[ 3 / biz / cgl / ck / diy / fa / g / ic / jp / lit / sci / tg / vr / vt ] [ index / top / reports / report a bug ] [ 4plebs / archived.moe / rbt ]

Due to resource constraints, /g/ and /tg/ will no longer be archived or available. Other archivers continue to archive these boards.Become a Patron!

/g/ - Technology

View post   

[ Toggle deleted replies ]
File: 90 KB, 1056x346, HAHAHA.jpg [View same] [iqdb] [saucenao] [google] [report]
63758367 No.63758367 [Reply] [Original] [archived.moe] [rbt]




>> No.63758395

I guess you can say.. that the performance.. of amd... has ryzen to extreme levels..

>> No.63758465

Good one Anon.

>> No.63758556
File: 109 KB, 2250x2324, 1512249231547.png [View same] [iqdb] [saucenao] [google] [report]


>> No.63758609

5.1GHz at Sandy Bridge IPC is not very impressive

>> No.63758617


AMD toilet-worshippers are getting desperate

>> No.63758619


>> No.63758668

I don't really know why people still say this considering it wasn't even true on day 1.

>> No.63758690

Since when it Haswell ipc the same as sandy bridge?

They only need a 10% ipc boost and clock speed boost to btfo Intel again

>> No.63758701

except ryzen IPC is better than skylake

>> No.63758721

>Sandy Bridge IPC
hahaha you fucking retard. It's haswell/broadwell IPC. Only 3-6% behind coffee lake. If you get the memory over 3200mhz it actually has better IPC than skylake, retard

>> No.63758834
File: 117 KB, 725x364, Ryzen Slide 3.png [View same] [iqdb] [saucenao] [google] [report]


>> No.63758836

Why are you guys so defensive about this? Even if Ryzen 2 hits 5.2GHz it would be another FX9590 housefire.

>> No.63758845
File: 11 KB, 501x585, 2132138910.gif [View same] [iqdb] [saucenao] [google] [report]


>> No.63759133

It's not being defensive, it's just living in reality as opposed to your mother's basement.

>> No.63759179

No mutli anymore? Anon did proofread his slides.

>> No.63759197

So first they sold ryzen 3/5/and 7. Then 9. and now 2?

what the actual fuck

>> No.63759207

Well if it's true I'd hate myself for ever buying a 1700X

>> No.63759211
File: 20 KB, 493x280, AMD-Ryzen-Vs-Kabylake-Single-THreaded-Test-IPC.png [View same] [iqdb] [saucenao] [google] [report]

sandy bridge ipc

>> No.63759263

found the thread

>> No.63759488

>12 cores on desktop processor.
So how much cores will we see in threadripper 2? 16?

>> No.63759710
File: 84 KB, 653x726, 1499957247669.png [View same] [iqdb] [saucenao] [google] [report]


>> No.63759748

If it maintains the same core count per die setup as current threadripper, up to 24c.

>> No.63759792
File: 380 KB, 500x375, mutley.gif [View same] [iqdb] [saucenao] [google] [report]

Mutley threading

>> No.63759987


>> No.63760012

48C per processor. 96 per 2P system. How does 192 threads per system sound?

>> No.63760073

>mfw mobo and CPU shit itself last month and bought Zen
Holy shit, I didn't except this much of a boost. They even upped the already great core count. Zen is already great but this Zen(2 or +?) will rip Intel a new asshole. Guess I'm either upgrading in a few months or just waiting for 7nm Zen.

>> No.63760104

so this ryzen + is the refresh they're going release in 2018 ? or the true ZEN 2 going out in 2019 ?

>> No.63760165

Isn't this fake? I thought Zen+ was just a die shrink, enabling higher clocks on the exisitng Ryzen line.

>> No.63760222

Its the node change most likely. The current node doesn't allow for super high clocks which is why voltage really takes off after 3.3ghz and most Ryzen and threadripper chips top out within the 3.9 to 4.1 range.

>> No.63760302

>12 cores
>24 threads

This would be amazing, but it's literally unbelievable. At best, assuming it's If not a house fire, you'd need a nuclear power station to power it.

>> No.63760452
File: 707 KB, 752x620, 1511946393182.png [View same] [iqdb] [saucenao] [google] [report]

Probably fake. I doubt you could get that without a housefire.

>> No.63760489

The slide is fake.
Zen+ is a better node plus probably different SRAM lib and design rules due to it being a client-only die.
Don't expect any more cores, but clocks...

