AMD SHITSTOMPS NOVIDEO AFTER SODOMIZING INTEL HOW CAN NVIDITS

File: chrome_2017-06-06_03-00-37.png (774KB, 680x678px) Image search: [Google]

774KB, 680x678px

>>60790709
Oh, the shiposters arrived. Like clockwork.

>>60790717
It's completely true

>>60790729
How so? You don't know anything besides Vega having 8 shader engines.

>>60790750
I know it's 50% faster than a 1080ti

>>60790755
How so? It's no longer geometry bottlenecked, yet it's still *only* 13tflops card.

>>60790770
TFLOPs is irrelevant for gaming.

>>60790709

Might be power related. If anything, this tells us Vega is more power hungry than gp104/102, which quite honestly, I think everyone already expected.

>>60790791
They are, very much so.
Mods please delete this thread, it's pointless, everything meaningful related to Vega was discussed yesterday anyway.

>>60790795
You aren't putting a 300W part in a AIO Mac, so I doubt that.

>>60790795
>uses tiled binning and larger L2
>uses HBM
Honestly are you retarded or what?

>>60790795
We already know that Vega 10 is over 250w. In their marketing slides they showed it as under 300w.
Vega 10 is a competitor to the Titan X Pascal with a 250w tdp.
>>60790802
That depends on the clocks they're using, and how many CU are active.
They can sacrifice some clock speed to increase perf/watt, and Apple obviously won't be using the same binning as your average consumer parts. They'll get cherry picked cream of the crop for voltage per clock.

>>60790824
Crapple uses 56/64 CU Vega10's.

>>60790824
The apple part runs at 150MHz less, I doubt you'd end up with more than 10% performance from such a mild downclock, the previous AIO Mac GPU was a 140W part.

>>60790802

All we know is that the chip inside the iMac is Vega, nothing else. Fwiw it could be an underclocked, undervolted one, which wouldn't be surprising considering the type of chassis it will be into.

>>60790854
We know its flops, 11 tflop fp32, and apparently it's using the cut 56CU version of Vega so any lower clocks won't be even needed, for maximum power efficiency Apple could have gone with the 64CU with lower clocks(shaders are more efficient than clockspeeds), yet they didn't.

>>60790871
No, both 56 and 64 CU versions are available for new throttlemac pro.

>>60790871
>trust me I know better than multi billion company

File: 55772_09_amd-vega-10-details-leaked-dual-gpu-teased.jpg (183KB, 1574x859px) Image search: [Google]

183KB, 1574x859px

>>60790824

>>60790908
Nvidia is fucking finished.

>>60790917
Wew really? Even Evergreen failed to kill nVidia.

>>60790923
It's a joke.
Actually more like a tradition, let me have my fun.

>>60790928
I wonder when will IBM buy NVIDIA.

>>60790908

>225W

After Polaris, I'll believe it when I see it. I remember AMD announcing ludicrous wattage numbers (some of <100w at load even), and look at how quickly it crept up once it was necessary to ramp up the clocks a bit.

>>60790969
Polaris has similar perf/watt as Fiji, you don't find that strange at all?
AT ALL?

>>60790969
Bega has longer bibeline for higher optimal clocks.

>>60790973
What's strange with half-Fiji at 14nm doing the same as Fiji? GCN was intended to clock at ~800-900mhz top.

>>60790973

>Different archs have similar perf/watt

What are you trying to imply?

>>60790989
Neither Fiji nor Polaris are at optimal clocks, 14nm finfets alone bring a good reduction in power.
Polaris is a pipecleaner.

>>60790750
No, we don't know that it's 8 shader engines; that's speculation.

It's my belief that it is 4 and that youtuber misinterpreted that die layout in more ways than one.
I don't know why there is a gap in the CUs, but it's there. It could be because of how the wires run to work at an inside-out fashion for less latency.

>>60791002
That's what I said.
>Polaris is a pipecleaner
And a good one, just look at Zeppelin yields.

>>60791008
No, it literally means it has 8 shader engines. AMD themselves stated their new scheduler could operate with >4 shader engines.

>>60790800
>They are, very much so.
Mods please delete this thread, it's pointless, everything meaningful related to Vega was discussed yesterday anyway.

