AMD Unboxed Debunk Zen 2 & RX 3000 Delusions

OH NOONONONONONONONO!!!!
youtu.be/iXr7_O1v8-c

Attached: AMD Austin.jpg (540x720, 252K)

Other urls found in this thread:

youtube.com/watch?v=8mqSkkhAWiw
dl.acm.org/citation.cfm?id=1905217
twitter.com/SFWRedditVideos

The Intel PR machine is in overdrive. I've seen articles and videos popping up everywhere in the last 48 hours trying to debunk the leaks (aka reality) and push some narrative that intel's 7nm EUV node is somehow "on track". Every single one of these "debunking" videos fails to properly address the very uncomfortable reality that is the chiplet architecture and the insane yield advantage it affords.

Rome is a LIEEEEEEEEEEEEEEEEEEE

Attached: 1542027477262.png (807x745, 205K)

This. AMD leaked the Zen 2 & Navi details themselves. Intel then panicked and within 24 hours revealed their hand with the 10 core Comet Lake-S. Then the XBox Scarlett details and some other stuff came out backing up the AMD leaks. Now Intel is shitting themselves, shoveling money into shills hands.

Attached: amd_leas.jpg (640x378, 37K)

Every debunking of the debunking, however, also fail to address the fact that AMD just doesn't need to undercut the prices of their *own* products by such an insane amount. Even their current products are already highly competitive in terms of price. Why would they choose to cut prices by such large amounts at the same time as they push such a large improvement to them, when they could instead improve their margins? It makes very little sense.

they just dont give a fuck

Right, truly the mark of good business leadership.

Maybe that's the fake part of the leak, AMD purposely put in low ball dollar figures to get the nerds salivating, so they'd spread it to their friends.

Attached: slip_and_fall.gif (240x300, 3.89M)

the pricing column values seemed to be all wrong on all those "leaks"
i dont get why dumb jim believes in that crap

I just want to believe.

wtf is she ok!?

Because they can and want to generate sales and an inventive to switch? Competitive pricing is not a difficult concept.

How many subscribers would this faggot lose if the leaks are actually true on release?

She got ZENED.

KEK @ that picture

>debunk
Why is this fake news' favorite word these days? Is it because they can just say it as a headline and trick normies into thinking it's true?

She burnt her ankles pretty bad, but apart from that she's fine

>have a bunch of gold
>set up shop next to a place that sells gold nuggets for $1000 each
>sell all of your gold bars for $50 each
>have no gold left and very little money left

>they can't be better than intel that's antisemitic !

I don't think we are running out of sand anytime soon...

GPU leaks sound too good to be true but CPU leaks are perfectly reasonable I really don't understand why anyone would think they are unrealistic.

>chiplet architecture and the insane yield advantage it affords
Kill yourself retard.

>Competitive pricing is not a difficult concept.
It is indeed not, but my point is that they already have extremely competitive prices. Competitive to the extent that there's no need to slash them further. Especially considering that Ryzen 3 is also a technical improvement on top of their already very strong position.

Though neither is sand availability the determiner of semiconductor manufacturing prices.

What the flying fuck was your point then?

What a retarded analogy

hey look repeat of 2017 logic again, how did it work out for you back then? ah right, AMD released competitive 8 core for peanuts compared to intel prices.

INTELTODDLERS SEETHING

Attached: tenor.gif (498x278, 1.69M)

based

But again, the point is that their pricing is already extremely competitive compared to Intel. Yes, the 2017 prices did undercut Intel by a great margin, but their current prices still do, and still will when Ryzen 3 is released. Why would they reduce them even further instead of choosing to increase their margins?

Why beat the competition when you can kill it?

Nuke the market.
The only way to properly coerce OEMs is to nuke the market.
You'll see what Rome will do to Intel server designs.

this AMD is out for blood

based Slutania poster

Don't you get it? If only the technical parts of the leaks are true, and AMD simply pretty much kept their current pricing, they'd still be killing Intel. Why would they choose to do so at half the price they could charge for it? It's not like AMD couldn't use the money for further R&D.

>tfw when 4670k
>tfw rig overhaul due soon
>tfw amd is back in time for it
I'm coming home

Nuke the market.
They will earn that money by eating into Intel's share.

source

You don't understand why Lisa Su is so confident, Zen is a superscalar and Intel shit isn't. They could probably go 128 cores for 5nm Epyc, this is how insane Zen uarch is.

>Zen is a superscalar and Intel shit isn't
Jow Forums - Technology

Ringmesh is fucking trash.

>half the latency of infinity fabric
>trash

?
You were talking cares and now you'retalking interconnects?

maximum damage control.
rx590 is just a ruse to throw off nvidia.

>gets destroyed in every way possible
>WE WON
INTEL YES

>every way possible
rollingeyesmiley.webm
Noone sane is questioning Intel's single-threaded lead, and the memory latency is an important part of that.

>gets destroyed in every way possible
>nuh-uh
ok this is epic

Attached: okay-1.gif (245x285, 916K)

BINGMESH

Attached: 1537707457895.png (1286x144, 13K)

>single thread
Get with the time, gramps

singlethread will be obliderated by rome. it's the new holocaust they have feared for so long.

Reminder this is how Intel looks even with a better 14nm node than Ryzen. Now imagine 7nm Ryzen.

Attached: untitled-3.png (672x794, 44K)

I sure hope that is correct, but I do have to say I'm worried about what the I/O die will do for latency on the desktop.

