NEW Radeon RX Vega 64 3DMark Performance

>>61834267
Are AMD, dare I say it, finished and bankrupt now that Threadripper and Vega are both confirmed flops?

>>61834291
>Threadripper
>flop
Pick one

>>61834267
>1080 performance level with the 1080ti TDP

At least Ryzen is good

>>61834291
Vega is a flop, Threadripper is a success

>>61834267
Holy shit what a disaster. No gamer has any reason to buy anything AMD now.

File: 1491259532131.png (108KB, 684x576px) Image search: [Google]

108KB, 684x576px

>>61834267
More of this please and spread the word.
With more of such publications, maybe they might be msrp at launch

>>61834267
How does that Gaming X compare to a reference 1080?

>>61834267
If Vega is anything like Ryzen it may be much better at work applications.

>>61834330
>With more of such publications, maybe they might be msrp at launch
That's not what's driving launch prices insane. The rumored crypto hashrate is.

File: hestia.png (250KB, 274x403px) Image search: [Google]

250KB, 274x403px

>>61834267
>there are drones who waited over a year for this

>>61834359
lmaoing @ their soul

File: 1494387422977.jpg (181KB, 888x1120px) Image search: [Google]

181KB, 888x1120px

>>61834348
Miners clearly inserting themselves by force into the upcoming timeline. A place will be made for them.

>>61834267
Was waiting to see if I'd build a full AMD PC this time around since I've had all the other combinations already but it's definitely not worth it to replace my 1070 with this

>>61834336
https://www.techpowerup.com/reviews/MSI/GTX_1080_Gaming_X/24.html

Vega 64 in the leak is overclocked

>15% faster in Extreme and ultra than a OCd 1080
>same as a OC 1080 in Performance and Timespy

Hmm, I wonder what this means

File: 8286_01_amd-finally-unveils-radeon-rx-vega-64-56.jpg (154KB, 2225x1066px) Image search: [Google]

154KB, 2225x1066px

>>61834415
>Vega 64 in the leak is overclocked
It looks like the lower clocked one is the aircooled model while the higher clocked one is the watercooled model.

>>61834419
1080 Gaming X =/= 1080 Gaming X PLUS

>>61834323
Even if it was 1080ti + 30% you STILL wouldn't be able to buy it.

File: perfrel_3840_2160.png (43KB, 500x1130px) Image search: [Google]

43KB, 500x1130px

>>61834462
Still looks like there's a difference.

The thing performs silly (in gaums) given the die size and uarch improvements.

File: gs3GTJc.gif (1MB, 320x240px) Image search: [Google]

1MB, 320x240px

>Stock Vega reference blower as fast or faster than factory OC and huge heatsink/fan 1080's
Looks promising.

>>61834419
This is the gayming x plus not the gayming x

http://www.guru3d.com/articles-pages/msi-geforce-gtx-1080-gaming-x-8g-review,1.html

>>61834491
It's 484mm^2 die.
For gaymen it's as shit as it gets.

>>61834491
Gotta "just wait" until September for partner GPUs. Shit sucks, should've been out 2-3 months ago.

>>61834491
>vega
>stock

The clock speed reported by 3DMark is a peak clock, it may or may not mean that the card was overclocked at the time of testing (1247 MHz base/1546 MHz boost vs 1630-1750 MHz boost for this test)

>>61834491
Vega is Oc, read the link

>>61834533
>(1247 MHz base/1546 MHz boost
That's 56.
We're talking 64.

File: 2017-08-10-12-20-44.png (98KB, 957x310px) Image search: [Google]

98KB, 957x310px

>>61834548

>>61834548
wut
>>61834457

>>61834267
All I'm interested in is to see if the Vega56 will perform better than a 1060. If the 64 results are anything to go by it seems likely, and if the price is reasonable where I live, it may end up being a great deal.

>>61834516
>For gaymen it's as shit as it gets.
Like gaymers would be able to buy it.