Mods please b& faggot for being faggot.
But not me because am not faggot like Anon, and op, but fuck you for thinking about b&ing me. Love you.

Pls b& everyone in thread.

>>60791019
They also literally said that Vega was 4 shader engines, fuck face.
If those are 8 shader engines, where are the matching geometry processors and ROP clusters for each, you ignorant twat who can only regurgitate shit you heard from a youtuber and who can't think for himself?

>>60791008
The improved load balancing slide implies 4+ engines, but it could also mean 4 engines but cut in half with each half having its own ROP/geometry units, which is technically similar

>>60791032

This. Shill thread evolving to people actually discussing tech? /g/ has gone mad I say

>>60790802
The mac versions are binned for efficiency. they're also only 11tflops so that means they're probably a couple hundred mhz downclocked for efficiency. don't expect consumer vega to be nearly as efficient.

>>60791055
>150mhz lower
>10% slower
>uses half the power

Yeah, I doubt that.

>>60791038
Where the fuck did they say about 4 shader engines you fuckwit?

>>60791073
Every-fucking-where, you fuckwit.
How about Anandtech? Fuck off.

A shader engine is more than just a group of CUs, you retards.
It's much more likely that they simply split the CUs into two groups per shader engine rather than it being 8 shader engines. From what I see on the die, that's what they've done. You dumb pieces of shits don't know how to interpret a die shot and are just regurgitating shit you don't understand from someone else. Fuck off, faggot.

>>60791095
It's not even a die shot you fucking faggot and
>anandtech
>official AMD source
My fucking god you're so fucking retarded.

>>60791054
>This. Shill thread evolving to people actually discussing tech? /g/ has gone mad I say

Yes. It actually took a turn for the better. Very confusing turn of events actually.

>>60791108
Of course it's not a die shot, it's a RENDER of the die shot and stylized using the actual die shot layout.

>>60791063

Well usual clock/watt graphs aren't linear. There's a point where small clock adjustments cause massive power draw differences. But this tells us either:
>1) Vega is still lower than that point, a clock reduction didn't lower power draw much, and Apple's cooling is a fucking miracle to handle all that inside an AiO
>2) Vega is clocked to the moon, and slightly lower clocks caused a massive drop in voltage, amps, and consequently power draw

Most probable one is maybe 2)

>>60791117
I think
>8 shader engines
is some kind of keyword.

>>60791063
You'd be surprised
Going from 3.4ghz (max all core turbo boost speed) to 2.6ghz (base clock) on my 2012 macbook pro brings the power consumption down from 45 watts to just 15
So 150mhz for half the power sounds about right

>>60791063
Fury x vs fury nano
>160 mhz higher clock
>100 watt higher tdp
You are wrong, friend.
Keep deluding yourself though. It'll make AMDs stock keep climbing and my wallet expanding.

>>60791108
Anandtech reported on what AMD told them at press conferences, you stupid shit.

Those very same new slides from 2 days ago say the geometry rendering performance per shader engine is 419% higher than the Fury X
If it's really 8 shader engines, that means it's 838% higher than Fury X. REALLY fucking unlikely.

I'm not going to argue with some ignorant little faggot who doesn't understand any of this for himself and just repeats what he heard from some nobody youtuber's (misguided) analysis.
If you were trying to argue this in person, I'd be choking you out on my cock instead, you worthless, ignorant little faggot.

>>60791155
Fury X and Fury nano aren't clocked near 1600MHz for that comparison to make sense

>>60791159
Then why the fuck would they split fucking shader engines? It makes even less sense than 8 shader engines you dumb mongoloid.

>>60791163
Well >>60791155 is wrong
It's a 50MHz difference between the R9 Nano and Fury X.

>>60791177
>wanting something explained to someone as ignorant as >>60791177

>>60791177
You're clueless, shut the fuck up retard.

>>60791163
I think the example satisfies the point that small clock changes lead to large tdp changes when you reach the upper clock limit of the chip.

>>60791196
Kys nigger.

>>60790717
If the shadder array is really separated, what's that in the middle?

>>60791229
I believe the middle is the hardware scheduler, which is said to be a significant amount of the die (~10%?)
And I believe part of those in the middle are the geometry processors and ROP clusters (the 4 sets of 2 equal sized rectangles at the four corners of that cluster of various sized rectangles)
So it goes from those hardware schedulers and likely various L1 caches into the 4 geometry processors, into the ROPs, and then outwards into the CUs.

