AMD Launching Ryzen On 28th February During GDC 2017 - Vega GPUs To Launch in May

wccftech.com/amd-launching-ryzen-cpu-gdc-2017/

wccftech.com/amd-vega-gpu-radeon-graphics-may-launch/

Other urls found in this thread:

youtube.com/watch?v=tK_C8NAchko
ark.intel.com/products/93790/Intel-Xeon-Processor-E7-8890-v4-60M-Cache-2_20-GHz
pcpartpicker.com/list/kkRhYr
twitter.com/SFWRedditImages

Cash ready awaiting benchmarks.

THANK YOU BASED LISA
THANK YOU BASED KELLER
THANK YOU BASED RAJA

Are you ready for Another Massive Disappointment? :^)

What-what-what?

go fuck yourself, amd

i have been waiting for too long, and now you say two months more? gtfo outta here

Only idiot weaboos would fanboy this hard for such a shitty brand

oh hey
that's on my birthday

My body is ready
2017 will be Intel's last year of relevance.

>AMD stock is down from shortsellers cashing in
Pussies.
Should I buy now or will it sink below 8 before it skyrockets?

You should have bought it when it was at 2.

Don't remind me. I had $2000 I was ready to invest when it was at 1.3 and then life happened.

Meh, my i5 6600k arrives today, with an RX480 of course. Either way good luck with Ryzen.

based shitwrescker

...

>wccftech

>wont be able to do pcie GPU memethrough due to lack of integrated gpu on 8/16 cpu

meh, and single core performance doesn't look good either, i guess i will go with 7700k

>single core performance doesn't look good either
Says who? Everything official we have to go on shows IPC somewhere around Broadwell/Skylake/Kaby Lake level.

>tfw delayed until April 1st

>Polaris was aggressively priced against its Pascal based competitors.

the 480 was more expensive than the 1060 at release here in germany. that information is more a let off for me

Pajeet fucker lied about the price too, the 8GB 480 still hasn't hit $200 dollars.

4gb was $200. 8gb was like $240 or something.

...

everything I see you post is stupid

turn off tripcode cunt

I'm 90% likely to get a Summit Ridge build in March, but honestly I'm more interested in Naples.

Will there be 2S platforms for 16c/32t MCMs, or will 1S 32c/64t be the only option?
Zen seems to get decent clocks at lower Hz, but I'd rather stick closer to 3 GHz for a 32c total build, and I don't want to get capped by a single socket power draw limit.

remember that broscience bullshit few weeks ago?

It was never at 1.3. Don't lie

Please tell me about it :)

I already have my R9 380X for that.
>3.5W idle

Also, AM4 is ready for raven ridge APU's

32 core naples will likely cost 7-10,000$
Intel has no answer for it, both in memory bandwidth and just plain poreformance.

as for the rante, I would assume amd has at least a 16 core and 24 core opteron in the works, and potentially skews from 8 cores with quad channel (2 4 core dies) through to the 32 core monster, likely in 2 core increments, but possibly in 1 core depending how mcming the chips works.

basicly, through to the 24 core, amd is likely going to undercut intel, possibly only through 16 cores as more cores then that would either make it hex or quad channel for memory.

Say I had the budget for a current intel 10core, whats the likelyhood of amd allowing 2 ocable 8core chips on a single mobo?

Always regretted not going dual socket in the 920 days and while its probably stupid I just wanna do it with current tech.

I'm not sure if I can see AMD doing a 2S*8c platform despite the higher power/clocking ceiling.

Zeppelin can probably get much, much higher interconnect speeds on a shared MCM, and that trumps core clock speeds in a lot of tasks, since you can start striping/interleaving physical address space over all the memory controllers on an MCM instead of just the 2 on each die.

>Vega in May
>Probably won't be able to build a Ryzen + Vega build until August
Nah fuck it, going with jewtel and nvidia

Let me guess, your 2 year old system is just not providing that 2 minimum fps bump you need?

seriously. I'm an AMD fanboy, but May? that's ridiculous. it's a year after 1070/1080, and won't even perform as well as a 1080

my friend is literally waiting for Vega before pulling the trigger on a 1070, but with this news, he'll probably just go Nvidia

AMD needs to fix their release schedule, because they're losing to Nvidia and Intel in every category

>May 2017
>May two thousand seventeen

Woah Vega will be a ferice competitor to the GTX 2050 baby Volta

My build is from [spoiler]2007[/spoiler], the only thing I've upgraded is the ram and I put in a ssd 2 years ago. I don't play video games anymore really but I like to think that I would if I built a new one. Too busy/lazy/tired being a working goy

It's great to browse the web with though.

...

AMD has been slow as shit lately, but it's not very realistic to expect that Vega 10 would be beaten by 1080s.

