because they all use the x86 instruction set. But it doesn't mean they shouldn't get more attention.

For example, one of the most appealing things about mantel, and subsequently vulkan, is better utilization of cores. We've had dual core chips for ages and quad cores almost as long. Yet DX and openGL stayed the course with 80% of the load on core 1, only haphazardly slapping on "multi core support" the same way api revisions have basically just had feature supports and specs slapped on since xp.

And a lot of focus since the early days has been shifting as much load to the gpu as possible because cpu's are so shit at anything beyond basic vertices.

AMD DID put mantle out for free, but devs didn't bite because they didn't want to alienate nvidia who flat out said they would never support it. And nvidia happened to own like 80%+ of the market at the time.

Having the openGL guys pick it up is the best thing that ever could have happened. The ps4/pro uses gcn and can't use dx12 so we may see more console ports with vulkan support.

And don't harp on dx12 too much. The examples we've seen were shoddy attempts at forcing dx12 support into dx11 titles very late into development. Which was possible before, but dx12 is completely new and needs to be considered early on.

>>58428181
then he would be really dumb, if he didn't buy any i7 if he uses his cpu to work

File: 1483629357673.png (181KB, 293x397px) Image search: [Google]

181KB, 293x397px

>>58429126
that's what amdcucks believe

>>58434181
Are you referring to the fury? Because that's the only time you could say CGN "bottlenecked itself."

And to the fiji's credit, had it not been the biggest possible die on an end of life node well past the point of being worth sinking more money and time into, a hardware revision probably could've fixed the pipeline saturation issues.

GCN as an architecture has stayed relevant for a good 5 years now without any major overhauls to the underlying framework. That's pretty badass, especially with how terrible general optimization has been over the past decade or so thanks to lazy fucking console devs.

>>58426496
>même
Ryzen confirmed for a meme

>>58429126
I'd rather trust a snail eater than street shitter.

File: 1481703908447.png (203KB, 500x420px) Image search: [Google]

203KB, 500x420px

>>58430271

Give me a good reason why I should upgrade to Zen from my Haswell i5

>>58430271
If only amd gpu guy as based as him, nope just get poo in a loo

File: the-truth-behind-datguylirik_o_4921903.jpg (55KB, 1023x387px) Image search: [Google]

55KB, 1023x387px

>>58433264
Lirik XDDDDDDD

>>58435199
8 cores you dipshit

>>58434844
1, if amd is to be believed, the devs went to amd to ask for lower level apis
2, i believe until dx12, it was the largest uptake an api had and was still getting more games when they gave it up
3, we have 3 vulkan games, with one planned to add support, and that's it.

If amd held mantle and did what they planned, beta it on their gpus, then open it, intel would have taken it up, and despite what anyone here says, intel is more relevant than both amd and nvidia as far as gpus for games goes, kind of a sad fact.

Another spectacular failure, 4.0 at 95W while Intel gets 4.5 at 91W

>>58435385
>Another spectacular failure, 4.0 at 95W while Intel gets 4.5 at 91W
>8c/16t at 95W
>4c/8t at 91w
yeah what a failure by intel

>>58434949
lets see here, the 900 series and all it brought to the table, on top of the die shrink now allowed full desktop versions to be on mobile platforms, meanwhile amd had.... from the 200 to the 400 series 9 laptop models, while nvidia had close to 1000, amd did have the apu, but that's just bare minimum to run the fucking os for the most part.

i'll give amd the 200 series because the 900 was not out yet, once that happened amd had the the 285, the 300, the fury, and polaris to come up with an answer, and they failed.

fucking amd drivers bottle neck the gpu
the way the card itself works bottlenecks it, or at the very least make it run less efficiently, for fuck sake pure number crunching capability, the 290x should match, and more likely because of how overclocking works on amd and nvidia crush a 980ti, but the card can struggle to consistently outdo a 980 and back when shit was newer, keep up with a 970

you can tell me the card isn't bottlenecking itself, but something is clearly fucked in either the way the card works or the fucking driver.

and really, why hasn't amd scrapped the driver, and just redone it yet? I mean they sure as fuck cant put a new one together that is actually worse then what they have right?

>>58435472
One core at 4.5 uses more power than 3 cores at 4.0, Intel could fit 10 cores in 95W with those small clcocks


It's another failure by AMD, you're only gettging foolded by hype and big numbers

>>58427618
anyone able to confirm that through something amd said, non smt versions of chips?

>>58435500
why is intels 10c/20t at 140W then?

>>58435522
Broadwell, Kaby Lake is much more efficient..
Also there's a lot more cache, quad core DDR4, much bigger chipset, multi socket buses, AVX512 etc

>>58427969
même means same in French.

>>58435540
>Kaby Lake is much more efficient..
it draws more power than skylake at same clock

>>58435589
Wow, aren't we butthurt?