File: 29575201691_1c48dab90b_o1.jpg (4MB, 4458x3778px) Image search: [Google]

4MB, 4458x3778px

Dat Fiji

Just you wait Nvidia. Just you wait.

>>60791291
Same inward to outward arrangement as I described in >>60791277 which strongly implies it's 4 shader engines like the 4 you can see there, that just, for some reason, has a gap between them.

Fuck tech youtubers and fuck the fucks on /g/ who have no understanding but regurgitate what tech youtubers say.

File: ultra kewl family.jpg (100KB, 640x410px) Image search: [Google]

100KB, 640x410px

>>60790709
>Machine-like, battle armor inspired design

>>60790709
more like they can only fit one watercooled vega card in that case

File: 1496830936288.png (459KB, 680x678px) Image search: [Google]

459KB, 680x678px

Should be mostly right.
Tell me if anything's wrong.

File: AMD-Fiji-Die-Shot.jpg (241KB, 718x796px) Image search: [Google]

241KB, 718x796px

>>60791291
where'd you find that die shot?
for ages the best thing floating around was crappy pic related:

>>60791599
https://www.flickr.com/photos/130561288@N04/29575201691/in/album-72157650403404920/

File: AMDGeoFront.jpg (291KB, 1851x497px) Image search: [Google]

291KB, 1851x497px

>>60791597
I think those aren't all ACEs, but is mostly correct.

File: AMDACE.png (281KB, 1583x456px) Image search: [Google]

281KB, 1583x456px

>>60791633

>>60791597
I'd replace the L1 with the display controllers and the IF/GMI one with fixed function hardware like DSPs, VCE and UVD

File: polaris10_die_full37uu8_smaller.jpg (2MB, 1736x1344px) Image search: [Google]

2MB, 1736x1344px

>>60791609
> https://www.flickr.com/photos/130561288@N04/
you're my fucking hero, anon, thanks.

>>60791597
looking at the way polaris is diced up, I'd say your shit is more than slightly fucked up.

> GCP/ACEs/HWSs are all believed to be in the center column, and ACEs aren't nearly that big
> front-end fixed-function units are also still probably in middle, between GCP and SEs
> L1 is spread out in tiny 16kB pieces per CU
> L2 was adjacent to RBEs, but that might change since they're getting coherent L1s now, what you have listed as front-end may be the new L2 location
> the new "HBCC" is probably a huge tag CAM for using part of the HBM as an associative n-way cache, which would probably fit in your red or light green boxes

all caps op should be auto ban

>>60790802
>Expecting good thermal design from the makers of throttlebook featuring 0.8Ghz

File: polaris10_annotated.jpg (2MB, 1736x1344px) Image search: [Google]

2MB, 1736x1344px

>>60791710

>>60791597
>>60791633
Holy fuck everything you listed is wrong.

What you have listed as ROPs is likely things like L1 cache, vector data stores, etc, for each CU. If it was rops, that's 192 of them when it has 62 and they aren't that alrge, you tard.
And the left is likely global data store. If it was a second L2 it'd be the same size as the right one.
What you ahve listed as "front end geometry units" is the memory controllers.
What you have listed as L1... what the fuck you have no idea how L1 works.
What you have listed as IF/GMI are hardware accelerators.
The only thing you MAYBE got right are the ACE.

File: FijiBlockDiagram.png (1MB, 3999x2250px) Image search: [Google]

1MB, 3999x2250px

>>60791710
I was going off this and the other Fiji diagrams.

>>60791789
That is not representative of how the Fiji die is actually laid out, dude.

>>60791789
functional diagram != layout, friendo

>>60791834
>>60791797
Nobody was actually kind enough to disect the Fiji floorpan before though.

>>60791844
anybody who can identify recurrent shapes, count, and read a functional diagram should be able to reach analysis along these lines:
>>60791762
> 16 PCIe PHY lanes, 256b DDR4 PHY lanes, 6 display controllers, ...

we'll see what comes of Vega, but simulated stuff like
>>60791597
is notoriously misleading about relative component sizes and placement.

