wccftech.com
>~90% increase in overall performance than AMD's last generation of APUs
>This literally kills and incinerates Inturd's Jewish attempts at kikening out a piece of integrated graphics that will ever be able to compete with AMD's offerings
JUST.
Other urls found in this thread:
servethehome.com
browser.geekbench.com
browser.geekbench.com
anandtech.com
browser.geekbench.com
i.imgur.com
twitter.com
The niggerfucking OEMs will still gimp it with single channel memory.
I hope AMD is not skimping on L2$ for DSBR.
Come te fuck on AMD
Just release this bad boy so I can pick up a new laptop ASAP, can't stand this piece of shit anhmore
They can't; the software for consumer Vega is not ready.
They've recently pushed DSBR into Radeonsi specifically tuned for RR.
It's coming Q4 2017.
It was already confirmed that no gimping will be allowed and will be directly supervised by AMD, considering the long-lasting big Zen contracts made with Dell, HP, OriginPC, iBuyPower, Alienware, Lenovo, OLDI, and Maingear. All of these make notebooks, not just custom pre-builts, and AMD sits heavily on their asses right now as to not allow for any of such shit to happen. The QA on AMD's side is insane, even when it comes down to watching over third parties.
>wccf
This is not a meme, Sup Forumsoy, they've became quite accurate lately. And from time to time they also simply just repost what's already on other outlets, like in this case.
Say his name.
Say it and say it LOUD.
You KNOW you WANT TO.
HOLY FUCK AT LAST.
A275/A475 refresh can't come soon enough.
Will motherboards come out with many sata ports? I want to build a nas with this.
>poobench
Bristol Ridge was FX garbage anyway, and all I always personally cared for in integrated graphics is only the ability to run something like PCSX2/Dolphin or at the very least ePSXe & PPSSPP while everything maxed out with stable 60 FPS and not turning my notebook into a fucking holocaust oven within ~10 minutes of play. No integrated graphics is currently able to fully accomplish this on all fields, not even 5775C's Iris Pro which up to now was considered the most powerful Intel's integrated video out there. Zen's APUs will be the very first ones which won't be only very cool and power efficient, but will also perform greatly enough as to allow 100% comfort Sup Forumsoyming in 1920x1080 or even 2560x1440 resolution. I cannot fucking wait to try out the most top gen1 Zen's APU offering by playing either Witcher 2 maxed out with and without Super Sampling while riding a train, OR by running modern Sup Forumsoyming Emulators with MAXIMUM GRAPHEN on it. I have almost no doubts that top gen1 Zen APU will be able to run all of this with relative ease while staying cool and efficient. Even if it'll just be a 1920x1080 notebook at 15" and 60Hz with IPS, this will be more than enough for me personally. AMD APU's are definitely "shit's so cash" right now.
> pic related is a 3 year old chip
> lowest end Crystal well
> Iris pro graphics 5200 (128MB L4)
will it be a botnet tho
>HQ
We're talking 15watts here.
Autism too hard there buddy
What's wrong with pre-built NAS from third parties?
Not all of them use Infail's garbage Atom's and other inept POS in them, you know? Even right now you can find decent third party NAS with AMD processors in them, and when first Zen APUs start coming out it'll be just a matter of months before first third party NAS solutions based on Zen will get on sale.
>geekbench
>ignoring that there are 2500U results with 25% more performance than that
Look at the frequency of that Zen APU, you dumb down syndrome-induced retard. It's an engineering sample, you stupid faggot. This shit will be at 2.8~3GHz by the time it goes gold and you can imagine what scores it'll get when that happens. Learn to fucking count, Brian.
It's not even it's final form, though. Just an IS.
I wonder when will AMD get bored enough of no competition to integrate HBM2 into the package.
Who cares, the CPU will be close to Intel, but the GPU will demolish Intel.
This is Vega 8, not Vega 10 even.
I wonder about the size of L2$ on these little boys.
>I wonder when will AMD integrate HBM2 into the package
Zen 2. They're already finishing it up by the end of this year and it's slated for release somewhere next summer or autumn. Absolute holocaust.
>only 4MB of L3
But why.
Though AMD uses 512KB of L2 (and it's 8-way) so the hit ratio must be decent.
Never. Low-cost HBM is more likely, but maybe not even that because of the size of the package will be disruptive for laptop motherboard form factors. Laptop motherboards are very dense so having a large package will piss people off.
>Never.
Oh i wouldn't say that.
>the size of the package will be disruptive for laptop motherboard form factors
And yet no one complains about HUGE slab of eDRAM that is Crystalwell.
