Engineer confirms Zen2/3+ prime metrics are perf/watt, as if

File: 1488662022312.png (64KB, 960x540px) Image search: [Google]

64KB, 960x540px

Here's an example of the perf/watt, the 6900k scores 1600 while drawing something around 130 watts.

File: 03_big.png (49KB, 849x621px) Image search: [Google]

49KB, 849x621px

Dam, you can already get the R7s to consume around 65 watts under load if you can undervolt them at stock base frequencies.

>>59767957

>find the Vega shit
>around 1:27:45
> a shitty 10 second animation with shitty sound

>WHAT DO U THINK? RADEON RX VEGA

>intermittent claps

>ok very good


zero new info but coming soon (tm)

>IF is used literally everywhere

Neat, the reason they're not using custom buses is because it takes too much time.

>>59767957
He also confirmed that Zen will be in fanless devices, so it very well scales down to 4-6W

>>59768244
>4 core 13" fanless Zenbook/Macbook/Carbon/Spectre/whatever


Oh golly gosh

File: 1489935357709.jpg (66KB, 568x612px) Image search: [Google]

66KB, 568x612px

>>59767957
So... literally nothing?

>>59767957
Ryzen 3, late Q2
Naples, late Q2
Mobile Zen, Raven Ridge BGA, second half of 2017, AM4 versions probably later.
Vega "just around the corner", also Q2

>>59767957
now imagine if they had kept their rolling release and continued releasing processors capable of 220w
their performance would be astronomical
but nooo they had to stop production and sit and rot while intel made bank off milking retards and laptops caught up in performance.

fuck these worthless laptop processors. aint nobody gives a fuck about the fraction of cents saved off 95w cpus compared to 125w. but the performance? would be something to be proud of. even more the performance we'd have if they continued releasing 10-15% perf increases every year or two, we'd have far better than this bullshit 30% increase over 5ghz vishera

fucking worthless laptops

>>59768041
Stupid tripfag doesn't understand power consumption

>>59767957
Short description of Vega

2x the Geometry throughput PER CLOCK(gimpworks improvement)
L2$ tied to ROPs
Tiled rasterization
Variable precision model, 8/16/32bit ops per clock
Ability to handle NVRAM/block devices

Rest is practically marketing nonsense, like "higher IPC, higher clocks, new shader engine, new XYZ" which doesn't go into detail besides being NEW


L2 <-> ROP
Tiled rasterizer and much faster Geometry engines are what you're looking for in games, should be a significant improvement.

>>59768041
whose gonna tell him...

File: power_consumption.jpg (210KB, 720x733px) Image search: [Google]

210KB, 720x733px

>>59768613
I don't get it. The graph he posted reflects real world power consumption tests and the voltage scaling of ryzen.

~65w power consumption is possible on a 1700x if you could undervolt it a little at 3.4 GHz.

Anyways this makes it pretty interesting for naples. Perf/watt might be lower than xeons.

>>59768778
>he
>not I

I hope Pinnacle Ridge is Q1

>>59768568
Servers yo, servers. Perf/watt is king in the data center, as every watt the chips use has to be paid for at least 2x over: Once for using it, and again to reject it from the data center.

If they can maintain (or better yet improve) performance while increasing the energy efficiency further, they will become the king of the data center. Having top of the line performance doesnt mean dick if it costs too much to use it.

>>59768896
Perf/watt is important but memory capacity and I/O is much more important.

Will AMD take the laptop market when the Ryzen APUs are launched?

>>59768939
True, but 128 PCI-E lanes and 8 memory channels per processor means Naples has that shit locked down.