>>60791844
idk dude it should be fairly obvious for many parts without labels.

>>60791762
"front ends" = geometry processors and ROPs, I think. Same that I think it is there on that Vega die render as I mentioned earlier.
And yeah, HWSs in the middle. Not sure about ACEs.

>>60791919
Shouldn't ACEs naturally be hanging off the scheduler/command processor? I mean their entire damn function is balancing compute and graphics and you don't do that without the scheduler

>>60791919
front-end is geometry/tessellation and rasterizers

R(asterizer)O(ut)P(ut)s = R(asterizer)B(and)E(nd) constituents are back end of the pipe

>>60791919
ROPs are the back end, anon.

Vega 11.. 40-44CU?

>>60792030
Should be close enough to fight off the inevitable 2070(which should be right near a 1080 or slightly better)

>>60791924
Yeah, that's right, actually. Command processor (aka HWs), ACE, geometry processor, and ROPs should all be grouped together, and in that center area both for Fiji there and the Vega render.
But I can't figure out what those skinny rectangles to the tops and bottoms of the CUs are supposed to be, then, on the Vega render.

>>60792030
Almost surely 28-36 CUs for Vega 11. Vega 11 is the RX600 series.

>inb4 some retards asks about
>"but what about 1070 competitor!!!!"
>when the 1070 is shitty price/performance and has shitty use cases.
>1070 is artificial marketing bullshit and almost always a bad buy

>>60792217
ROPs and rasterizers are backend functions, they are in the CUs, ACE/HWS/geounits are not, they're in the frontend.

>>60792230
no they aren't.

>>60792230
see
>>60791633

Clear distinction between CU and CMD processor and what goes in what.

>>60792243
Why would something that needs to wait on the other units and is basically the last part of the pipeline be in the frontend?

>>60792271
They are side by side, not in line.
CUs go back and forth between the processors/schedulers. Thus it's best to place the ROPs by the processors/schedulers.
Sheesh.

>>60790709
Poor Volta.

File: 8890_Slide10.png (4MB, 2628x1470px) Image search: [Google]

4MB, 2628x1470px

>>60792309
I'm talking about the GPU pipeline, in short the ROPs are the last stage of the pipeline, just before the display I/O

>>60791229
Interconnects, L2 cache and random logic, most likely.

>>60790770
>*only* 13tflops
Titan Xp in 11 TFLOPS.

>>60792473
It's a shitload more when it boosts.

>>60792473
Tflops are just clockspeed*ALU count, they take into account basically nothing coming from drivers, APIs, geometry units, scheduling, caches and so on and so fort.

>>60792498
>It's a shitload more when it boosts.
Yeah, like 12 TFLOPS.
>when it boosts

>>60792516
Well aware, I just thought the comparison was stupid since Vega does have more grunt than literally any nVidia GPU in terms of raw FLOPS.

Who here is of the opinion AMD is slightly delaying Vega because they're taking time to do some better voltage binning instead of dumping GPUs in the market with 5-30% different power draw differences at the same performance

>>60791008
>No, we don't know that it's 8 shader engines; that's speculation.
Thank you, I came here to say this.

>>60792536
TitanXP can boost in _muhgayms_ close to 1800MHz at stock.
It's marketed boost clock is far below that.

>>60792544
Power draw is only important when AMD isn't winning, see Intel vs Zen

>>60791452
Is the market for 14-year-olds with heaps of money really that huge?

>>60790709
So basically Vega 10 is faster than dual GTX1080 W E W

>>60792554
>1800MHz
So then it also reaches 13 TFLOPS. Still strange to say that Vega is less.

File: 1496851255264.jpg (302KB, 1012x1010px) Image search: [Google]

302KB, 1012x1010px

Real die shot just hidden from view by photoshop? changing color levels clearly shows stuff underneath

>>60794299
Could be anything under there. It's just a mockup, deal with it.

>>60794317
Then why bother making a clear die floorpan underneath if it's just a simple render? Makes zero sense, also using nonuniform surface mounted cap line, not to mention the interposer isn't uniform either, one side is fatter than the other.
No real marketing mockup would do that. It's fucking pointless and makes your product look WORSE.

I stand by my argument that this is a real dieshot just overlayed with images.

>>60794372
It is a real dieshot, the memory is Samsung one.