>>58435346
Bulldozer had more cores as well and it was shit, I ain't falling for that meme

>>58435610
>he doesn't know the difference between a long pipeline CMT design vs a short pipeline SMT

That's why nobody asks your opinion on technology.

>>58435346
MOAR CORES doesn't matter. my i5 has never been overloaded unless I ran prime95

>>58435599
You really showed him

>>58435540
what are you talking about? Kaby Lake won't get more than 4 cores. Also Skylake-X will have less cache than Broadwell-E

>>58434858
>any i7
Not any, 6700K is only 4C/8T.

>>58435622
You obviously don't do anything substantial with it. I have a 6600k and its just not enough.

>>58435346
But all the games I play are shitty and unoptimized and use like 1 core for 90% of the load.

I will never buy a low single-core performance CPU.

>>58435704
Fuck off back to /v if all you care about is games.

>>58435748
no

File: 1478857869799.jpg (47KB, 587x600px) Image search: [Google]

47KB, 587x600px

>>58435704
>But all the games I play
Then nobody give a fuck about you, newest games use all 8 cores - Battlefield 1, Deus Ex Mankind Divided, Forza Horizon 3, Gears of War 4, Rise of The Tomb Raider, Fallout 4, Witcher 3, Gta V, even older unique games like Planetside 2 scale well with 8 core.

So i guess you only play trash weeb visual novels so yeah your pentium should be enough.

>>58435784
oddly enough Intel still rapes AMD even on those games

shows how much of a joke AMD is

>>58430816
>>58430845

The OC with RAM ratings isn't referring to overclocking the RAM chips, rather it's referring to overclocking the memory controller on the CPU/Northbridge

>>58435784
>newest AAA garbage use all 8 cores

ok chad

also you're gonna have to source those claims kid, because i don't believe all those games actually use 8 cores.

>>58435784
arma tho

>>58435480
But the 1050/1060 have a different core count from their desktop counterpart

Woops you guess it's another 3.5 meme you fell for.

>Actually believing Nvidia

Oh I can't fucking breath

>>58435704
Everything is pointing to very decent single core performance. With 8 cores I could run Windows in a Vm with 4 cores, have muh linux running on 2 on my second screen, run pfsense router on 1 core, and have 1 core running plex/kodi/seedbox/whatever all at the same time. Can't do that with 4 cores, friend. Not even going to mention the ability to encode muh chinktoons twice as fast.

>>58435822
because all amd has out right now is shit that has 8 threads, and when a game uses all 8 threads shit works about as good as an i7 till haswell generation.

>>58435704
you don't need the best single core performance, you need good enough. we are already moving away from single core engines, and even apis that further expand the departure from single core reliance.

>>58435868
mother fucker don't make me have to defend nvidia... the 1050, 1050ti, 1060, and 1080 are their desktop counterparts...

funny enough, the 1070 is the one that's gimped.

now the better part, the fuck do I care? prior generations were half the gpus their desktop equivalents were, this one you pretty much have the full desktop chip in everything but... oh wait, I got that wrong, the 1070 is actually stronger on the mobile version then the desktop.

look when nvidia does a good, and they clearly did a good here, have to give it to them.

>>58435961
Not trying to argue, but both 1060s are clearly gimped

>>58430271

why do they call jim keller the shit wrecker, somebody put me on to this meme

File: 1483974609716.jpg (540KB, 1100x1002px) Image search: [Google]

540KB, 1100x1002px

>>58436006
>but both 1060s are clearly gimped
AYYYMD shilling as always, gtx 1060 6gb is superb card much better than your poo rx 480

>>58436115

this is like your 3rd post shilling for nvidia, jesus christ man they don't even pay you to be this retarded

>>58436115
The difference is less than 5% worst case scenario.
1060: no sli, 3gb, less shaders, clearly gimped.

>>58436167
He was saying that 6gb was gimped, yeah i know 3gb is trash. But 6gb was not gimped except in meme sli which would be garbage anyway so who cares about sli.

>6c12t
>65w
Holy shit. If they can field that, this is revolutionary. It would fuck up Intel's "energy efficiency" garbage meme..

>>58435480
You're right. The 290x is a number crunching machine. That's why they're so good at async compute.

But that was a forward looking measure, not unlike the revolutionary implementation of a hardware t&l engine in the geforce 256. Or like when ATI implemented their "quads" strategy with the R400 chipset, essentially laying the groundwork for the way we still make GPU's today.

Unfortunately, like the Truform Ati tried to push way back in the day, it was too forward looking to gain traction. Async compute is a very un-GPU like thing for a GPU to do, and devs would have to have created the game with it in mind to make the most of it. Drivers can help to a degree, but there's only so much you can do.

And AMD's drivers have been pretty great these past 2-3 years or so. The rx 480 was a bad example, but from the 200 series up till then, they've been really consistent in having drivers for new releases and with adding more features and improvements.

You're also overstating mantel's relevance. It was only in a few big titles, one of which was a forgettable tomb raider reboot and another was a generic fps whose only selling point was a gimmicky xray headshot thing.