Vega 10 will have 60% more bandwidth and 40% more fp32 FLOPS, so the only way it will even be close is if the drivers are completely broken or yields are so bad a third of the chip has to be turned off.

I still think that GP102 will be better, but not even strong overclocks of GP104 should beat it.

So ill more than likely be looking at hoping opterons can OC and buying a higher corr chip? I game some but mostly im slowly getting into video editing so the idea I had was rendering out on 2nd cpu while still having my computer be useable.

Right now my projects arent huge but with family/friends knowing in doing it I have had hints at a lot of holiday/group trip footage heading my way.

Oy vey. 2 shekels for this lot. Good work chaver!

Okay I exaggerated. Was a little below 2 though. We're talking like 3 years ago or so.

maybe, but I can't see it beating a 1080Ti, and we all know once the Ti comes out, the 1080s will drop $150ish and be around $450. who's gonna spend $650 on a Vega when you can get an Nvidia one for a similar price? it's like Fury X again: if Fury X competed with 1080, it'd win, but since it had to compete with 980Ti, it lost

Dude fp32 flops =/= gaming performance. It's good for parallel computational tasks, namely shit like machine learning.
I don't know how well vega's gonna play games (probably still pretty good) but it sure as shit is gonna grab the attention of enterprise markets.

>So ill more than likely be looking at hoping opterons can OC and buying a higher corr chip?

Who knows?
Intel still sells 8c (1.6/1.7 - 3.2/3.6 GHz), 6c (3.4/3.7 GHz), and even 4c (2.6/3.2 - 3.5/3.7 GHz) dual-socket Broadwell models, so maybe AMD will throw a few similar SKUs out and see what sticks.

Isn't naples 8 channel memory?

Head over to pcbg, post your build and maybe we can find some cheap upgrades without a full new system

Deep Learning is increasingly using fp16 (2x rate on Vega and GP100), and it's more scientific simulation stuff that uses fp64.

fp32 is in kind of a weird space in the middle where some financial models etc. rely on it, but fp16/fp64 algorithms seems to be bigger growth areas.

Doesn't matter to me at this point if Vega is late compared to Pascal, I refuse to pay the GSync tax so until nVidia supports FreeSync I won't be buying their cards.

yeah, just got verified in slide leaks this week.
obviously only true for the 32c/64c 4-chips MCMs though.

With alibaba going for ARM -- and possibly giving other major players funny ideas -- and if AMD delivers on Naples Intel will have a bad bad 2017...

Even if everybody sticks with x86, Intel's enterprise Xeon margins are going to absolutely collapse if Naples is even remotely as good as it looks so far.

Enjoy your nogainz

Amd will make a dual or quad socket chipset for workstation/server use, but they may also come along with one for non server cpus or even ask allow non server cpus to work with that chipset.

there is quite a lot of use for a dual cpu even in a non professional setting, such as streaming, recording video, or just the overhead of doing something taxing. but this would likely be a mid cpu cycle thing, not an out the gate.

do not get more then one cpu for video editing, its one of the fun areas where the programs refuse to see more then one cpu.

depending on the chips changes how many channels there are,
1=dual
2=quad
3 (if possible)=hex
4=octa

it has to so with how each cpu die handles memory.

youtube.com/watch?v=tK_C8NAchko

Problem with that is sockets, I doubt they're going to have 2/3 de parts if the boards are all octo channel memory, which is why I only expected quad channel memory so 2/3/4 die opterons could be used for a top to bottom range, I mean intel's only getting into hex channel with their new lga3000 which is going to sit above the also new lga2066. It's like amd have hit a niche that intel didn't feel at threat with this 8 channel memory thing.

Another user did mention the 16/32 opterons will be on a bga platform competing with intel xeonDs, wasn't the old 16core leak a low clocked part as well? would seem to tie in with that rumour.

Not falling for the usual retarded hype.

Reviews needed

>Intel has no answer for it, both in memory bandwidth and just plain poreformance.
ark.intel.com/products/93790/Intel-Xeon-Processor-E7-8890-v4-60M-Cache-2_20-GHz

funny post i saw in there.

So I just heard that Microsoft wants to reveal the Xbox Scorpio before E3, so before June. So maybe AMD and Microsoft are going to Launch their products together. AMD is already providing the hardware for a 4k Xbox.... I'd imagine Xbox will have some form of both Ryzen and Vega since AMD has been taking their sweet ass time.

thoughts on my asspull speculation?

only 24 cores
only quad channel

As the other user stated, intel has no answer for Naples, at least not yet. They'll be releasing a 32 core Skylake Xeon sometime this year to compete with AMD's new Opterons. Though it will probably be clocked even lower.

as far as i know, the ps4 pro is using AMD hardware as well, maybe Microshill sucked AMD off for better hardware, we'll see though.

microsoft and sony use booth amd software for their consoles

AMD did have a little Xbox Scropio sign at their CES booth, and it did have the Ryzen logo on it. AMD wouldn't comment about it, but its pretty obvious why it was there.
The new Xbone will have newer hardware in it. It'll be Zen based.