>>60790709
Because Vega is a housefire with a AIO

>>60794790
Why the fuck nVidia does not offer watercooled Titan Xp version? Their blower sucks ass.

>>60792554
>for 15 seconds

>>60794806
it got VAPOR chamber

>>60794923
Cool, Sapphire used them back in 2009. NVIDIA blower still sucks ass.

File: Jen-Hsun-Huang-Tesla-Model-X-P90D.jpg (159KB, 1500x750px) Image search: [Google]

159KB, 1500x750px

>>60794806
Because fuck the customer. Design it to be good enough to run benchmarks for the review results and if you expect it not to throttle during normal extended use then you can just suck a small Chinese dick for all they care.

Oh you want 4GB? You fucks won't even notice it being 3.5GB while my bank account will, and who do you think I respect more? Eat it up like our market share you filthy faggots

Hey guys, see that water under the map or the complete details of that road block? No none of you fucks can but we will force it to be rendered anyway because fuck AMD user oh and fuck our own users as well because you still get a performance hit also. Maybe it's not as much as AMD to make it unplayable but its enough to fuck over #60 and #70 Nvidia owners over also, we could let you render it as fast as the #80 users but we want all you fuckers to buy our higher margin products instead, we didn't even give you the choice to reduce the tessellation setting manually like AMD so you are doubly fucked by us hahahaha.

Now buy more of our shit because I need another Tesla you pathetic zombies. P.S. Assholes.

File: 1496157700717.png (806KB, 1038x819px) Image search: [Google]

806KB, 1038x819px

>>60795310
>the way you're meant to be cucked

File: nick-young-confused-face.jpg (30KB, 505x431px) Image search: [Google]

30KB, 505x431px

>>60795310
So you admit that the only logical course is:
>stick to Nvidia
>upgrayd ur gpu every year

poorfags smdh

>>60794806
I hate AIOs

>>60795719
I hate loud as ALL shits blowers more. It's a fucking 250w card with no aftermarket versions.

>>60794902
Enough time to run benchmarks.

>>60795719
but why

t.just upgraded my 1080 ref card blower to an AIO cuz no trip fan upgrades could be found

>>60790709
VEGA 10 >>>>>>>>>>>>>>>>>>>>>>>> DUAL GTX1080

>>60796163
No fucking space, can make more noise
>>60796870
You talk about the TDP?

>>60797100
>You talk about the TDP?
225w is not that high.

>>60797149
>225w
>vega
sauce of this

>>60797218
AMDs own ROCm roadmap.

>>60797236
link? because why the hell they need a aio

>>60797283
Because they don't need an AIO. AIO Vega FE is completely optional. Blower one is the standard version.

>>60797283
>need a aio
It's a luxury option.
You can go aoi because it;s better, you dont really need it.

>>60797301
It's stupid

>>60790709
The reason they only have single AMD Vega 10 there is because it will have near zero availability.

>>60797333
Why? They give you an option for better and quiter cooling ootb.

>>60797348
But Vega is not hot>>60797149

>>60797369
And blowers are loud as shit. As i said, they give you an option of quiter cooling for cards with no aftermarket variants.

>>60797392
I don't think there won't be AIB cards, but obviously AMD wants to directly sell cards like Nvidia Founder Edition is, it cuts into the profit of their partners.

>>60797369
But idiots and PCMR types love AIO water, so it's just filling a niche.
>bbut I hate that niche
So? Why shouldn't AMD make fat stacks because you're an autismo?

>>60797451
There will be AIB cards, just not for Vega FE.

>>60795310
That was a gift from Elon.

>>60797149
>225w

Bullshit

>>60797545
How so?

>>60797558
Be forgot the extra 0

>>60797568
1.7 woodscrews fuck off Huang.

>>60795310
Reminder that there are 5 cars on this picture.

>>60790750
The one game they showed off someone counted the tearing in the stream and determined based on the tearin it was running said game about 20% faster than a 1080ti, and that was with the crossfire penalty of 10-20% peak performance per gpu, I forget the game they used, but it was damn near a best case scenario for amd. That said, it looks like they got something with vega.

>>60790800
amd's cards are typically 30%+ more powerful than nvidia in flops but slower or equal in performance.

flops is a poor way to judge cards outside of peak theoretical performance.