>>58436115
The 1060 is like 1% faster lmao

>>58435697
Well, what do you do with it?

in my wildest dreams i barely dared to believe zen would match haswell in overall efficiency. if vega turns out to be decent then i might even just build a whole new anti-goy rig.

>>58426496
I can't wait for the 9590 of this generation where they push the arch to its limits with massive power draw. imagine a 6ghz 8-core at 160w

>>58436371
>anti-goy rig
Ok, Pajeet Tsung

>>58436403
no worries incel yid :^)

>>58436433
I don't worry, premium Nvidia and Intel hardware will always be supreme

>>58434694
If he can get a 6c12t CPU for around $300-400 and slightly lower IPC, then why not?

>>58436016
Its just a joke that I started a while back and it caught on.
Keller has a long list of accomplishments, so when singled out he looks like something of a superstar in the semiconductor world.

File: perfect_01.png (265KB, 501x382px) Image search: [Google]

265KB, 501x382px

>>58435540
>quad core DDR4

File: 2017-01-09_2318247ukfb.jpg (534KB, 1920x1080px) Image search: [Google]

534KB, 1920x1080px

Some info on AMD's new Infinite Fabric go released, but no major sites reported on it.

File: 2017-01-09_2318178sk5w.jpg (378KB, 1920x1080px) Image search: [Google]

378KB, 1920x1080px

>>58439212

File: 2017-01-09_231822rdkz6.jpg (412KB, 1920x1080px) Image search: [Google]

412KB, 1920x1080px

>>58439223

>>58439212
>>58439223
>>58439240
>buzzwords
>buzzwords
>buzzwords

>>58439240
>multi-socket
>multi-die

>MULTI-DIE

OH IT'S HAPPENING

>>58439715
You must be over the age of 18 to post on this site.

File: Multi_functional_extension_socket_634591664681364674_2.jpg (57KB, 1024x729px) Image search: [Google]

57KB, 1024x729px

>>58439783
>multi-socket


ITS HAPPENING!!!

FINALY QUAD CHIP MOBO'S 24 CORES 48 THREADS

UNLIMITED MINECRAFT BLOCKBREAKING POWER!!!!!

>>58439956
Come on, at least try to be funny.

>>58439783
Its the fabric used for scaling MCMs, namely their server Opterons. The Naples platform was shown off in a 2P configuration, 64c/128t.
I doubt any vendor is going to sell a consumer dual socket AM4 board.

File: AMDTest_1.jpg (1000KB, 2697x1530px) Image search: [Google]

1000KB, 2697x1530px

>>58439981
they did this once
they will do it again

>>58439240
Think Nvidia has a multi die fabric? NVLink doesn't seem like it.

>>58439212
>>58439223
>>58439240
where are the slides, faggot?

I need more info on inter-die interconnects, since I care a fuckload more about the potential of Naples than about Summit Ridge or Raven Ridge.

>>58434355
i might upgrade from 2500k if that is true

File: AMD-RYZEN-ZEN-CPU_Market.jpg (393KB, 3999x2250px) Image search: [Google]

393KB, 3999x2250px

>>58440079

File: AMD-Zen_Naples-Server-SOC-840x280.png (175KB, 840x280px) Image search: [Google]

175KB, 840x280px

>>58440101

File: AMD-Naples-Zen-Platform_5.jpg (2MB, 2728x1816px) Image search: [Google]

2MB, 2728x1816px

Rest of the slides here

http://wccftech.com/amd-zen-naples-server-cpu-vega-gpu-platform/

>>58440079
> 8 channels

my dick is diamonds.

Is there any info on 2 die/4 channel configs?
16 core at ~3 GHz seems like it would be great for a workstation.

>>58440131
Holy cow look at the size of those chips.

>>58426496
TWO FUCKING MONTHS?!?

TWO MOTHERFUCKING MONTHS????

GO FUCK YOURSELF, AMD

motherfucking street shitters keep kicking the can everytime they have to launch shit

>>58440148
The dual die platform is called Snowy Owl.
Its BGA, like intel's Xeon-D, but this scales to a far wider range of TDPs.

>>58440152
Those look bigger than 4 Zeppelin dies would really require.

There's either extra spacing to aid cooling, or they've padded the dimensions to support HBM-using future enterprise APUs.

>>58440152
Just you wait when Intel starts using MCMs again, they'll have 96 cores

So if I wanted to buy a Ryzen CPU, I guess I would have to buy a new mobo right?

I'm using an 4690k at the moment, which is LGA 1150 I think? Is Intel and AMD still doing the meme where they separate the sockets so that you have to change motherboard every time you want to switch over from the other manufacturer?

>>58440131
thats a big cpu. looks like 2 maybe even 3 intel cpus.

File: What the fuck are you talking about.png (938KB, 868x578px) Image search: [Google]

938KB, 868x578px

>>58440193
Is this 1999?