I'm pretty sure that's what i just said.

a lot of us are 6+ year old systems and were waiting on intel to drip 6 cores down to not retard price points, if not 8 cores, but intel jewed us hard.

my sticking point right now is just the cpu, so amd putting it out within 6 weeks, I can easily get it and if vega sucks, I can stick it out for another year or so on my gpu.

why would they bother making the hardware if no one buys it? this is the point most people can't answer me in a good way on. the market has never, every time they were better price and performance and again when it was just better performance rewarded amd. Why go out of your way with catering to consumers? put some shit out at a sweet spot price/performance and call it a fucking day, the high end can be a byproduct of pro parts, and that's it.

I can't understand why people think amd should bust their ass to make a good part just so nvidia lowers their price for you to buy nvidia, everyone apparently wanted a monopoly and then bitch when they get it.

there are no upgrades that are at best side grades. assuming his pc is from that era, he would best be served with getting intel's 60$ 2 core 4 thread cpus.

pcpartpicker.com/list/kkRhYr

the case is a bit cringe, but get rid of the led fan in the front with something not glowing, and it would be great airflow for the cheapest price possible.

24 cores opposed to amd 32
quad channel memory apposed to amd's octa channel

currently intel has nothing to go against this. and when intel puts out their 32 core cpu with hex amd still has the bandwidth advantage, and no multi socket tax.

Seriously this thing won't be in stores until Christmas 2017 at this point.

Nevermind, AMD software?

you said ps4 pro and scorpio. but the xbox one and standard ps4 use them too

i'm retarded, i mean hardware

oh yes, i didn't know the normal xbone used amd but i was aware that the normal PS4 used amd as well.

I'm honestly interested in the specs of the 4k xbone.

That's what i read it as anyway.

PS4 pro stuck with gimpy Jaguar CPU cores and only really added 2x fp16 throughput to a doubled GPU.

Scorpio is anyone's guess, but it would stand to reason that it could have launched substantially earlier if it didn't include Zen and/or Vega tech.

Also, Rynen launches at the end of Feb, Vega will be in May. Microsoft can do all the backflips they want at E3, but Vega will have already (but very recently) launched.

amd is consistently beaten out by nvidia on cards with almost 40% less flops, and when they finally get their shit together on drivers, nvidia has cards with equal flops doing 10-20% better.

Its likely vega will beat out a 1080, but nvidia has the wiggle room to shipt prices down drastically, making a titan xp 500-600$ is not unrealistic but amd would have to force them to do it, but a 1080ti would likely satiate people enough they don't have to touch the titans price at all.

no, we all assume that is what will happen, realistically, if nvidia is not apposed, they will release a 1080ti for the 700-900$ price range and leave the 1080 alone. Hell, you want me to be smart about it

the 1070 will be apposed by a smaller vega, the 1080 will be opposed by a bigger vega and potentially a 1080ti will match a big vega, its all on amd to force those prices down, but if I was nvidia, unless vega was more than 10% better at a lower price point than any of my cards, i would not touch prices at all. because why the fuck would i? No one buys amd.

they raised the msrp and made it 200 for the 4gb, you fucking retard
>2 shekels have been deposited into your account, powered by novideo gimpworks

...

this is really assuming nvidia has anywhere to go architecturally, anyone know of any gpu techniques that came out that would increase performance on lesser hardware, this isn't an area I know much about?

its kind of assumed amd will mcm for navi, and nvidia would love to do that for smaller dies too, this is likely the next area that both try to hit.

if amd goes this way and nvidia doesn't, amd is going to win that generation for performance, 2 polaris dies without the crossfire tax would likely cost less than 400$ and would beat out a 1080
3 dies would beat out a titan for 560, and 4 would beat out nvidia's best for under 720$. This is assuming scaling is not an issue, and that there is no driver support needed, the mcm and talking is firmware and nothing can fuck it up through bad code.

But again, no one buys amd, so why would they cater to a high end market? going the mcm route, this allows them to make a great sweet spot gpu, and then scale it outside of the lithography, wasting no money or resources on niche products, it would allow them, once the gpu is done, to have an entire lineup at the same time, hell, you may see quicker gpu iteration, where there would be a 1-4 die range for polaris, then with vega another entire 1-4 die lineup ready to got.

This could potentially kick nvidia straight in the teeth if they are not trying to do this already. but again, no one buys amd for gpus.
cpus, even as shit as they are, people still get them, but give them an amd gpu that is half the price of nvidia and twice the performance, they still fucking buy nvidia.