>>61834663
True that.
If hashrate is as good as "rumored" - the cards would never even leave China.

At least SSG is hilariously good.

Even bigger Polaris would've performed better in gayms.
Why didn't they make a bigger Polaris for consumer market? The die would be smaller and cheaper than Vega10.

>>61834797
>they put a 2TB SSD on the PCB

>>61834291
Vega was trash
Ryzen was a homerun living and surpassing the hype.

>>61835421
And it works!
Wonderfully so. It's so good they basically got Hollywood on lockdown.
Also Vega's HBCC is their sikrit sauce for both Vega and Navi.

>>61834267
it's a mining card

>>61835453
Vega is absolutely incredible for big datasets.
In facts it's the only GPU uarch to handle large datasets.

File: fy dog bongos.gif (416KB, 333x500px) Image search: [Google]

416KB, 333x500px

>>61834359
SO glad I didn't. Bought a 1080Ti a couple months ago and I've never been happier.

>>61834291
>intelshill always try to shoehorn CPUs into a thread about GPUs because they know AMD is kicking Intel's was and all Intel can do is accuse AMD of anti-semitism

>>61835466
compute* :^)

>>61835637
V100 is a better compute card though.

>>61834267
This might be worth it a month from now when partner GPUs are out and Vega 56 is confirmed to be close to Vega 64 speeds while not costing $500.

File: why can't i hold all these memes.png (806KB, 1038x819px) Image search: [Google]

806KB, 1038x819px

>>61835643
It costs $18000-$20000. That's "buy a house" money. Good luck.

>>61834727
someones gotta buy it

>>61835637
>loses to a 1080 in opencl benches
>good

>>61835643
>muh big dies

>>61835465
>Wonderfully so. It's so good they basically got Hollywood on lockdown.
Is amd using a different type of ssd for the ssgs?

>>61835690
It probably uses some enterprise-grade Samsung drives.

>>61835465
>they basically got Hollywood on lockdown
source?

>>61835690
No, it's just physically close and is treated as video memory and doesn't have to go through from storage to system RAM to the gpu and doesn't need the OS to handle all i/o.

>>61835683
>>61835664
Well, it's more powerful than Vega10.
GV102 will also be bigger and more powerful than Vega10.

>>61835702
>It probably uses some enterprise-grade Samsung drives.
That doesn't seem like a love affair that would last too long.

>>61834658
56 beats the 1070 so yes its the only good card of Vega

>>61835741
Bigger and more expensive.

>>61835731
>>61835690
It's literally just an SSD that is closer to the die
still magnitudes slower read and write speed than the actual HBM on the card
16gb+slower 2tb
Literally 3.5+.5 tier marketing, but it's ok when AMD does it

>>61835755
Shame we will never see it because of mining.

>>61835713
They have a publicized interest from RED for the ssgs.

>>61835748
What do you mean?
>>61835765
Anything to maintain mind- and marketshare.

>>61835768
It works like a large (VERY large) L4$ chunk due to Vega's HBCC.

>>61835777
>publicized interest from RED=Hollywood on lockdown
alright buddy

>>61835768
Except they explicitly market it as having 2tb ssd on board rather than saying it's 2tb vram with mystery memory

>>61835773
Just buy 1070.

>>61835796
Doesn't matter what it works like
Still an SSD attached to a graphics card

>>61835816
They're now more expensive because of mining too. At this point it's better to buy 1080 or 1080ti.

>>61835823
Yes, it's just that, but that still makes it fucking amazeballs for doing work with massive datasets directly on the GPU with no overhead.

Stop shilling, stop fanboying, stop sucking JHH's tiny azn penor. It's not even a MUH GAYMAN GPU you're trying to shit on. You're just pathetic at this point.

>>61835823
It's low-latency storage attached directly to GPU that's also paged by Vega's HBCC.
Please stop damage controlling, nvidia will have something similar in ~two years.