>>58440193
no you can use your intel mainboard for ther ryzen cpu but you will have to remove some of the amd specific pins that dont work with intel mainboards

File: seen1.jpg (65KB, 500x519px) Image search: [Google]

65KB, 500x519px

I am buying Kaby q7700

>>58440290
Cool..?

>>58440209
Is that a no?

>>58440222
Please no memes, that sounds like it would only break the mobo

>>58440131
Reminder that AMD is finished, it needs 4 chips to compete with one Intel chip.

>>58440193
it's well enough that they both get to make x86_64 instruction set chips, despite all the licensing hurdles that brings.

using the same socket is a bit of a far stretch. especially since the platform on top of which the CPU runs is not only a massive R&D endeavor in of itself, but it's also deeply tied to the CPU's architecture itself. forcing both companies to use the same socket would stifle innovation and we'd probably still be stuck with north+south bridge motherboard designs.

>>58440493
Reminder that Intel is finished, it's letting its competitor with a 5 year gap in architecture and 2 year gap in fab catch up to them.

>>58440493
Reminder that AMD doesn't have to deal with the horrendous yields of a single massive monolithic die like intel does, so profit margins on their enterprise parts will be even higher.

Intel's E7 8890v4 with 24c/48t can't even compete against AMD's 32c/64t Naples Opteron.

>>58440563
Intel has a 32 core too

>>58426496

I'm making a gming computer

I was planning on buying the cheapest intel 7 processor but my friend said to wait until this was released.

Am I wasting my time by waiting? I don't know shit, has this been confirmed to be worse than that or are we still in the waiting phase?

How much longer do I have to wait?

>>58440594
Which they announced right after Zen's IPC was leaked, followed by HT pentiums and unlocked i3.

Desperation at its finest, modern Intel isn't used to competition, these are just bandaids until their lawyers with start visiting OEMs with unsuspicous suitcases full of 'partner gifts'

>>58440594
They're planning to have 32c SKUs with Skylake Xeons. Looking at the clocks of the Broadwell Xeons, 2.2ghz for 24c, I don't expect anything better from a 32c part on the same troubled process.

The top end Skylake chip will probably be clocked lower and have higher power consumption.

From Arse Technica
How AMD's turbo core and XFR work in Zen:

"A lot of processors have pre-programmed clock speed voltage tables," explained AMD's Robert Hallock at CES. "We don't. This is very algorithmic. We analyse power consumption limits, thermal limits, silicon utilisation limits, and out of that boundary, if none of those limits are being met, you can just keep raising clock speed until one of them is. Then you level off the boost and then try to sustain it as long as possible. The system is smart enough to know what's going on inside itself, and adaptive enough to prevent sudden drops in clock speeds."

>>58440801
This isn't all that much different from Nvidias, who do state a 'boost clock' but the chip can go over that at times.
Intel does this too, but only on Broadwell-E since it's the only one with Turbo 3.0 IIRC.

>>58440829
It would be very different if it ends up that a 3.6 base 8 core Zen can turbo a single core to 4.5 in serial workloads.

Intel doesn't have an answer to that.

>>58440563
>AMD's 32c/64t Naples Opteron
Which I desire with all my heart.

Too bad it's going to be $4k or something similarily obscene.

>>58426496

IT'S OVER AYYYMD IS FI- oh, wait...

>>58440650
50 days according to the latest rumors

>>58440878
Thats wishful thinking, the chips with the F4 OPN have a 4ghz boost, and thats single core. An extra 500mhz isn't going to get squeezed out.
Maybe they'll get there on 7nm, maybe even higher.

>>58440879
$4,000 would be a bargain compared to intel's $7,000~

>>58440878
That would be amazing for normalfaggots who can't overclock and enterprise folks who don't overclock, but for anyone here not OCing a 8 core CPU would be a waste unless it's a workstation given to you by work.

>>58440908
From what I remember the frog Zen ES that was tested had a 3.1base and 3.3 boost but they said it boosted to 3.5 on a single core.

>>58440935
It was 3.15ghz base, the listed turbo was supposed to be 3.5ghz, but it never went over 3.4ghz in any circumstance. That was A0 silicon, the very first lot of engineering samples.

>>58440908
>$4,000 would be a bargain compared to intel's $7,000~

Well, there is always hope they are going to drop off servers one day. Just like you can now get G34 opterons for peanuts.

>>58440650
It will be very close to Intel and it will bring down prices if you don't like the performance. A couple weeks at most.

>>58430271

ABSOLUTE PUSSY DESTROYER

>>58440905

fuark this thing better be fucking good, I already waited December because my friend said it was coming out in January.

>>58440951

Will the prices respond that fast though if it is close?

>>58440951
It's nice to dream but even when AMD was demolishing Intel with K8 Intel never lowered its prices.

>>58440967
>Will the prices respond that fast though if it is close?

It already did. Intel is shifting its entire portfolio down. Pentiums are becoming I3s, I5s will drop to I3 and so on. In a year entire Intel i7 range is going to be 6+ core.