>>59768968
Not easily with Intel contra revenue

>>59768896
>>59768939
even then these aren't server chips, there are specific models for those specific tasks.
consumers though? fucking give us MOAR POWARRR for as much as we're willing to pay in electricity/PSUs

>>59768684
Tell him what? My 1700 runs whisper quiet with the stock cooler under load. I was able to get 1.1v at 3.2 GHz on mine

>>59769063
Well, if it pleases you there's a rumor going around that AMD plans on dropping a 12 and 16c bombshell on Intel for the high end enthusiast/prosumer space. They wouldnt even need to reduce core clocks if they base it off the 1700 spec. Just put 2 on the same package, double the TDP (to 130W), and fucking go.

There's also the possibility that the 12/16C processors will be unlocked. How does 16 cores pushed to 4ghz sound?

>>59767957
>Krzanich

Isn't that the retard who delayed 10nm lines because he thought Samsung, TSMC, and GloFo could never catch up?

>>59769129
>low frequency
>low voltage
Wow I wonder how it got so quiet

>>59769136
I see no reason why they wouldn't be overclockable, your only limit is board VRM.

>>59769002

Intel has no answer in the mobile market.

Paying OEMS would be the only way they would keep them out, and AMD has already been linked in rumors with Surface Pro 5, custom APU in MBPs.

And I don't think Intel wants to be slammed with an even larger fine the world over for doing it twice.

>>59769194
>Paying OEMS would be the only way they would keep them out,
Good thing Intel has a lot of money, ey?

>>59769159
Yeah but what impresses me is that I'm getting this on an 8-core processor.

Isn't OC'ing and undervolting significantly more unlikely on chips with 8 cores vs 4?

>>59769240
No

>>59769240
no not really
They are about the same

>>59769251
Elaborate. Because what I understand is that the frequency and voltage is applied to ALL 8 cores, increasing the chance of one of them not being able to function because it requires more volts or can't handle higher frequencies.

>>59769284
You're at 1.1 v 3.2ghz on an 8 core
That's about all you need to know, there is no difference. Turn your trip back on

>>59769220


>And I don't think Intel wants to be slammed with an even larger fine the world over for doing it twice.

Violating the law and manipulating the market, again? That's a 2 billion dollar paddlin' for each country you tried to do that with, was it worth it?

>>59769332
As if that would fucking happen, especially now that Intel is practically a government owned company.

>>59769323
I'm not the tripfag you useless sack of potatoes.

Can anyone who's not retarded please explain why OC'ing and undervolting is the same on an 8-core ryzen or 4-core i7 processor?

>>59769351

A man can dream...

https://www.bloomberg.com/news/articles/2016-10-20/intel-should-win-fight-over-1-16-billion-fine-court-aide-says

>>59769365
It's the same, that's literally just the way it is, you were wrong
It doesn't take much effort to maintain 8 cores compared to 4 cores when you're already at a low frequency in the first place.
I'm running my 6700k at stock voltage 4.5 ghz

File: 1466495411030.jpg (74KB, 500x707px) Image search: [Google]

74KB, 500x707px

>>59768244
>He also confirmed that Zen will be in fanless devices,

Like my 65W TDP AMD A8-7600 with a huge 140W TDP heatsink?

> Current CPU temperature: 41C
> MINTEMP=hwmon3/device/pwm2=48
> MAXTEMP=hwmon3/device/pwm2=54
> MINPWM=hwmon3/device/pwm2=0

CPU-fan kicks in after about 20 minutes if I do something that puts 100% load on all cores like video encoding.

Not sure if it'd would work well without a total of four 800 RPM 12" case fans but it sure is nice.

If I do upgrade to Ryzen or a new Intel CPU then I'll absolutely make sure it's 65W or less.

>>59769435
I think he meant laptops and industrial design/tough chassis' that need IP68.

You know, those things that can't afford to have 5 pound of copper

File: one core to rule them all.png (67KB, 602x314px) Image search: [Google]

67KB, 602x314px

>>59769435

You do know Zen scales all the way down to Jaguar levels now right? So when the new Xbox scorpio talks about "Modified jaguar" It's very likely talking about a low power Zen core.

