this kills Jow Forums
This kills Jow Forums
not really, she's a great businesswoman and lifted AMD out of poverty. who cares anyway.
>engineers create something
>lisa takes the credit
Stupid bitch can’t even speak English correctly
Typical American comment
>engineers who haven't made anything good except under Lisa's leadership
Whoa i wonder what it is
LISA MOMMY PLS GIB MILKIES
You're thinking about raja
what? Jow Forums loves Lisa
Nothing good happens without good leadership. That being said, my next build will be intel because their products are still better.
I'll hold my judgment for when the competition is back on again
right now I could just as easily chalk AMD's success up to Intel's hilarious incompetence and complacency
>not pictured, Jensen Huang
mommy
Better at what? Lighting your house on fire?
we will all drink from her milk on the 7th of July
baste
I enjoy this metaphor
youtube.com
Nigger please
she still can't save RTG, is she not involved? it seems that way, too busy with zen
IPC.
youtube.com
better than 99% of americans.
HAHAHAHAHAHAHA
Imbecile.
Cope harder poor fag. You're only fooling yourself.
/v/ tourist?
Oh no no nononono. We can’t let it end like this, intelbros
Kék
Intel has worse IPC as reflected by their higher frequencies
Kek.
Let's go over the facts, shall we:
Intel
>some vulnerabilities found, but ALL of them have since been patched
>Intel ME has been locked down, audited by security experts, and is no longer a security risk
>patches have NEGLIGIBLE impact on gaming
>storage IO takes a slight hit
>much higher clocks (better for older games)
>much lower core and cache latencies (very important for gaming)
>more industry-adopted instruction sets, better optimization in Unreal Engine, Unity, Frostbite, etc.
>uses less power-per-clock and has better frequency-to-voltage characteristics (why Intel will always hit 5.0GHz and AMD will never be able to)
AMD
>Ryzenfall and 12 other vulnerabilities STILL NOT COMPLETELY FIXED
>PSP has been fully compromised and gives attackers complete access to all parts of an AMD CPU; AMD REFUSES TO DISABLE PSP ON ANY RYZEN CPU
>has performance loss with its already lower IPC (turning a shit hatchback into a shittier hatchback)
>literally slower NVMe throughput by default because AMD can't into IO queuing optimizations
>lower clocks than Intel's First Gen Core architecture
>almost twice or three times the latency depending on which cores are accessing which cache; made even worse with multi-die Ryzens (like all of the Ryzen 3000s)
>literally no one optimizes for AMD's instruction sets; locked out of the instruction sets the games industry uses
>uses far more power per watt because Infinity Fabric eats power the higher you clock it (guess what was exponentially increased on Ryzen 3000...)
Other notes - PCIe Gen 4 literally doesn't matter
>storage IO loss literally doesn't matter for gaming because 99.99% of games doesn't come close to using a tenth of an NVMe drive's IO throughput
>4K 144hz gaming doesn't even saturate PCIe 3.0 x8
She put Joanne Plummer in charge of Radeon who was a lead engineer on Zen 1.
>not the callarypasta
hardware isn't their problem, it's marketing and doa strategies that make zero sales with good hardware.
>(x) literally doesn’t matter
Ebin
>factspasta again
You get out of here with your facts. Go on git.
WE GOT COCKY INTCEL BROS
let it go, user
believe in MOMMY
She's an engineer herself and is Texan.
why does it kill Jow Forums?
weve been worshipping lisu su and amd for the last 2 years.
God, so many comments about how some F1 presenter should
1) know who the CEO of a semiconductor company is
2) assume that a Chinese-looking woman in Germany speaks English.
Why are people so retarded?
Wrong. GCN is garbage. Navi is GCN with double precision stripped out and more emphasis on rendering faster, not wider like Vega. 5700XT has less shaders than Vega, but it's got faster shader bandwidth. This makes it use less power.
why. if s(he) really is a biological female she deserves the praise for having done an excellent job.
whereas kiketel splurged billions on diversity she focused AMDs limited resources on proper engineering by white and east asian men while ejecting the streetshitter that was ruining the graphics division. 2 years and amd will rule graphics as well.
nothing but respect for mommy su.
Management is also important user.
I think AMD were in trouble a few years back and some key mgmt decisions reversed that situation. She also used the sword on intel when she needed to.
>he doesn't know
Let's go over the real facts user.
My computer uses an intel CPU.
It is slower now than when I bought it because it was incorrectly made the first time.
That's all people care about.
> years and amd will rule graphics as well.
> Sony is AMD
> MS is AMD
> Stadia uses AMD GPUs
> Samsung licenses patents from AMD
> Intel licenses patents from AMD
> Apple uses AMD GPUs
Why the fuck would she go against the powerful desktop/datacenter enemy which has concentrated all of its money and R&D just in that segment? More money is always better, but this standing looks like a money pit.
Bait.
>gaymers
LMAO
Face it Amd is competing with Haswell for single thread and IPC. Admittedly Amd has better hyperthread performance, but 16 kids with novelty hand shovels wont beat 4 day laborers with spades at clearing out a sandbox
VMs run better on brute strength vs limpwristed acrobatics.