Prices will also drop if Ryzen delivers. If you are on the verge of buying a PC, wait a bit. Unlike usually, shit is going to be stirred hardcore.

>>58440967
They probably would have had wide retail availability a bit sooner if they didn't keep pushing clocks higher. AMD's presentation for GDC this year is titled:

>Join AMD Game Engineering team members for an introduction to the recently-launched AMD Ryzen CPU
>recently launched

So they'll be available before February is over, and AMD confirmed there is no paper launch. When they're released they'll be on sale everywhere.

>>58441015
Is Coffee Lake (6c) still gonna be an APU, and if so, what exactly can AMD do to compete in that market?
Raven Ridge will just be 4c + Vega, so AMD better hope that the GPU side is at least decent.

>>58441044
Coffee Lake is middle of 2018

>>58441044
>Is Coffee Lake (6c) still gonna be an APU,
Yep, they're still pushing big IGP.
AMD's Raven Ridge will probably still have the advantage in IGP performance, intel will hold on to the advantage in CPU performance.

Its funny that intel is literally the more cores company now.

http://arstechnica.com/gadgets/2017/01/amd-ryzen-motherboards-hype/

At CES, the company [AMD] openly asked journalists what it should be priced at. Price an eight-core Ryzen at around £400/$500 and you bring down the cost of eight-core chips to that of a six-core—a solid move for consumers, but not one that will greatly affect mainstream performance. Price it the same as a quad-core i7-7700K—about £300/$330—and you dramatically shake up the industry. The decision is yet to be made.

"There are a lot of discussions going on," says Hallock. "We're capturing the feedback. We wanna take share, we wanna be the best price/performance option, we wanna be the first on people's minds. That's part of the bounding box for pricing discussions as well as paying off the R&D investment... We're looking at what Intel does—and we're not gonna do that. We think people want the choice, and need the choice. The market needs the choice—hopefully we can turn it around."

Get hype niggers

>>58441142
>£
>400

that is too much.

>>58441181
In case you don't remember, your currency is crashing.
It should acutally be higher than 400 to meet the 500$ target.

>>58441181
Then get the poorfag 6 core.

>>58431202
Oy vey 0.02 shekels have been deposited into your account

>>58429126
Snails are everywhere in french gastronomy :^)

>>58440950
shit you can get most intel chips used for 1/10th the price on ebay (used)

I got a few 2670's for $50 when the new generation came out

still $125 on there.. I think it's 1.5k retail

File: Certified Shit Wrecker.jpg (421KB, 1680x1050px) Image search: [Google]

421KB, 1680x1050px

>>58430271

>implying 65W TDP means 65 watts
AMD always lies, Poolaris was 150W TDP and it uses almost 200

>>58442026
very perceptiv of you to see that TDP is not Watt! what gave it away? the missing W?

File: 02-Overview-Gaming.png (24KB, 600x450px) Image search: [Google]

24KB, 600x450px

>>58442026
Oh look, a little shitposting /v/irgin retard stopped by.
The RX 480's die draws 110w, the VRAM draws 40w~. Only the reference cards on release were pulling slightly more power, totaling about 165w instead of 150w for the total board.

>>58442052
AMD does treat TDP as typical board power.

>>58441911
I already do. I am on 2011-3 and hope to catch one of those 16 core+ xeons one day. But 32/64 - goddamn.

>>58442026
Polaris is mediocre overall and poor on the efficiency front, but not nearly that bad.

If you count ms-scale transient levels, every single card from any manufacturer has fluctuations from the VRMs.

>>58436167
sli and multigpu is, has been and always will be a shit solution. in nvidias bid to made sure people dont get 2 1060's instead of 1 1080, they likely saved a fuckload of people a lot of strife.

>>58436115
no the 1060 is not clearly better, the 6gb is at best equal to the 480, and in the next year, likely the worse option as texture sizes are set to bloat now that the minimum mid range gpu, aside from the 1060, is now 8gb.

the 3gb variant is inexcusable though.

>>58436006
the 3gb card is and there is no way to excuse nvidia on that one, but the normal one isnt. may not have sli but that is shit regardless.

>>58436280
1% faster right now, 2gb less ram, for former will change in 3 or so months time, the latter will hurt you for many years to come.

>>58436016
He is a one man team building/planning machine. a co author of the x86-64 and lead designer on many teams that built cpus that clock for clock kick intels ass at best, all its competition's asses at worst.

the fact he does it so often in any project he is in, got him the name shit wrecker, and how consistently it happens gave him the certification.

someone should print him off a diploma like thing when zen comes out, just so he has a physically trophy like thing to tell him he is know as the shit wrecker and the internet certified him.

>>58442205
It will be interesting to see how Vega's "High Bandwidth Cache" deals with memory pressure.