>>60798046
That was before Vega, anon.

>>60798066
oh, I know, however I look at polaris as the pinnacle of amd's performance per flop to nvidia, and treat it as about equal to a 1060 6gb. and that had what a 23~ flop disparity between the two.

If amd gets that disparity down to 15% they match a stock 1080ti, and from there a realistic overclock/aftermarket clock could put them at titan Xp

But until amd either shows us fps comparisons, i'm putting half way between 1080 and 1080ti at stock.

>>60798122
>i'm putting half way between 1080 and 1080ti at stock.
That's simply Fury X at 1600mhz. Vega is not Fiji.

>>60790802
polaris had a 90-160 watt range, we never seen the 90 watt parts as they all went to embeded and pro parts.

if vega has the same range, we will see some 180-200 watt embeded/pro parts and 300~ watt gamer/consumer parts.

>>60790969
consumers never got the high binned dies, we got the bottom barrel shit for a VERY long time.

the xfx gtr is the closest to binned we got, and these cards can pull the power use we seen in the amd demos showing efficiency.

Just because we likely won't see the sub 225 watt vega, doesn't mean it isn't real, its just not our segment.

>>60791002
I don't want to call polaris a pipecleaner, it was just the only part they had that was ready for production at the time.

>>60790906
me knowing more than apple... in almost every regard but design and branding.

>>60792217
>and ROPs should all be grouped together, and in that center area

ROPs/RBEs were formerly adjacent to the memory controllers since their color/Z caches were incoherent given more direct access so they didn't swamp the chip-wide internal crossbar.
They're gonna be coherent in Vega, so who the fuck knows where they'll go next.

>>60792230
>ROPs and rasterizers are backend functions, they are in the CUs
no, the rasterizers are FE and what turn triangles into the tiny square fragments that get fed to the shader array.
The ROPs/RBEs catch the results of shaded fragments and do final memory write aggregation, color blending, MSAA combining, and depth/stencil testing, but they are tied to memory controllers and are explicitly NOT in CUs, even if there is potentially a per-SE limit on RBEs.
> ACE/HWS/geounits are not, they're in the frontend.
correct

There's no fucking way it will be a 225w GPU.
The 480 when overclocked pulls more power than a stock gtx1080 does.

I'll admit power usage is meaningless, if temps are kept at a reasonable level.
But I don't believe 225w for a second

>>60799999
Vega is not old GCN. It's optimal clocks are much higher.

>>60799999
You're basing this off Polaris/Fiji which are technically the same thing just in different sizes when Vega is made to be as different as possible to them.

If AMD wanted a bigger Polaris, they could have released it in 2016.

File: 1480615635847.png (11KB, 689x448px) Image search: [Google]

11KB, 689x448px

>>60799999
>using the worst binned polaris chips (desktop) OVERCLOCKED as a reference for Vega power draw

>>60800223
>implying Vega could ever be as efficient as Pascal given what we know about it's front-end
There is a reason why Maxwell and Pascal are so energy efficient, because they trimmed the energy-hogging processes at the expense of raw compute power. Vega is not taking the same approach, but adding more shit into the front-end.

>>60799999
Wrong.

>>60800471
Maxwell and derivatives are only efficient due to static scheduling, tiling and beefy L2.

>>60800471
You seem relatively clueless.

>>60800471
Vega is going big on graphics-specific work load reduction.

> tiling/binning rasterizers - shaders don't need to calculate pixel values that will be obscured by closer pixels
> coherent RBEs - deferred shaders can pull g-/z-buffer values from cache instead of requiring a flush to external memory and heavyweight shader synchronization points

beyond that:
> moar L2 - saves memory bandwidth
> 2x rate fp16 - low precision shaders don't need to clog up ALUs and registers as much

I'm pretty hype, 1450MHz Fury X is pretty impressive, much less a 1600+ Vega

>>60800674
I thought no one besides miners bought Radeon.

>>60800684
What does that have to do with my post?

>>60800684
got him good
XDDDDDDDDd
buy $NVDA

>>60800684
It's okay to be wrong sometimes.