>the integrated video of gen1 Zen's APU will demolish Intel
That was the point to begin with, though. Just like OP-post says - the "~90%" performance increase is overall, meaning on ALL fields, not just computation or just video, but BOTH. I presume with a pretty big portion of confidence that using top version of gen1 Zen's APU will be something like putting a full-sized i7 2600K (after all, 8 core RyZen without SMT is LITERALLY "two 2600Ks" by sheer PPC, and gen 1 Zen's APUs will have only 4 computing cores at the start) in your notebook + new instruction sets + less power consumption and lower thermals, and then adding a GTX 1050 Ti/RX 560-level card over that to the boot. And all of this just from the CPU alone. This is the most realistic course of action in my personal opinion. But I actually think it might turn out to be even better than that.
>And yet no one complains about HUGE slab of eDRAM that is Crystalwell.
Because Intel still subsidizes their chips and development of products. AMD doesn't do that and can't afford to do that, so they don't have the power to tell OEMs to shut their filthy peasant mouths and know their place.
This is tested with a single memory channel.
Check the AES and Memory bw scores.
Engineering sample. A literal "proof of concept" stone.
>they don't have the power to tell OEMs to shut their filthy peasant mouths and know their place.
They did exactly that with EBYN.
OEMs were eager to comply.
Sandbagging? Nice.
Or the R5 is only 4MB and R7 is 8MB to differentiate a bit more between them.
Both are still 512KB L2@core. Zen lives and dies on bretty large, low latency and fast L2. It's quite the opposite of faildozer (L2 killed it).
>they don't have the power to tell OEMs to shut their filthy peasant mouths and know their place
Except that this is EXACTLY what they're doing right now by heavily supervising OEMs for not gimping out pre-built systems and notebooks based on Zen APUs via single-channeling. Long-term contracts, bitches.
L3 is one thing, but cutting L2 would be too destructive to the performance.
I don't think it's even possible to cut L2, it's not made into subslices like L3 is
Ye, long-term partnership with chinknovo is that.
>L2 killed it faildozer
Wrong.
What Keller'd FX is FPUALU-trash architecture and artificially inflated corn pseudo-Hz. FX was doomed from the very start simply because by sheer IPC and PPC there was a formula in place which literally translated as "3 FX cores = 1 Sandy core", and this was due to FPUALU approach. They advertised FX APUs as 4 core solutions while IN REALITY there were only 2 actual cores, same with desktop - FX's pseudo-8 cores were actually 4 in reality and pseudo-6 core stones were only having 3. And now compare Sandy's (for example) 4 TRUE ACTUAL cores to FX's 4 "pseudo-8" cores...this battle was lost even before it began, because FX's IPC and PPC were absolute garbage shit and just 1 (ONE) Intel's Sandy core LITERALLY performed like 3 FX cores.
No you fucking retard.
4-wide decode per 2 sets of ALUs and horrendous cache latencies killed Bulldozer.
Stop pretending you know anything.
Not just Lenovo. There are a shittion of big OEM/third party players contracted with AMD on Zen right now, see here .
>Only 4 out of """""(((8 supposed)))""""" cores were working at one time, with other turned off
>FX could never turn all cores on simultaneously, only working on a "flickering" methodology
>WAAAAAH, U LIE! THERE WERE 8 REAL COARZ I SAID!!!!
Amusing.
>Only 4 out of """""(((8 supposed)))""""" cores were working at one time, with other turned off
>FX could never turn all cores on simultaneously, only working on a "flickering" methodology
Bulldozer is not a barrel processor.
You've just described Sun's Niagara.
Please stop shitting in this thread you tech illiterate moron.
If you want some delicious RF ID chips alongside with your CREAMY EQUINE MAYONNAISE under the lid - AMD is not the fast-food joint you should go to.
Faildozer was EXACTLY working like that, you DUMB fuck. Entire FX family works this way. There were only 4 real cores at one time, because ONLY 4 cores were working at any one given time, while others disabled themselves automatically. This shit was so bad it was literal "monitor panel PWM"-equivalent of CPUs. FX flickered out of it's goddamn ass ALL the time, because it was BUILT that way.
No you fucking retard it's not a barrel processor.
Stop shitting this thread you tech illiterate moron.
Alright, fine, I admit your defeat. Go and educate yourself, you reality-denying inexperienced imbecile.
>abloo abloo Bulldozer is a barrel processor!