It's honestly hard to tell if it's
> A: AMD foresaw small-ish HBM capacities 2-3 years ago and build a solid solution to transparent virtualization
> B: a hack from ~1 year ago to deal with poor HBM2 yields
> C: complete marketing bullshit for what will be mostly a driver-managed band-aid

In the case of A or maybe B, Vega might do OK with 8 GB SKUs. Otherwise, AMD's pretty fucked.

>>58440002
dont overlook it, unless amd makes the consumer chips not able to do multi cpu, vendors WILL do it at some point.

granted, software still doesn't work properly with multi cpus, so it would largely be used as a '2 computers in one box' kind of option, you know there are a lot of uses for just using a secondary computer to render shit while you continue to do things on your main one, well here is a motherboard that likely reduces the cost over all of the need for 2 systems.

it will be touted as a workstation thing though. but a workstation without the intel tax... god i cant wait.

>>58426496
it's over.
intel is shit wrecked.

>>58440189
>Just you wait when Intel starts using MCMs again, they'll have 96 cores

let the core war begin

>>58442407
>software still doesn't work properly with multi cpus
What a retard.

>>58427618
turning off smt is pretty dumb other than jewing out the consumers.

>>58442486
SMT isn't a free feature, SMT disabled lowers voltage and increases OC room

File: [HorribleSubs] New Game! - 09 [720p].mkv_snapshot_23.22_[2016.08.30_05.04.54].jpg (120KB, 1280x720px) Image search: [Google]

120KB, 1280x720px

>>58427947
waiting for zen+ apu with hbm.

>>58442486
>I don't understand how SMT is actually implemented
If there is a defect in the registers needed(the highest instance of defect rate due to their size and variability) for that second logical thread SMT can be disabled, and you still have a fully functional core. There is no reason to throw away an otherwise working die.

Don't make retarded assumption, you gamer child.

>>58441044
>>58441077
amd does not need to compete with intel's 6 core i7, not really at least. as amd likely has their bottom 8 core 16 thread at an i7 price range, if you are in for a high end cpu, you are doing something that benefits from a higher end gpu also, so really, intel is just wasting die space on i7's really only the i3 line and below should have gpus on them, should move the igpu off to a motherboard option instead.

>>58429126
CanardPC are serious people.

File: 1409603110438.gif (3MB, 300x237px) Image search: [Google]

3MB, 300x237px

>>58432813
Haswell was pretty much the end of the line for Intel's current design. The only real improvements since then have been due to IGP improvements and faster clock speeds. Otherwise you're looking at like no more than 3% better IPC on average, despite the fact that Skylake was a "new architecture". Coffee Lake is a Kaby Lake refresh and Cannonlake is just a die shrink, so 6 generations of Intel CPUs will have basically the same IPC.

We all know what happened the last time Intel resorted to chasing clock speeds for performance boosts (Pentium 4)...

>mfw AMD is about to fuck their shit up

>>58442505
in the grand scheme, SMT is about as close to free as a feature of its level can possibly be.

90%+ of the work comes from having a robust register renamer in your pipeline. Splitting the other buffers is not nearly as rough.

>>58442538
you are a retard who has no idea how SMT is implemented.

>>58442574
Not to mention that AMD will unlock all of their Summit Ridge CPUs for overclocking and solder their heatsinks so they don't start house fires like Intel's shitty chips.

>>58442574
In reality, latest Intel CPU's are just reworked Pentium 3's. Same architecture with shit bolted on, die shrinked and with power consumption improvements.
All their other architectures were failures.

AMD actually has a better track record in that regard.
Early AMD cloned Intel chips, then started doing their own thing and then released K7 and K8 devastating Intel. They then had an horrible new architecure that almost killed them, and now an even newer architecture.

Intel has perfected their architecture to unbelievable levels, but sooner or later they need to start developing something new.
Also, Cannonlake is a massive failure for them, it should have arriven what, four years ago?

>>58440993
no, they just forced oems to not use it, something Im not sure intel could get away with again.

>>58442610
>I'll just pull things out of my ass since I don't understand defect binning
This segmentation is not artificial.
Stay in /v/, you dipshit gamer child.

PRICES WHEN?

>>58442642
>solder the heatsinks
Not confirmed.

File: 850px-skylake_block_diagram.svg.png (113KB, 850x1075px) Image search: [Google]

113KB, 850x1075px

>>58442654
lmao, do you actually think that SMT cores have multiple, independent register files?

>>58442669
Release is meant to be in February so in a month maybe.

>>58442680
I think an AMD spokesperson mentioned it in a presentation a while back.

>>58442712
>Release is meant to be in February
Source?
Inb4 wccftech.

>>58440916
not necessarily.
Imagine if each motherboard has its own safe limits pre figured out, the chip itself knows where it can and can't do, and all you have to do is slot it in and its auto overclocked assuming you can deal with thermals.

this is a massive win for everyone who doesn't overclock, which is 99% of the computer market+

those that do overclock, it will be interesting how far amd pushes this and if they do make it how we envision it, how much more overclockers can get out of a cpu. because what if it was only an extra 200mhz?

what if amd sees only 1-4 cores are being used, and hits the high 4ghz range, doing nothing at all on the users end.

can not wait to see what they do with it.
im expecting just boost clocks with more granularity, but the way they talk about it has me thinking bigger.