>> No.63760518

Avoiding a housefire is as simple as soldering the die to the IHS and having proper cooling. 5.1GHz is unbelievable though, that sounds like a straight up miracle so I very much doubt it's true.

>> No.63760624

Hey another hardware engineer.


>> No.63760668

How much higher clocks are we talking?

>> No.63760712

Guys, AMD already has a 12-core part. They could absolutely release a mainstream 12-core. Did you forget what IF does?

>> No.63760714

The 2018 refresh will see a meagre 200-400 Mhz base clock boost and better memory compatibility. Nothing else.

>> No.63760786

>damage control this far ahead

>> No.63760808

Depends on the workload. In some cases Ryzen's straight up 30% better. But gets its ass kicked when AVX or AVX2 are involved

>> No.63761086

>Did you forget what IF does?
Eh, it's a protocol.
So what?
A lot more clocks.
Also housefires, but who cares anyway?
Current Zeppelin (ZP-B1/B2) is 14LPP with HD libs (78/9T).
PR (probably ZP-C0/C1) will be 12LP with 9T HP or even 10.5T UHP libs.

>> No.63761121

>gets kicked on avx2

so basicly 99% of the times ryzen clock to clock is faster since no one uses avx2 for anything relevant what so ever

>> No.63761143

Okay, now what is that in MIPS?

>> No.63761994

>Eh, it's a protocol.
... IF is an MCM technology. It means that AMD doesn't have to change ANYTHING about the Zen architecture and could release a 12 core Ryzen part. There are two limiting factors for how many cores AMD can put on the mainstream Ryzen parts: 1) TDP and 2) cost.

For TDP, AMD has 30W of play before they get back to 125W, which has been the target for their high end mainstream parts for quite a while. AMD has even gone as high as 140W for mainstream parts in the past, so TDP really isn't an issue. As for cost, we're talking about a smaller node, which means more complex chips at the same cost. Maybe only marginally, but AMD also has a very mature 14nm node to lean on for profitability in the server space while they claw back their reputation in the consumer space.

tl;dr, if your only supporting argument for saying that AMD won't release 12 core Ryzen parts is that they're not redesigning the architecture, you haven't been paying attention at all.

>> No.63762589

Nope, you're oversimplifying things.

>> No.63762678

No, They are shrinking the process. Which always improve temperatures, and even the current Ryzen is not limited by temperature, but a lot ofther things.

When that is said, i believe this is fake. That much IPC and Clock gain, is not gained by a simple die shrink, but a redesign. Which this isn't.

>> No.63762736

I believe it's fake but it's not unheard of that a CPU was released with features missing or disabled because they weren't ready in time or had stability, security and errata issues.
Actually Vega is a good example, Intel's Skylake-X also has some features it should have disabled, likewise the only really known disabled feature of the current Ryzen we know is the LDO and the PCIe lanes(probably more due to pins and motherboard design than stability)

>> No.63762896

Obvious fake, what they're releasing on February is Zen+, which is just a process improvement. It will most likely have better clocks, but I highly doubt they will increase core counts. Zen2 is coming late 2018/early 2019 and it's 7nm.

>> No.63762945

People were saying the exact same thing about Threadripper.

>> No.63763037

>... IF is an MCM technology.
What were Magny-Cours/Interlagos/Abu Dabi?
>could release a 12 core Ryzen part
New die, many manhours; not worth it.
>tl;dr, if your only supporting argument for saying that AMD won't release 12 core Ryzen parts is that they're not redesigning the architecture
They are not going to waste time making a new 12core die.

>> No.63763135

Wheelchair CPU architects, shut the fuck up, do you have any idea how old CPU architectures we were looking at before Ryzen hit? Check when Pentium M was released, that should tell you, in what sane mind are you putting something mint new with obvious bottlenecks and comparing it something old as shit? Even fucking Bulldozer, that piece of flaming low-IPC-by-design turd gained 15-20% IPC in the span of 2 years and AMD basically abandoned it in less than a year! Only leaving a few people to work on it so we're not stuck with Zambezi.
Where do you get off that a new core which is much better is basically gonna trail Intel?
If x86 IPC was impossible to improve, then why the fuck is Intel making a new x86 architecture in 2021?