Kill yourself. Even with infinitely scalable, ideal data-flow architectures, there's still a sequential critical path through every algorithm.

youtube.com/watch?v=8mqSkkhAWiw

cope

Core

Front-end
Improved branch prediction unit
Improved prefetcher
Improved µOP cache tags
Improved µOP cache
Larger µOP cache (?? enters, up from 2048)
Increased dispatch bandwidth

Back-end
Increased retire bandwidth (??-wide, up from 8-wide)
FPU
2x wider datapath (256-bit, up from 128-bit)
2x wider EUs (256-bit FMAs, up from 128-bit FMAs)
2x wider LSU (2x256-bit L/S, up from 128-bit)

Security
In-silicon Spectre enhancements
Increase number of keys/VMs supported

I/O
PCIe 4.0 (from 3.0)
Infinity Fabric 2
2.3x transfer rate per link (25 GT/s, up from ~10.6 GT/s)

>fuck science, just slap moar coares on it, software devs are LAZY!!!1

cope 2: electric boogaloo

he's right, you know
you're just being retarded right now
strong single core performance is as important as having multiple cores

>everything that ever happens is constrained to a single global serial execution path
wow holy shit. why did we ever bother with more than 1 core in the first place?

lets see how intel double their core count

Attached: 1537204806160.gif (108x147, 284K)

Yeah but all intelel has left is single thread so I like to shitpost about it

>the truth must be that algorithms are either infinitely parallelizable or not parallalizable at all
Pretty stupid statement, senpai. Obvious the truth is in the middle, which is exactly why both multi-threaded and single-threaded performance matter.

What a retarded thing to say. Obviously scalability is the downside of the ringbus, noone is disputing that. But that doesn't mean it doesn't have upsides as well, and the main such upside is its latency.

PCIe 4.0 is downward compatable, right? so we could see every single every SKU operating on 'older' mobo's and having PCIe 4.0 on the leaked x570 platform as a premium feature first (and better power delivery specially for R9's if true) but i guess we won't see 4-channel Memory on it, yet.
isn't too far fetched or is it?

Attached: 1529567124804.jpg (757x627, 87K)

imma get me one of these new ganfled zen chips
when do they lunch though

waiting any day now till memory latency doesn't matter to fill the list even further for delusional Intards

>so we could see every single every SKU operating on 'older' mobo's and having PCIe 4.0
No, the boards need a PCB respin to handle PCIe 4 signaling.

We will know early next month

x370/470 using PCIe 3.0
'new leaked' x570 using PCIe 4.0 only
thats why i ask if it's downward compatable

Pcie 4 is only forwards compatible. CPU, motherboard, target board all need to be compatible. You can run the cpu, motherboard and peripherals as pcie3 for backwards compatibility.

Always read the manual from your motherboard and CPU vendor when in doubt of link speeds.

I wasn't talking about cores I was talking about Zen UARCH being ABLE to SCALE to 64 and maybe even 128 cores without breaking a sweat while Intel glues two ringbus garbage dies together.

>maybe even 128 cores
drop the maybe. if AMD doubled it once they will double it again when a newer process provide the needed shrink/density. propably 5nm EUV process can manage that

It is very definitely downward compatible, yes.

While I'd very much like to see that, it's quite unlikely that AMD is going to win in latency if they're adding an I/O die to Ryzen 3.

Yeah but Intel has 4k Netflix, take that amd shills

That's not what "superscalar" means, though.

>no Mads Mikkelson as aged Geralt trying to save Ciri produced by HBO or even Amazon

Attached: 1541065888542.jpg (899x1280, 396K)

>if AMD doubled it once they will double it again
That is assuming IF scales indefinitely, and there's no intrinsic reason to assume it does.

fair enough, time will tell

delet this, it hurts too much

If anything, it's pretty annoying how AMD isn't revealing any technical details whatsoever about the infinity fabric. Even Intel is more forthcoming about their coherency protocols.

no.

Attached: henry.jpg (1500x999, 69K)

Exactly.
>Competitor has sold no gold until you're completely out
>You end up with all of the consumers' money
Except in this analogy, you can produce rather than harvest gold.

>triss is a mutt
>yenn is a gypsy
>fringilla is a nigger
I SAID DELET

Attached: 1541368683304.jpg (1040x768, 299K)

You're right, the right term is hyperscalar, which sounds even "higher" than superscalar, thanks for the confusing terms, faggots.

What you're looking for is "scaler", not "scalar", user. There's a pretty big difference.

???
dl.acm.org/citation.cfm?id=1905217

That is a completely different thing from what AMD is doing.

LOVE IT ALREADY!

Attached: 1538770348655.png (374x253, 32K)

I just pretend the show doesn't exist, even when I see those images I don't feel anything.
The show will be shit btw, not because of the cast, it will simply be shit.

Attached: su.png (1400x1000, 986K)

That is so beautiful, user! I love that they're taking diversity seriously, and show that a rainbow is more beautiful. LOVE PREVAILS!!!

Attached: hqdefault.jpg (480x360, 18K)

MOMMY

Attached: 1519821338768.gif (120x97, 78K)

Everyone is so used to Intel's dominance, the idea things will flip in 2019 disturbs them.

I know. Look at how many people go full COPE mode after Zen2 leaks. 2017 was just the beginning. 2018 was already very entertaining.

Well, this video is already retarded.

>It makes no business sense to launch a product at half the price of your only other competitor's product in the same market segment! That would never happen!
AMD literally did *exactly* that with the 1800x.

Intel 6900k price: $1,000
Ryzen 1800x price: $499.99

So congrats to Hardware Unboxed for confidently asserting opinions that are totally wrong.

Except 6900k is still better.

Attached: chrome_2018-12-07_10-23-41.png (892x186, 27K)