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

482KB, 3999x2250px

>>58442654
>>58442710
SMT is a technique revolving around tagging and mostly dynamic resource partitioning.

>>58442642
>>58442680
http://www.bitsandchips.it/52-english-news/7784-our-leak-confirmed-am4-cpus-will-have-th

Godavari and Bristol Ridge have a soldered IHS. Credible leak says Ryzen does too.

>>58442710
>proving you're a little shitbrained /v/irgin kid
They size of the register files are large enough to accommodate two threads. If they're lacking then they will not pass validation for running SMT, the exact same reason why intel's i5s exist. If they could sell every one of those dies as a $350 i7 they would. Defects do not allow them.
It is not artificial segmentation. Not every otherwise functional die can have SMT enabled.

Stay. in. /v/.

>>58442727
See: >>58441023

http://schedule.gdconf.com/session/optimizing-for-amd-ryzen-cpu-presented-by-amd

>>58442727
Not 100% but AMD have said that "Q1 usually means the end of March, but not this time", plus there's some Jim Keller presentation in early March whose description mentions "recently released Ryzen".

>>58442760
>>58442771
Cool, thanks.
Wish vega would arrive around the same time.

>>58442110
it has fantastic efficiency, consumers never got the higher binned chips, and likely even the xfx gtr chips aren't the highest binned ones.

it makes me worry about vega and what gamers will get there, but good overall at least for me because this means oem/enterprise bought amd enough that high binned never touched gamers.

>>58442760
>They size of the register files are large enough to accommodate two threads. If they're lacking then they will not pass validation for running SMT, the exact same reason why intel's i5s exist. If they could sell every one of those dies as a $350 i7 they would. Defects do not allow them.
Not true, they need to ensure they cater for all markets in which they can profit. Even if every chip had functioning SMT there would still be i5s, otherwise they know customers who might've been willing to pay for an i5-level product but not for an i7-level product would just go for i3s or not bother at all, which harms their profits. It's just a happy coincidence that chips with defective SMT implementations can still work perfectly as i5s. AMD used to do it all the time with 2 and 3 core Athlon/Phenom IIs as well.

>>58442803
GTR owner here. 69 asic on mine

>>58442654
>>58442760
x86-64 only exposes 16 int GPRs and 32 AVX registers (except for AVX-512).
Your argument reeks of bullshit that you think there could be many dozens of independent defects that trash half a register file but leave every other critical piece of a core intact enough to sell.

Find some actual sauce backing up your argument or continue looking like a pathetic poser.

>>58442643
>meanwhile at Intel
tick
tock
tick
tock
tick
tock
refresh (haswell refresh)
tick
tock
refresh (kaby)
>let's just call our model "process - architecture - optimization" now! hurr durr
refresh (coffee)
>wait, another refresh? where's the "process" stage of our new release model?
process? (cannonlake)

>find out more in the next episode of:
>Where did my 10nm go?

>>58442381
likely, and this is what i beleive, most of the use of ram is going to be textures soon, 4k requires textures to be there and be good, however do you need a texture that is... 200 meters out to be the best possible?

If this works how I would assume it to work, amd will use system memory as a psudo gpu memory, and pull in textures that are up close and personal to gpu, and things further away but would still be in gpu memory otherwise to the system memory, if you dont have enough, ssd, and if you are like me, a hdd you store all your games on.

this will alleviate some need for mass ram on the gpu, and if things are decided this way, likely a better solution then just gpu memory, but it will have drawbacks when games actually pull in more then 8gb for textures alone.

>>58442881
>I'm going to fallaciously claim that selling a binned chip with SMT disabled is nonsensical jewing
>oh shit I actually don't know what I'm talking about
>oh shit I immediately got called out for it
>oh shit I need to keep bluffing when I'm surrounded by people far more informed than me

Stay. In. /v/.

>>58442449
you have any idea how many pro applications just wont work with more than one cpu? they will use the entirety of one, and not touch the other. this was a reason i didn't jump on a dual cpu motherboard and cpus when they were cheap a while back.

>>58442932
Do you have any idea of how many pro applications work with multiple CPU's (since the 90's, even!)?

>>58442927
nice sauce, faggot.

So will AMD's 8 half cores manage to beat Intel's real 4 cores this time?

Stay bolted to your seats for the next itteration of "ANOTHER MASSIVE DISAPPOINTMENT!" In early March.


-Coming to a forum near you

>>58443048
>Zen
>half core

Adorable shitpost, anon.

>>58443085
>Zen
>not being half the core strength with double the cores
Stay in /v/

>>58443048
Shitposting aside I'd pay good money to watch a 4 hour long in-depth movie about the history of the CPU.