>> No.63763175
File: 125 KB, 960x720, c7fix5jkqy201.jpg [View same] [iqdb] [saucenao] [google] [report]

>> No.63763184
File: 108 KB, 1442x719, 1509484800491.png [View same] [iqdb] [saucenao] [google] [report]

They would have to either make a bigger die, or release it for the X399 platform. Both options seem really unlikely, AMD roadmaps show Pinnacle Ridge parts with up to 8 "Zen" (not "Zen2") cores.
This would be believable if it was actually about Zen2, but whoever faked the slide fucked up by claiming that this is what AMD is going to release in February, which we already know it's a 12nm part, not Zen2 which was confirmed to be 7nm.

>> No.63763211

>ryzen 7 gaming 5

AMD Marketing at its best.
Why not call it Ryzen 7 Gaming 5 CPU 3 Core 8

>> No.63763231

It's Ryzen 7/5 Gaming (something like 2500G in AMD naming scheme).

>> No.63763247

>Ryzen Gaming, mobile
Now that sounds interesting.

It's Ryzen Gaming 5/7, lmao.
>ryzen 7 mobile apu 5

>> No.63763272

>Now that sounds interesting.
CES will be even more interesting.

>> No.63763430
File: 1.99 MB, 400x300, 1511663546051.gif [View same] [iqdb] [saucenao] [google] [report]


Cant believe you are such a brainlet or are baiting.

Its ryzen 7 and ryzen 5 you mongoloid.

>> No.63763454

But the "Gaming" is part of the name like Threadripper is, meaning it's not the same chip as whatever before it.

Why the differentiation? Different die?
Remember it's under the premium mobile column, what is it doing there? a 40W notebook part? Doesn't seem premium mobile to me.

>> No.63763481

Good luck fitting 2 dies in an AM4 socket.

>> No.63763504

It's a "G" part in their naming scheme.
Means whatever.
I though M was high-power mobile.
Maybe, maybe it's a APU + dGPU part akin to KBL-G, but thicket (with 2.5D interposer and without EMIB)?

>> No.63763526

Maybe some APU with integrated HBM like people were rumoring years ago? I remember reading about cheaper HBM coming soon, but I can't find the link now.

>> No.63763534

RR die has no HBM PHY.
Besides, Papermaster said "NO" to HBM integration this gen.

>> No.63763555

>Doesn't seem premium mobile to me.

Good thing no one listens to you.

>> No.63763584

>low, cheap, stupid & retarded, as always

>> No.63763605
File: 18 KB, 678x440, 1484837419115.jpg [View same] [iqdb] [saucenao] [google] [report]

Then no idea what it is. I doubt AMD will compete against their own GPUs by releasing something like the glued togetherâ„¢ CPU+GPU that Intel announced.

>> No.63763646

>I doubt AMD will compete against their own GPUs
They won't, KBL-G is basically an Apple product, and Vega11 will be more powerful (assuming IT FUCKING WORKS) than whatever Polaris intel got.

>> No.63763787

This board is still as garbage as it was when I left several months ago. What a fucking shithole. Gibbering fuckshits still repeating crap about IPC and clockspeed that they don't even understand.

>> No.63764269

What were Magny-Cours/Interlagos/Abu Dabi?
Parts that didn't use IF-- that is to say they did not scale in the same way that IF does, which allows mainstream parts.

>New die, many manhours; not worth it.
It's not a new die. It's three CCX instead of two using two dies. I'm getting the impression you have no sweet clue what you're talking about.

>> No.63764288

>Parts that didn't use IF-- that is to say they did not scale in the same way that IF does, which allows mainstream parts.
They used HT3.0, which IF is vaguely based on, and scaled actually very well, it's just the cores were meh (K10.5 was outdated and BD was shit).
>It's three CCX instead of two using two dies.
CCX is merely 4cores + L3.
It's not a die.
Zeppelin is a die.
Contains 2CCX, a 2channel DDR4 IMC, 32 PCI-E lanes, 4GMI-links, 4 GP PHYs and something else I totally forgot about.

>> No.63764371
File: 42 KB, 700x374, ryzen.jpg [View same] [iqdb] [saucenao] [google] [report]

Look at the 1800X. There's space on that chip for another CCX.

AMD could release mainstream 12 cores whenever they want. The question is whether or not it makes sense in their product stack. The answer to that question really depends on clock speed, which is what I doubt most in that slide, not the core count.

>> No.63764401

So if you understand how IF works and what the CCX is, why are you insisting that AMD would have to design a new die to get a 12 core when you understand how they would achieve that without designing a new die?