Nvidia has more recently beaten out AMD cards by choking the geometry and ROP pipelines, but Vega supposedly has almost a 3x clock-for-clock boost in geometry throughput over GCN.

Nvidia literally cut ALU FLOPS from Kepler to Maxwell for comparable models but still did better due to the selective bottlenecking in addition to legitimate advances like their tiling rasterizer.

>it's a year after 1070/1080, and won't even perform as well as a 1080

That doom benchmark puts it between the 1080 and the Titan.

It's why Nvidia is delaying the 1080 Ti (a cut down titan) until Vega is out, to fuck them up the same way they fucked up the Fury X launch with the 980Ti.

>its kind of assumed amd will mcm for navi,

I had kind of thought so as well, but AMD quit tagging "scalability" as a feature on their marketing roadmaps a while ago, so who knows?

I just know that Navi is supposedly a 2019 product now though.

no, that would be retarded to not have 2 die parts at least, 3 is just an assumption they can.

imagine this

"We are amd, and we want back into the server market, here is out only cpu, 32 cores 8 channel memory"

They would be catering to the smallest segment of the market, the people who need 32/64 threads and need high bandwidth memory.

intel has 4 cores at the low end up to 24 cores at the high, amd will hit them along the line as not everyone needs more than 4 channels of memory, or their needs don't have more than 8 cores.

they would not give up a majority of the market just to have one sku above intel, unless this sku is under 2000$ so it fits everyone's need and budget, but that would be retarded to price it there if its better then everything intel has.

Microsoft unlike sony has the money for some major sucking (sonys game division is the only one making money). With Phil Spencer at the helm, they might be doing something bigger than expected since they have been pulling some moves to Merge PC and Xbox. especially since they have been expanding their Play Anywhere program and apparently a Game Specific mode in Windows 10.

AMD fucked itself with Fiji, and the 980 Ti at $650 was a preemptive overreaction by Nvidia, who apparently thought it was going to kick their ass.

Vega seems too similar for my tastes, a big chip with a lot of priority given to HPC customers at the possible expense of the gaming market.

>28th february
>may

Nigga I want to build me a new PC right the fuck now. My i4 4690k is getting old. Fuck waiting another month, I'm going Kaby Lake i7

Where did you find an i4?

It beats it in Doom which uses Vulkan which is an API that already favor d AMD. Kinda how like the fury x beat the 980ti in it, even though the 980ti beat it in 98% of the titles. Pretty sure Doom and Ashes of Singularity are the only 2 games I can think of where the fury x beat the 980ti. Chances are Vega will still be worse than the 1080 and definetly 1080ti, but at a more attractive price point. Forcing nvidia to drop prices.

Shenzheng Chyna probably

My first time I came to Sup Forums I learned that Phenom II x3's could unlock to x4's with a BIOS setting. Then I built my first custom PC with a Phenom II and a Radeon 4850.

Since then it's as if picking any other processor besides current unlocked i5 has been useless.

AMD has practically ZERO market share in the server space. They have nothing to lose and everything to gain. I'm 100% convinced that Zen and Vega were designed first and foremost as server/work station focused parts.
We have info that 8/16 Zen is just over 95W. Intel's 8c is 140w. Are you fucking kidding me? They have an efficient arch with comparable IPC and will most likely a better price/performance/watt. AMD can start and win a price war in the server space. They have nothing to lose and everything to gain. Whether Zen is a PC hit or not won't matter. If AMD even got 10% server market share, that's YUGE money.

I got some stock now but I wish I had more money to buy even more.

Part of the TDP delta is the fact that Broadwell-E(P) has 4 DDR4 controllers and the cores have much higher AVX throughput, but I agree with most of what you're saying.

Its not just their higher throughput potential but the circuitry itself.
If you have a high resolution thermal camera and look at a board from the back side you can see through it, and see the specific structures on die under load. The larger datapaths in the FPU are noticeable. The moment they switch on power consumption increases, even before they start processing instructions. That means that there is static cost in power to have them on. They're normally gated off, and the second they become active power draw per core increases significantly. They large structures, they account for a lot of transistors, and they're a lot of thermal budget.

>buying the literally cucked 290x 4 years after release
LOL

>buying a i5 after 2015
KEKED

good goy

Make AMD Great Again

amd are the opposite of jews you fucking faggot

>being so retarded as to misinterpret the Intel/nVidia shill

good goy is a insult you douche

>reading comprehension
learn it

Yes, I know, I was insulting the shill.
Do you have a bad case of autism?

Im the amd fag with a i5 4690 and a 390x

weird mix but runs pretty good

Well yeah, it should.
It's not a weird mix at all.
Both of those components are great value with excellent performance.

Too late to care. Fucking things should have been out months ago...

The moment of relevance has passed.