>>61835780
Those ssgs will be constantly stressed, and will need replacement more often than usual but I'm pretty sure RED can afford the ssgs since they already go through a lot. I'm only specualting, so take it what I say with a grain of salt, but I think these ssgs might burn through more ssds than usual. Again just specualtion, and not an attack on amd. I don't have my hands on this card, so I'm just making an assumption right now (also I'm pretty tired).

File: dN1gYXAO3eaa4zLs.jpg (354KB, 3049x1534px) Image search: [Google]

354KB, 3049x1534px

>>61835807
Are they now?
They are trying to call it "2tb of SSG memory" instead of "We put an SSD on our graphics card"

>>61835851
>>61835852
>Stop shilling, stop fanboying, stop sucking JHH's tiny azn penor. It's not even a MUH GAYMAN GPU you're trying to shit on. You're just pathetic at this point.
>Please stop damage controlling, nvidia will have something similar in ~two years.
You guys gotta stop projecting and drop the nvidia boogerman
It's an SSD attached to a GPU, nothing will ever change that

>>61835805
>alright buddy
I didn't claim "hollywood lockdown" that was another anon. I just gave you a possible reason as to why that anon made that claim.

>>61835575
now you can play games while your mom will make a dinner on your GPU

>>61835768
>>61835823
>trying to compare GTX 973.5 to Radeon Pro SSG

This is why nobody takes Shillvidya seriously.

>>61835889
Stop damage controlling for nvidia, anon.
They already did that for you (by raiding 3 NVMe drives).

>>61835864
Meant ssd's

>>61835864
Enterprise-grade MLC drives have great endurance, anon.
And you can RMA it or replace drives on your own (probably).

>>61835889
it's a really fast ssd, at least that is good

File: three point five.gif (1MB, 300x169px) Image search: [Google]

1MB, 300x169px

>>61835867
Call me when AMD loses in court over it. https://www.gtx970settlement.com/

>>61835904
You're projecting a lot man, not everything is a big shill conspiracy

>>61835931
What's important is latency, not speed.
This SSD is attached directly to the GPU.

>>61835931
It's just an NVME SSD, it's really not any faster than consumer m.2s

>>61835935
As I said, stop damage controlling for JHH, he already did that for (You).
>>61835953
It's much lower latency.

>>61835927
I don't think ssds would last that long under a constant state of read/writes like volatile memory, but I could be wrong since I have no clue as to how they work. I'll just have to wait, and see.

>>61835958
>it's much lower latency
ok cool, still an SSD with far slower read write speeds than the memory on the card. The latency won't cover that up

>>61835985
>The latency won't cover that up
Oh, it does, considering it does live 8k playback.

File: Tin.jpg (56KB, 320x240px) Image search: [Google]

56KB, 320x240px

>>61835958
>As I said, stop damage controlling for JHH, he already did that for (You).
Do you have some medicine you need to take? What the fuck is JHH?

>>61836013
>Oh it does
No it doesn't, live 8k playback doesn't mean anything. My computer can playback 8k video from my camera, what's the point?

>>61836038
But can it do raw 8k playback?

>>61836043
Yes, at least from my friends rented RED camera it can. But it's the lower 15k model but I doubt the higher ones change much
32gb DDR4
1080ti
R7 1700x

>>61836077
https://www facebook com/jarred.land/videos/10154878406240415/

>>61836104
>https://www facebook com/jarred.land/videos/10154878406240415/
They are using a 4 core CPU

>>61834291
i will so fucking laugh if the new bios is as good as they say

cant wait to see the threads here

>>61834491
~1900MHz is basically stock for Pascal, good custom cards when OC'd end up >2100MHz, so this is actually looking really bad. I bet those 1750MHz results are from the liquid cooled ~350W card too.

>>61835484
Vega is not efficient like Volta and Pascal

>>61836189
It's as efficient as anything ~"14nm" FinFET gets.

