Zen 7nm 16cores 15% IPC gain

>if you cram 2x the core amount in it

Meanwhile in reality
>2700x reaches 4.2GHz
>2990X also reaches 4.2 on fucking air
Where's your logic?

NEVER, EVER THREATEN INTELAVIV AGAIN OR YOU WILL SUFFER CONSEQUENCES THE LIKES OF WHICH FEW THROUGHOUT HISTORY HAVE EVER SUFFERED BEFORE. WE ARE NO LONGER A COMPANY THAT WILL STAND FOR YOUR DEMENTED WORDS OF CORES & TDP. BE CAUTIOUS!

Attached: 1528343393510.png (900x980, 104K)

you're comparing a cpu with a single die with one that has 4 dies, the heat/area is exactly the same
it's different when you have 16 cores on the same die, even if it's 50% more power efficient

IBM Nazi black magic

Isn't AMD producing Zen 2 at TSMC?

No, just Radeon 7nm
Probably Epyc as well? I think

>Epyc
What a gay name. Xeon is clearly the superior name.

INTEL IS FUCKING FINALLY FINISHED
(((THEY))) WILL NEVER RECOVER
INTELSHILLFAGS BURIED AND DEAD

no, only the gpus are at tsmc
it's not that easy to port a design from one foundry to another

>Xeon is clearly the superior name.
|
|>
|
|3
|
|

AMD? More like GAYMD.

>not just
>
3
come on man, efficiency

Attached: 1528222249083.png (1196x676, 1.05M)

OY VEY! You really showed to these goyim!
Good one, fellow hand scrubber!

תודה לך חבר

גויים מלוכלכים

GPU's might be done by TSMC, GoFlo will do CPU's.

Attached: 1232353465345.png (657x527, 226K)

>being this upset ebyn is superior

Attached: 1498254502257.jpg (2048x1536, 1.49M)

Ryzen 2 is going to be AMD's Sandy Bridge, screencap this.

Attached: 1487297768743.jpg (960x878, 125K)

Who cares?

You're an idiot if you buy a new cpu every single year. My Current AMD CPU will last me easily 5 years+

don't worry, everyone with a brain knows

DELID DIS

Attached: 1497881424635.png (5098x1500, 2.29M)

But EPYC/Ryzen is more power efficient than Intel. And has roughly similar levels of IPC. Intel's last advantage is the roughly 10% clock speed advantage.

Does that mean AMD is superior to Intel given they can air cool 32 core threadripper @ 4 ghz?

>2018
>add more bingbus and overclock it to 5GHz

Attached: 1506427974891.jpg (940x433, 89K)

Sidementalityposting a shit.

> tfw still using a 2600k and will probably keep it at least another 4 years
Who comfy here?

>Those jew hieroglyphs

Top kek wasnt expecting that

I expect it to have 12 cores (hexacore CCX) and up to 10 % IPC, but more likely 5 %. It will still clock higher, at least 4.6 GHz due to the new process.

How can Intel EVER beat AMD at moar coars when AMD started the coars race?

Attached: 1449058731632.jpg (618x696, 79K)

no one said it was or wasn't more efficient than intel, but it gets to a place where you have to choose between cores or clocks on a die if you don't want to build a housefire
you can't make clockspeeds better while doubling the core count whithout having a big power/heat penalty, even if you're going to a better node

>16 core on the mainstream platform
Literally why. Just make relatively high clocking 6-8 core procs for the mainstream platform.

No Spectre Inside(tm)
No Meltdown Inside(tm)

>MOAR NIGGAHURTZ
Just wait, in 10 years or so you'll have 10GHz CPUS...at 9001W

>process claiming 10% faster clocks made the CPU go from 4.1 to 4.35, a .25 difference
>but the process that claims 40% faster clocks will only make the CPU go from 4.35 to 4.6, a .25 difference
logic

>40% faster clock
>6GHz
yeah not gonna happen

>No Meltdown Inside(tm)
I thought that Meltdown affected both AMD and Intel. Isn't it only Specter that gets Intel.

>Glofo is lying on their spec sheets

>MOAR COARS
Yeah because that's exactly what we need in this market segment that can barely take advantage of 6 cores as it is.

another pajeet that writes "tech" articles.
fuck those shitskins... don't even know what and how IPC works
you double the cores, you get +15% IPC uplift.
you went full pajeet.
never go full pajeet.

More like they don't know what they're talking about

Blame the software developers for being lazy pieces of shit. Niggahurtz race has barely doubled in speed in 20 years.

>silicon foundry talks about the silicon process they're developing without knowing about its properties

Other way around
Meltdown and specter effect both but meltdown is specifically for Intel. Specter takes someone who knows exactly what their doing to use the exploit.

>effect
REEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE

>6GHz
>6GHz
>6GHz
Sorry but you're delusional and completely retarded.

Its realistic. I expected to Zen+ to be 4.4 at maximum, but more likely 4.2-4.3, which is what happened. I'm pretty happy with it. It matches my max OC 5820K which was a worse OC, but consumes much less power and is 40 % faster multicore wise. Ryzen is really efficient in low power states compared to Haswell or older CPUs, especially the HEDT ones.

>muh cores
>still gonna lose to Intel in all relevant metrics

If they could get threadripper chips to shut down at idle and light use that would be amazing. Right now its just twice the idle of a 1800x for no benefit.