>> No.63764410

I've been waiting for this product to take shape forever now. I want just like basic 1080p gaming capability in a laptop that's practical to bring around with me all day and that chip is fucking it.
Nvidia could do more in the laptop gpu space but this competition is going to be awesome.
I don't know how everyone isn't hype about a cpu/hbm/gpu package.

>> No.63764434

>understand how they would achieve that without designing a new die?
They won't.
You need moar crossbar ports for another CCX, which means severe modifications to Zeppelin.
Fucking hell please stop.

>> No.63764459
File: 105 KB, 1920x1541, apuu.jpg [View same] [iqdb] [saucenao] [google] [report]

>i maek gigaheartz

>> No.63764468
File: 697 KB, 650x462, 1505145947750.png [View same] [iqdb] [saucenao] [google] [report]

It's not that simple, you need space for the interconnects too.

>> No.63764522

Oh, I misread your post. Yeah, there is space for another set of 4 cores, but they would need to redesign stuff, like how to connect the new CCX to the other ones in an efficient way, and it would also increase die size, which could hurt yields. Seems unlikely to happen until 7nm is ready.

>> No.63764542

Finally someone gets it!
They don't even need more cores, since they can trade density for clocks.

>> No.63764605
File: 706 KB, 1517x856, 738f3dc3e8fccbdb72fa42580af90ca0.jpg [View same] [iqdb] [saucenao] [google] [report]

So let's be modest here and say that we'll see 4.5ghz out of the refresh and higher memory clocks. The memory thing might be optimistic.
If we look towards ln2 ryzen runs, taking the negative view of no ipc gains and all frequency, we see a cinebench score of 2022. This is on a 1800x. Comes out to 126 a thread. Pic related.
Looking at the 12 thread 8700k running at an obtainable 4.9ghz(I know they can go higher) you get a cinebench score of 1591 or 132 a core.
These things are going to be great for competition against Intel but will still loose to 5ghz chips in games that only scale to about 4 threads. Same story, nothing will change unless they also manage a ~6% IPC uplift.
If that happens it will come down to specific workloads, memory bandwidth, and inter-core latency.
As a gamer(I know I suck cocks thanks) I don't think it will be superior to a delidded [email protected] and ddr4 4233 cl19 but that's just because I buy at the high end. I suspect the ryzen+ 1600x will fair very well.
For production ryzen is better than consumer intel now and will be even better on the refresh.

>> No.63764609

Zeppelin is the base for Ryzen, Threadripper, and EPYC. They're all the same die. The MCM is what makes them different, but the MCM is not expensive to manufacture.

>how to connect the new CCX to the other ones
The same way they do in Threadripper, but the MCM can be smaller since they don't need to add another memory controller or any of the other extra features on TR.

>> No.63764633

My fucking god I am starting to despise AMD for their marketing.
Fuck me this is truly the end times.

>> No.63764643
File: 60 KB, 1014x1024, 1512307649147.jpg [View same] [iqdb] [saucenao] [google] [report]



>> No.63764654

But what does that have to do with Ryzen though

>> No.63764657

I have no idea what you're getting at

>> No.63764658

August next year?
1 year to EOL seems decent.

>> No.63764671

>12 cores
if this is true Intel is shitting 10nm bricks right now

>> No.63764695

They can't shit 10nm bricks.
They don't yield.

>> No.63764704


>> No.63764738
File: 2.05 MB, 742x610, 1493098009707.webm [View same] [iqdb] [saucenao] [google] [report]


>> No.63764752

>The sky is orange
>It's blue
>Why you gotta be all mad tho?

Also, Ryzen is a very efficient architecture. FX was already hot and power hungry before they pushed it to its limit, so the result was a foregone conclusion. Ryzen has a lot of headroom before things get out of hand. And FX was competing with Sandy Bridge which ran cooler and more efficiently so that only increased the gap, whereas Ryzen would be increasing power draw and heat output up against Housefire Lake, so it would only level the playing field at worst.

Short version: Ryzen is not FX, stop it already.

>> No.63764818

>CES will be even more interesting.

>> No.63764829

A lot of RR laptops, possibly various smaller Vega, possibly 2000 series announcements.

>> No.63764873
File: 22 KB, 718x512, 1485130431686.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.63764886

Nothing uses avx2 so who cares.