238% the performance of a single RX 580

You saw it here first

File: CherryPicked.png (1005KB, 1280x721px) Image search: [Google]

1005KB, 1280x721px

>>60800762
Perspective
Keep in mind crossfire scaling is never 100%

File: CherryPicked2.png (1003KB, 1278x720px) Image search: [Google]

1003KB, 1278x720px

>>60800776
Narrow bus on the 580 limits the cards in crossfire in some testing scenarios

>>60799999
Did i just waste my life searching for this

File: CherryPicked3.png (595KB, 1282x716px) Image search: [Google]

595KB, 1282x716px

>>60800803

>>60792448
fuck, you're dumb

>>60800762
It'll likely be in the 200-300% range, depending on game and resolution.

Lilmo is lame

>>60801649
for 400$? no way

>>60802200
The 1700 is like 10% slower than a 6900k at 27% the price.

AMD can indeed surprise on pricing.

>>60790717
http://wccftech.com/amd-radeon-rx-vega-die-shot-fake/

>>60802957
It's stylized, but in no way it's fake, see memory modules markup.

File: ellesmere_annotated.png (4MB, 1264x973px) Image search: [Google]

4MB, 1264x973px

>>60791710
>>60791762
Here's a nonshit dieshot that's annotated

File: fgsfds.jpg (16KB, 239x251px) Image search: [Google]

16KB, 239x251px

>>60790709
Oh look. AMD is supposed to kill all competition and rise to it's former glory.... again.

How long are we going to circle jerk the AMD meme?

>>60802200
Who said $400? Not me. Retard.

>>60803264
uhhhhh that's way too much die area to be ROPs.
I'm fairly certain the ROPs are there with the HWSs, ACEs, etc.
Idk I guess maybe it is, given that it's a 36CU die.

>>60799999
The 1200MHz part is 225W, but the final card will be hitting 300W because they need to push the core up to 1600MHz for it to compete with the 1080 (non-Ti).

File: EfKCIJ3.jpg (1MB, 2048x1614px) Image search: [Google]

1MB, 2048x1614px

Polaris die has that "line" through the CUs.
Looks like some kind of L1 cache that is shared between 2 CUs each.
So yeah. Again, I'm pretty fucking positive it's 4 Shader Engines. The render just has a line through there.

I hate how stupid people are.

>>60803803
No, it's at 1600MHz to compete with the $6000 GP100, not the GTX 1080 non-Ti, you fucking retard.
Seriously kill yourself.

>>60803803
>nvidia shill on full damage control

>>60803803
>denial

>>60803803
lmaoing@you're life

>>60803292
I've been hearing this about AMD for a decade...prior to motherfucking r9 290/390, and killing it in midrange with rx 470/480/570/580.
It is hard to ignore that the need only to step up their game just a little bit in uber high end departement, and low end offerings.
Also, if ryzen is an indication, amd can kill something rather than itself.

>>60805363
Looks 1.7% yields to me.

>>60805061
They need to scam people like Intel.
>LOOK OUR CARDS ARE SO FAST YOU CAN GET THAT TOO $130 FOR 750Ti SAME TECHNOLOGY YOU DUMB KEK

They were selling cards sometimes 3 or 4 generations old under the same generation as their new ones. The 710 isn't even 1/4th as good as the iGPU of the G4560, because it was really a GPU that was like 6 fucking years old when it was new.
AMD has done some rebrands, but nothing like the blatant scamming people like Nvidia did. But that scamming made Nvidia money which they could further invest in marketing to further scam people. Now they're so big they dont' need to scam people as bad. Just planned obsolescence with broken memory controllers and low VRAM.

>>60805370
The farmers won't be happy

File: 54ca7c9ba8b0f_-_hp-firebird-430-0109.jpg (32KB, 430x287px) Image search: [Google]

32KB, 430x287px

Voodoo is still a thing, more to the point HP VOODOO IS STILL A THING?

>amdrones think vega will draw 225w

>>60806773
It's on the roadmaps. And tiled binning and enlarged L2$ does exatly that. It makes GPU more efficient, see Maxwell.

>>60806773
Considering Vega is a line of cards in different configurations, the chance of at least the lowest desktop Vega in the brand being 225W or less is actually very likely.
Vega mobile GPU's will undoubtedly be much less than that.
The next consoles might use the Vega architecture and a very mature process which will see it being very efficient.