>AMD doubling core and thread count
>while Intel removed HT from all CPUs other then i9™
AHAHAHAHAHAHAHAHAHA

>muh 5 extra fps in gayming

>games
>relevant metrics

תא גזים

OH NO, 5000$ i9 will have 20fps more than 500$ amd cpu in skyrim hd remaster

245 fps instead of 265 fps oh no

3570k here, waiting for 2400ge to have confirmed iommu working and cheaper ram to upgrade.
>50w machine that can host a headless linux host with win8 guest native gpu when gaymen and can plex 4k mkvs from a rock64 nas

>2400g
I was gonna build a freeb one of those for my cousin who is still gaming on a fucking A6 apu. But fuck $70 for 8 gb of ram

i9 is such a housefire that even thicc gayming notebooks with giant venting ports get maximum of 15% performance increase, and I am talking about MSI turbo jet, noise n' vents edition. For any slim mainstream notebook it's an overkill of the cooling capacity. All AMD needs to do is not to produce a housefire while maintaining decent gains. Then it will be a cheaper CPU with better or at least same performance minus housefires. Pair it with Vega and you have a perfect APU for 13inch notebooks with GPU much better than what Intel can offer.
>notebookcheck.net/Some-Intel-Core-i9-laptops-are-not-running-any-faster-than-a-Core-i7.317268.0.html

That good or bad?

If AMD makes a smaller but incredibly more efficient die, then it's all over. They could probably scale that to 128 cores, Intel needs to get rid of its monolithic design and fast.

of couse it isn't 6GHz you bunch of fucktards, it's 40% over 14LPP, not 12LPP
you idiots

remember the difference between nehalem and sandy bridge?

much more likely to have 3x4core ccx since they don't have to completely redesign the interlinks for the ccx itself in that case.

>look it up
>14LPP is targeted at ">3GHz" operating frequency
>this is for 7.5T
What the fuck

what does that even mean?

Attached: 1527517259377.png (211x239, 3K)

oy vey shut it down!

Attached: 3278347578457845.png (453x557, 50K)

It means AMD picked a process targeted for mobile and datacenter applications for Zen...which actually makes sense since the focus was the server market.

Basically there's no way Ryzen 3 won't see 4.5GHz+ base clocks and boost above 5GHz, considering they got 3.6GHz base out of a process targeted at 3GHz and 7nm is targeted at 5GHz.

going by the spec sheets, 7LP is 6T

Yes, so if AMD does what I think they'll do and design 16c dies for EPYC/TR using 7nm 6T and 12c dies for Ryzen using 7nm 9T EPYC will rape Xeon in power efficiency and Ryzen will rape Core in frequency (and with a small IPC boost, in single core too, which considering that AMD gets more efficiency out of each additional core and more efficiency out of their SMT implementation as well, would mean Core is going to be behind in every workload).

it doesn't look like it has a 9T option, only 7.5T and 6T, unlike 14LPP, which has 10.5T, 9T and 7.5T

7LP is also not in production yet. Maybe they only use 7.5T or maybe they still stick with one die and use 6T for all of it, my point still stands. I would be surprised if the top-end Ryzen 7 has a base clock lower than 4.5GHz.

I should clarify, 7LP is not in production so their offerings may change.

I think the future is going to be a cool place (for some things)

>7nm
Intel can't even do 10nm yet.

Attached: 1528445007753.jpg (691x771, 112K)

Simply wrong.

3800X:

205mm2 die size
5-7% IPC integer,
8-11% IPC floating point, native 256b AVX
8-core CCXs,
32MB L3$ using the same structure,
2Ghz Infinity Fabric,
PCI-e 4.0,
200-300Mhz stock clock increase with remaining power budget,
110w TDP.

fite me

If we forget nomenclature and standardize by announced feature sizes,
Intel's 10nm is worth 8.6 real nm. IBM/GF/SS's 7nm is worth 9.5, TSMC's 7nm is worth 9.3

source: my ass

Good, of course.

Who is onboard to save money and buy cheap used EPYCs in the future?
One of those will be literally 10 years of comfy computing.

too bad ST performance on them is bad
also, I think a cheap used 2990X is going to be cheaper than an used 7601

How bad is it?

the max Turbo on it is 3.2GHz, a r3 1200 has a max turbo of 3.4GHz

It isn’t that bad, they are server chips. Everything about them is made for highly multithreaded server workloads.

What about 7 nm EPYC?

higher, but no one outside of AMD knows exactly how much higher
It'll probably be considerably better than the 14nm ones, for sure though

Making fake specs is quite a fun sport.
I miss the times everyone created their own nintendo revolution.

Well, I would save for these.

We'll get 15% more performance, not 15% IPC.
Improved IF and IMC is the lowest of hanging fruit to achieve that, atop generally improved/larger/wider cores. The weak FPU is the only thing stopping Epyc from taking over x86.

I ran some numbers after shopping a single Zeppelin die into a 16-core unit with larger FPUs.
I measured the resulting fake "Zen2" 14nm die by dividing the new image dimensions by the old dimensions, and multiplying by 213mm2.
I then used GlobalFoundries' announced scaling metrics for logic and SRAM, and borrowed an Anandtech forum user's calculation for I/O scaling (borrowing from Intel)

With a 95% confidence I declare a single "Zeppelin 2" die will be between 198 and 220mm2
You can call bullshit all you want, I do statistical analysis for a living.

No

oh really? and what if that 16c rumor is (obviously) wrong then?
>15% higher performance
lol

ok kid