>>59769525
>It's very likely talking about a low power Zen core.
No you fucking retard.

>>59769525
>It's very likely talking about a low power Zen core.
Fucking hell, no. Zen is completely different.

>>59767957
the marketing and iconagraphy, jesus they went overboard to make it more interesting. Kinda funny.

>>59769152
>Krzannich delayed a node

That's not how it works, Intel has issues getting it yielding properly, the CEO has nothing to do with it

>>59767957
https://youtu.be/N0QrGZSary0?t=1005

"We read every single one of your comments"

RYZEN BACKDOOR CONFIRMED

HARDWARE SPYWARE INCLUDED

File: IBM-Power8-Core.jpg (83KB, 796x597px) Image search: [Google]

83KB, 796x597px

>>59767957
Kind of disappointed they won't do an IBM and build a 5 GHz 8-way SMT 16
wide monster core for Zen 3. A man can dream though

>>59770031
That's a 600nm+ die AND tops out at 22 or 24 cores, can't remember.

Though shitload of threads from SMT

>>59769332
they still haven't paid.
they owe half of yuro a couple billions and amd a couple more since ~2005
AMD has certainly used that iou as a collateral many times already, but they still haven't seen the color of the money yet.

>>59770349

They're fighting it in the EU and are very likely going to win against paying it.

And people ask me why I'm an AMD fan boy. Because there's no other company in existence who is as familiar as being fucked over by corporations and government entities like AMD. As someone who routinely gets fucked over it's inspiring to see them battling as hard and as well as they are.

>>59767957
> not having an adult job where you earn lots of money but still living in your parents basement so you don't have to care about electricity costs
You are quite literally cucks.

>>59772684

Heat.

>What does he mean by that?

>>59769525
they probably inserted some ZEN IPs, but nothing major
they have to keep structure the same for compatibility

>>59767957
The current Ryzen's are literally the Athlon XP's, with zenver2 being Athlon 64
The only part missing is Intel going full Netburst and going back to quad pumped ALU's and 50 stage pipelines
>>59768810
>>59768684
Dumb Intel shills
>>59768778
Any R7 can run with a cTDP of 35W senpai
They're efficiency monsters
>>59772668
>They're fighting it in the EU and are very likely going to win against paying it.
I miss when the EU wasn't immersed so much in corporate cuckoldry, even US lobbies are tame in comparison nowadays
>>59769152
No, he thought 10nm would just werk, even though he made the wrong call with 14nm
Intel can't keep the node lead nowadays, he should had accepted that and pursued other ways to keep a lead over the competition
Instead he keeps doing the same fucking thing, pretending engineers will shit rainbows and make the fabs work just because investors see him as a decent CEO since Intel has kept decent revenues in his tenure

Q1 earnings should be REALLY interesting.

>>59768778
Why buy amd when the more powerful I7 runs on less power?

>>59772970
Can't read, huh?

>>59772919
dumb ayymdpoojeet

>>59772735
What is a room fan and central AC?

>>59772995
You aren't getting paid with such low amount of effort, also stop smearing shit on the keyboard

>>59768968
A lot of my friends already have an amd laptop, so i believe their share will grow with Ryzen

>>59773074
I feel sorry for them. Must suck being poor and in constant fear of 3rd degree burns.

>>59773163
You'd be surprised. I have a used A10-4600M laptop I got for $300 on ebay and overall I'm pretty impressed desu.

While encoding video, the cpu stays at 2.7GHz. BF4 nets me ~40-50 FPS on low 720p settings in most maps.

Granted I only get ~3 hours of portable battery life but the battery is used and I can't really complain since I got it so cheap.

>>59773008

Jesus Christ.

Less power means less heat means less cooling which means smaller sizes which means denser server racks which means more performance.

If you're too much of a fucking faggot to consider the possibilities of performance per watt get the fuck off this board you normal faggot.

>>59774759
Hahaha, I bet he's the one making those ARM server shill threads.