>>60806835
Consoles already use some kind of Vega IP, most notably the PS4 Pro

>>60807011
Yes, PS4Pro uses Rapid Packed Math or whatever their 2xFP16 is called.

Where would 2x geometry per clock put Vega 10 in relation to 1080ti?

>>60807058
If it really works - 10% higher.

File: gcn cu.jpg (250KB, 1253x706px) Image search: [Google]

250KB, 1253x706px

>>60803264
those annotations are basically garbage except for the GDDR5 PHYs.

each CU is 4 separate 16x SIMD + register banks, plus some common scheduling, texture AGU/filtering, and shader data store stuff.

if a "CU" doesn't at least have 4 largish similar blocks, you got something wrong.

File: B3Dpoly.png (5KB, 518x269px) Image search: [Google]

5KB, 518x269px

>>60807058
Very close and beating in the upper case, it's not 2x flat, but per clock, and Vega is a good 60-70% clocked higher than Fury.

Polaris itself had a rather notable boost over Vega itself, there's a good reason it can be pretty close in performance with half the shaders and ROPs

>>60807171
>Polaris itself had a rather notable boost over Fury itself,

>>60807171
I find it pretty impressive how the Fury X could actually stand up to a 980ti and a VRAM limitation at stock with that super gimped frontend.
And now that should be equal with 60% at least higher clocks? Things are looking fucking sweet.

>>60792554
Fuck it can, I have it and that piece of shit can't be boosted for shit without breaking.

>>60798046
Because of the draw stream binning rasterizer and many other highly efficiency improving things nobody knows how much performance will actually increase. While GCN and it's 5 iterations were small upgrades here and there, Vega is a basically a whole new GPU. Gaming performance so far is absolutely unknown.

>>60807171
Too bad it won't have 2x geometry output per clock - AMD never delivers on promises like that.

File: Screenshot from 08-06-2017 14:47:17.png (39KB, 752x508px) Image search: [Google]

39KB, 752x508px

>>60807249
Nvidia's turbo is really good, but they should really consider actually advertising real boost speeds in their specs, because the turbo is already fucking huge at stock, people are left wondering why there isn't all that much performance to gain from overclocking, at least compared to Maxwell.

So people end up thinking "wow from 1500MHz to 2000MHz, that's like 30% more performance!"
Which ends up in barely 10% after actually overclocking because Nvidia leads you to believe that the cards runs at like 1600MHz in games.

>>60807286
They overdelivered on Zen, and a lot of transistor budget goes into geometry throughput with Vega, why shouldn't they achieve that?

>>60807286
>Too bad it won't have 2x geometry output per clock - AMD never delivers on promises like that.

they've been pretty firm about 11 tri/clock using "primitive shader" (probably = also using ALUs instead of just fixed function).

that's somewhere in 16.5-17.5 Gtri/s, which is just nuts.
I don't see them going with 8 tri/clock (12-13 Gt/s) fixed function though.

>>60807557
So after this would Nvidia have any hardware advantage anymore? Fury X already has similar ALU throughput to a 1080 and it's clocked 70% lower.. than a 1080

Vega10 will be 4096 and 3584SP
Vega11 <3584?

>>60790709
holy straw grasping batman

just wait for FE before shitposting

>>60807314
they could, or they couldn't. this industry is not easy to understand, let alone predict.

some people think R&D perfectly scales with money invested. they are morons. some people think that because Zen was good, Vega must also necessarily be. they are also morons.

what are the odds of even buying a card when it comes out? won't tell all just get snatched up by miners?

>>60808018
buy it from europe we still have loads of 580/480 here

>>60807945
36-44 CU I think

>>60790709
That's just because the miners are going to grab the whole stockpile again and they can't find enough of them to stick two in a single machine.

>>60800834
HurrDurr RotTR can't into CF.

>>60808165
OEMs don't buy off the shelf hardware, AMD and Intel will always sell a unit to OEMs before selling one to retail in the case of supply issues, ALWAYS.

>>60803292
>it's former glory
>it is former glory

>>60795310
I can believe it's a real hworang.

>>60791291
I built a new system recently with a 500w psu, am I forced to get a 1080 or what?

>>60808621
Vega will have 225-300W models, nothing stopping you from getting a 225W one or even a 1080ti with a slight underclock