Navi isnt GCN. The single fact that the 4 cycle latency and 64-wide wavefronts are gone is proof the ISA itself had fundamental changes made to it. It's hybrid only in the sense that j
They somehow worked in compatibility/emulation for GCN functionality
Momma Su in here rightful position.
oof
Jow Forums has always been an amd board because Jow Forums in general is contrarian by nature but our trust is finally paying off
>engineers are finally able to create something good because the CEO actually understand what the fuck they're talking about
>CEO is not in any way responsible for that
That doesn't mean it's not based on gcn. That just means they have streamlined everything from Rajas crap being slow and wide and huge and hot.
it doesn't matter, any of it doesn't matter from business standpoint. it has competitive performance
it doesn't have competitive marketing. nobody wants the thing.
No valid retorts yet.
No one want to say the boring "the AMD IPC is higher now".
WHAT DID HE MEAN BY THIS????????
??????????
?????
It does matter. Because it's faster than the competition while being cheaper and consuming less power
GCN is an ISA, not any one specific hardware implementation.
Long ago AMD filed patents for a GPU arch which was a derivative of existing designs but with variable CU size, more scalar units per CU, improved schedulers. They planned to have a mix of full 4x16 CU, 4x8, 2x16 etc all in one GPU with schedulers intelligent enough to distribute the workload where it needed to go for the highest perf/watt.
Navi is clearly evolved from this but instead of varied CU sizes they simply moved away from 4x16 to 2x32 to increase perf/watt and power per op.
she's also a pretty good VLSI engineer, she got her BA, MsC and PhD at MIT
>tfw you will never wake up in the middle of the night to Mama Su staring down on top of you, thrusting into your butthole
She probably has more education than you'll ever have
>Rajas crap being slow and wide and huge and hot
This is also rajas crap. A brand new uarch like this would take 5 years or more to make according to what people like anand have historically reported. 2-3 more for planning, design, and verification, then next several years in driver development and getting real die
>she
>engineer
you do know that the very reason cpu's broke the speed wall was because of her right?
she literally invented a way to use copper as pipelines without having to think about quantum tunneling
Holy fuck one cherry picked test.
Are you sure you're not just remembering something? Gcn has 4x16 wide SIMD that compose the "wavefront". A variable CU gpu sounds bogus AF and would probably be wasted resource hell
Misremembering
>t. can't even write a single line of HDL
This kills AMD
Protip: check the clockspeeds
They totally have this ready for 2020 on 10nm just you wait. It's not like they have nothing and decided on 9900KyS, nothing of the sort.
>not coming to desktop until late 2020 at the earliest
>won't see clock speeds anywhere near Coffee Lake
>yields are significantly worse than 14nm++++++, so prices will be even higher
And that's assuming this is even accurate, considering there is no documentation of the test systems or verification of this result anywhere, and it comes from WCCFTech of all places. That's higher than even Intel is claiming in their marketing slides.
>3:58:38 No.71469410
>3:58:37 No.71469411
wut
M-mods?
Oh ye clueless babs.
freepatentsonline.com
I don't misremember anything. Ever.
Again, GCN is an ISA. The execution width is not a static feature.
Intel's had a really rough time improving IPC by more than 3% per generation for the last, what, 4 generations?
Its' 10nm process was a complete shitshow by all accounts
You're claiming that somehow combining both of those factors they whipped up a product that delivers results like that?
They did some public demos, their biggest performance uplifts come from shit like WinRAR.
>AMD had a really rough time for 10 years
>Its single core performance was a complete shitshow by all accounts
>You're claiming that somehow they whipped up a product that will compete with Intel?
Sure it's an early leak, but keep in mind this is a laptop SKU so clock speeds are naturally lower.
Desktop products will probably be on par with Zen2 clockspeeds, meaning mid 4ghz.
Zen was a complete ground-up redesign. It achieved roughly the numbers you're claiming Intel got from a process improvement and minor microarchitecture improvements.
Not even the other users of Chiphell believe that chart.
>completely new design vs zombiefied skylake for the 5th time
So for consumer CPUs AMD can finally compete with Haswell?
Ryzen 2xxx series already competes with Skylake in single thread and Coffee Lake in multithread. But you knew that.
IPC on Zen 2 is already up to date, it was the clock speeds that weren't stellar
>better than 99% of americans.
That's not exactly hard.
WCCF
Are the biggest AMD nutt garglers on the web. If they are showing AMD getting beat by something, then its fucking legit.
>Ryzen 2xxx series already competes with Skylake
Before or after IF latency?
I was actually mistaken, that image comes from a Chiphell thread which has since been deleted. WCCF was just the first site to report on it.
>single thread
Jesus titty-fucking Christ you shills are dumb
>retard doesn't realize Lisa is a engineer
Try harder.
Although not a new ground up arch, it greatly differs from past process refinement/refreshes we've seen from Intel over the past few generations.
Also all of the previous ones were on 14nm while this is 10nm.
>implying this hard
>engineer
>woman
Try harder
>it greatly differs from past process refinement
How? Intel hasn't really detailed any of this. They have also claimed similar improvement numbers from architectural revisions in the past.
She legit worked doing chip design for Texas Instruments.
>Working with various IBM design teams
right
Read the wiki and fuck off.