Too bad all I'd get is Hollywood's "hurr electron move because electro-fairies"

>>58427845
>>>/v/

File: Screenshot from 2017-01-01 10-59-34.png (63KB, 536x955px) Image search: [Google]

63KB, 536x955px

>>58443132
What are you trying to accomplish with this shitpost?
I mean.. Come on, even children can find evidence to refute your stupid, it's everywhere.

>>58427307
>I don't get this humor.

Tu n'es pas Charlie.

>>58433200
>more cores is bad
Explain this meme to me. Is it about games?

>>58443155
>believing fake news

>>58427920
He'd get the same performance but have twice the cores (four times if you count HT), and without the chips hitting 80C due to shit thermal paste. And he wouldn't be supporting a jew company anymore.

Not a bad upgrade all in all.

>>58442950
funny enough, not the ones I considered dual cpus for.

>>58443155
How does the Intel 2+2 AGU work? Last I checked Intel was still on a 6 integer pipes so I assume that +2 is conditional or something?

>>58443172
yes and no.
amds more cores initially worked worse then their prior generation cpu, and in practice, ran crappy over everything but render loads.

then you have amds gpu and nvidia's gpu, where nvidia gets shit done usually better with less cores.

so more cores became a retarded meme.

when amd looked to do as good as intel now, 'but it has more cores so it cant clock as good as my i7, moar corz lol' shit is now happening.

>>58443155
Man, it's kinda embarassing to see how small a bulldozer core is.

>>58443384
Might make a good embeded core at low frequency, it's small and cheap.

But that's all I guess.

What's a good source for CPU benchmarks, so I can compare Zen with other CPUs when it arrives?

>>58443582
CPUboss and HardOCP

>>58443582
Look for sites sponosored by Intel since they're paid professionals who know what they're doing.

I don't believe anything until the majority of tech sites have reviewed the cpus and give us hard benchmarks.

And without any date when Ryzen 6/12T is hitting the market, I am not gonna playing the waiting game.

File: execution_unit_overview_large.jpg (173KB, 991x743px) Image search: [Google]

173KB, 991x743px

>>58443284
Intel's handling of load/stores is radically different. Every port on the crazy scheduler has at least some fixed function abilities. They added some dedicated stores so that other ports can handle loads in the same cycle and increase throughput.

The usage of AGUs in AMD's core while serves the same ultimate purpose, is not actually doing the same thing.

>>58443582
Anandtech is the biggest, and they have a decent cataloged bench tool, but they're not totally objective.
Tomshardware can also be pretty decent, but they're not the most fleshed out reviews. Where they shine as of recent is in their power consumption testing.
Computerbase.de is stellar, as is Hardware.fr.

I rely on those four sites primarily.

>>58428109
Given that Intel's 14nm node is better than GF/Samsung's and AMD is offering similar power efficiency to Intel, there's a chance that Zen is actually more power efficien than Skylake on a purely architectural level. Possibly

>>58442908
> I've lost my 10nm, a book by Intel

>>58443604

Top kek

>>58443159
ah oui

>>58440193
Yes, its a new socket, AM4.

I'm ready to upgrade from my i7-3770k. My mobo started shitting itself at the end of last year to the point where I can to revert to stock clocks.

I'm ready to go back to AMD again dammit! I just hope they come through.

>>58443582

Techreport since they're one of the few sites that places more importance on frame times than FPS for games and they are fairly unbiased.

File: faggin.jpg (22KB, 530x382px) Image search: [Google]

22KB, 530x382px

>>58443132
NHK did a superb 6 + 9 parts series back in 1991 & 1995 about electronics, semiconductors, microprocessors and so on; both domestic and the rest of the world
lots of backstory, explanations, interviews with key silicon valley people & jp company heads (my god did they ever copy..), and so on
obviously also includes intel's (among others) story and the i4004. especially the i4004 as it originally was for a jp customer during their calculator war and they could also interview the people in japan that were involved

you'll find them over on ニコニコ動画. it's in japanese, of course. but that shouldn't be problem for anyone visiting a site dedicated to jp culture, like 4chan

>>58430271
*apologizes*

>implying you guys know how SMT werkz

yea right fagits that shit is witch craft

>>58443771
befor feb 27

>>58445015
the basics is this, cores were to powerful on the compute end and could not be fed adequately or they were to slow on processing and parts that did that processing were going to waste, so instead of trim it down, they let 2 threads go at once. in applications that support it, its a positive 25-30% (50% theoretical max) and in applications that don't its not wasting much die space

>>58445385
AMD the last day of March was not their trajectory. Therefore, 30th of March.

>>58445903
http://schedule.gdconf.com/session/optimizing-for-amd-ryzen-cpu-presented-by-amd

>Join AMD Game Engineering team members for an introduction to the recently-launched AMD Ryzen CPU
>recently launched
>before GDC

Ryzen will launch before the end of February.