>> No.63764904

Sounds like an even larger reason for Shintels to boast over single thread performance

>> No.63764905

Asian Sam Hyde

>> No.63765113

>tfw waitied for Ryzen 2
Thanks for Beta testing

>> No.63765117

>50%+ performance improvement in 18 months
Moore's Law = alive and well

>> No.63765141

>12 core, 24 thread
>implying cheapo 350 boards can handle all of that

>> No.63765142

Not for Intel, heh.

>> No.63765188

it's alive for Intel, too, but they got fat and lazy and don't want to hurt their profit margins.

>> No.63765189

Why not?

>> No.63765203

lol literally who even fucking cares?

>> No.63765215

t. Brian Krzanich and his retards from TMG

>> No.63765226

Oh look, a fucking Logan shill

>> No.63765256
File: 248 KB, 400x580, srsly google.png [View same] [iqdb] [saucenao] [google] [report]

If that's 12nm, what's going to happen on 7nm in 2019?

>> No.63765265

Upgrading in 2019?

Ryzen+ is 2018 (just a refresh)
Ryzen 2 is 2019

>> No.63765272

Why must you feed this shit to the amd shills, they are a disease and you are only perpetrating it.

>> No.63765283

Rome will happen.

>> No.63765344
File: 350 KB, 672x794, 6541684984949948.png [View same] [iqdb] [saucenao] [google] [report]


>> No.63765367

That is exactly why Coffee lake is so shit

>> No.63765368

Nothing, we learned a long time ago that increased clockspeed has a diminishing effect. Pushing it up that high is mostly a marketing thing.

>> No.63765405

fuck off, Sam is a basementdwelling neckbeard

>> No.63765406

lol I never even said that i'm for intel, I was just saying amd fanboys are lurking all over /g/ and its annoying.

>> No.63765422
File: 54 KB, 1024x576, lelouch.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.63765448

Why is the bald guy from kitguru on that picture? he hasn't made a video about it yet.

>> No.63765528

Just buy one now and upgrade when 7nm Zen2 Threadrippers arrive, if you need it. Waiting is a meme.

>> No.63765543
File: 1.64 MB, 1299x728, PROMO_SLIDE24.png [View same] [iqdb] [saucenao] [google] [report]


>> No.63765567
File: 26 KB, 326x327, Cum-slut.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.63765588

same socket?

>> No.63765594

Of course.
AMD is commited to SP3/SP3r2 until Zen3.

>> No.63765602

u wot?

>> No.63765608

SP3r2 aka TR4.
Read the post properly ffs.

>> No.63765618

>Falling for Poo lies every year

>> No.63765650

>S U I C I D E W A T C H

>> No.63765655

Why does it say Ryzen 2 when, from AMD's own materials it's actually Ryzen+

Isn't the actual Ryzen 2 coming in 2019?

>> No.63765667

Ryzen 2, not Zen 2.
Ryzen =/= Zen.

>> No.63765680

It's an AMD "markup" apparently, preparing for the year / getting promo ready, or totally fake, great shitpost if true

>> No.63765684

wait, so this is am4 right?

I might sell my 1800x for the 2800x to be quite honest

>> No.63765699

It's Ryzen 2nd gen, not Ryzen 2

>> No.63765701
File: 14 KB, 423x453, 1488641250981.png [View same] [iqdb] [saucenao] [google] [report]

Can we lay off the hype train for this one? Then you won't look like such an idiot when it finally comes out. Again.

>> No.63765727

If you haven't noticed, AMD CPU division was executing so well Intel is shitting bloody bricks.
Besides, AMD started leaking info, which means they are preparing the hype train.

>> No.63765742
File: 614 KB, 992x1043, 1506963063554.png [View same] [iqdb] [saucenao] [google] [report]

>If I pretend it doesn't exist it can't hurt me

>> No.63765769

It's doing well, but it's not living up to anything near the hype that it had before launch. Temper your expectations.

>> No.63765770

>until Zen3.
didn't they say zen2?

>> No.63765819

Ryzen 2 is Zen+, Ryzen 3 will be Zen 2

It's marketing move, people looking at it think it's a next gen CPU rather than just improvements on current gen. Kind of like Intel's 3 ticks before a tock

>> No.63766021

Yes. My 2500k is fine for now

>> No.63766056
File: 153 KB, 600x339, Intel Afraid.jpg [View same] [iqdb] [saucenao] [google] [report]


Name (leave empty)
Comment (leave empty)
Password [?]Password used for file deletion.