Please stop.
fucking this, ishygddt
just make a small machine with dual channel, i want a new laptop
INTEL BANKRUPT
INTEL FAGS ON SUICIDE WATCH
>just make a small machine with dual channel, i want a new laptop
SOON, just wait
THANK YOU BASED JIM
APU'S are CUTE!
>fx flickering
Dude stop.
You're being too technical to understand you.
Well not that Xeon-SP and KBL-R are dead it AMD has CFL-S to deal with.
I think they'll a round of price cuts the day Coffin Lake launches.
>taking wccftech seriously
LUL
i want to succ lisa su
THANK YOU BASED JIM
A-CUTE
>it's the stupid faggot who thinks Intel's $400+ spaceheater mobile CPUs are the same shit they put in $500 OEM craptops
...
servethehome.com
Even the cheapest of EBYNs is doing fine.
What are these third party nas with AMD processors? Really intrested here. I don't like giving my money to those kikes at this point arm is better.
And you know what's the worst part of all the shitfest?
Intel damage controlled so fucking hard instead of staying fucking silent.
There's QNAP ones with full fat Ryzen 7's.
To be fair EPYC BTFOs Xeons in 80% of use cases. They see the dollar signs. While Raven Ridge is very good, so far it doesn't look so much better that AMD can slap people in the face with their dick.
>To be fair EPYC BTFOs Xeons in 80% of use cases
So does RR.
iGPU alone makes it ages ahead of anything Intel.
...
>over FIFTY
>SKUs
>from THREE dies
HOW
anandtech.com
Meanwhile Ice Lake will have 8c/16t a year from now, and you know what that means.
Just to show that the R7 2700U is alive, though a INVALID
>From THREE dies
I have very bad news for you - i.imgur.com
No-no, i mean Xeon-SP.
It's THREE DIES.
AND OVER FIFTY FUCKING SKUs.
THEY ARE SEGMENTING EVERYTHING EVEN MY FUCKING ANUS.
>that multicore score
?
BASED
AMD
JIM KELLER
JEW KILLER
>>that multicore score
>?
Bugged obviously, it's an ES
>?
ENGINEERING SAMPLE™
JUST WAIT™ WHEN IT GOES GOLD
Why is it so much faster on x86 android?
They would be stupid to cut their products performance in half when Raven Ridge will actually be good.
Still have no idea why they had single channel APU products before though.
>No-no, I mean Xeon
>He doesn't know that Inturd's "desktop" and "XXTREME ENTHUSIASSD" processors are LITERALLY cut Xeons, LITERALLY same shit just BACKWARDS
No, i mean actual Scalable Family Xeons.
OVER FIFTY
50
SKUs
FROM THREE DIES
THE SEGMENTATION DOESNT MAKE ANY SENSE.
>Still have no idea why they had single channel APU products before though.
*smiles in shekels*
Why do you write like that?
Nothing stops me from doing so.
Oy vey
>8 core Intel mainstream
>you know what that means
Of course I do, Brian. We all definitely do:
INTEL MOBILE HOUSEFIRES
I want my GPUless ITX box and I WANT IT NOW
>Mobile 8 cores
>Intel
Pick one.
45w SKUs.
>Meanwhile at AMD
>~15W
>ha ha we have a node advantage
And then 7LP comes in and tears Intel a new assholes.
I have no idea how 7nm LP will do for low power since it's an IBM node.
It's >60% lower power as ISO frequency or >40% more performance as ISO power.
It's as good as nodes can get.
...
That means 60% lower power at 3.0GHz? (no zen2/+ improvements)
>It's as good as nodes can get
Except that they're already working on going 5nm and 4nm. If it'd had too extremely diminishing returns or was unfeasible, everything would've stopped at 7nm, but no they didn't.
Exactly, the stats are compared to 14LPP.
And it's >60%.
Probably ~70% lower power for non-EUV 7LP.
I mean as good as single node shrink can get.
Denser than Intel's 10, with better performance, perf/watt and yields (7LP has lower mask count compared to Intel's 10) to boot.
Except that's typically at target frequencies, which is 5GHz or something for the 7 leading performance node. Scaling down usually doesn't turn negative but it can certainly be less than 1:1, which is already mediocre considering scaling down on Zen becomes >1:1 after a certain point
At ISO frequency, user.
It's >60% lower power at 3ghz, which is 14LPP's sweet spot.
But they plan to seriously put Zen 2 at 5GHz as a "mainstream" frequency.
You do understand that mobile offerings clock lower?
Intel's 14+ can achieve nuclear clockspeeds yet mobile chips don't do that.