>>61835867
Yes they are

https://pro.radeon.com/en-us/product/pro-series/radeon-pro-ssg/
>16GB HBM2 + 2TB SSG
Unlike Nvidia AMD tried to be transparent

>>61836336
For the same jobs, Pascal/Volta draw less power. If you buy like 400 GPU, this is a big deal.

>>61836404
Compute-wise they both draw the same amount of power.

>>61836189
>>61836336
If it could feed everything it would be efficient. As it is it's just wasting power on shit that doesn't do any work but isn't gated completely

>>61835867
so? they patented it it was their idea they did it
regardless of how you spin it its still better

>>61834267
I TOLD YOU SINCE FEBRUARY. VEGA IS DOA

>>61836679
but amdfags told me it would trash 1080ti and take blows with volta

>>61836693
the volta on the market isnt the volta nvidia actually talked about....

its just maxwell 4.0

>>61836717
Maybe nVidia's next uarch will be different...

>>61836731
they gonna have to since vulkan dictates that it needs a hardware sc...
and it will be hilarious when their power draw skyrockets because of it

File: wheresmymic.gif (2MB, 320x240px) Image search: [Google]

2MB, 320x240px

>>61836752
DX11 forever! DX12 bans multithreading! Vulkan API is shit! Yay DX11! I love DX11, don't you?

>>61836752
Hardware SC barely affects power draw.
What matters is ALU count. Moar ALUs= moar powah. And we're almost out of node shrinks.

>>61836768
nivida card handle vulkan pretty well though unlike dx12

>>61836777
a true hardware sc matters A LOT we saw a pretty nice drop on amd when they change the ACE to hws with the new gating...
add to that the FP perfomance and you have a recipe for laughs in nvidia there is a reason as to why no one has done a power draw test after the "magica" driver nvidia released for titan

>>61834323
are you delusional? i'm more worried about not being to find an aftermarket cooler vega 64 for a reasonable price once they're out in september. i hope to god more stores adopt microcenters 2 gpus per household rule.
a gtx 1080 equivalent for AMD with freesync is going to be absolutely phenomenal. imagine a gtx 1080+xb271hu or whatever overpriced gsync monitor you can think of but actually affordable. the only people i can think of that will complain about vega gpus are people who already over paid for pascal gpus.

>>61836829
I am complaining about Vega because the die is too big given the gaymen perf.
It just doesn't make ANY sense.

>>61836768
nvidia trashes amd even in vulkan, retarded amdrone

https://archive.rebeccablacktech.com/g/search/filename/wheresmymic/

File: 580 1060 game breakdown.png (505KB, 1920x1080px) Image search: [Google]

505KB, 1920x1080px

>>61836877
Oh no equivalent AMD GPU is beating equivalent NVidia GPU in some games, better go nuclear rage mode.

>>61836877
>nvidia trashes amd even in vulkan
HAHAHAHAHAHAHAHAHA
HA
Good joke, lad.
Nvidia is still trying to push VkHLF?

>>61836877
lol no

>>61836867
It actually does. Shit memory, shit memory controller, and shit scheduler. AMD's fucking retarded. And then they proceed to stuff it with an ALU block that they expanded for higher IPC and higher frequency. Even polaris is bottlenecked just like Fury, just like the 290s. They haven't fixed a fucking thing. The 13TF didn't magically disappear, if they feed it that's exactly what they'll get and in ALU stress tests the performance begins to show but not in games and not in actual compute with varied memory ops

>>61836899
im going to laugh so hard at this especially with the new gpu profiler amd created

>>61836768
I sure love artifacts and stutter in vulkan while running rpcs3 on RX poojet gpu. Just kill yourself.

>>61836829
>the only people i can think of that will complain about vega gpus are people who already over paid for pascal gpus

this level of delusion

>>61836948
>alpha quality emulators have bugs
Wow!
>>61836926
>Shit memory, shit memory controller, and shit scheduler. AMD's fucking retarded. And then they proceed to stuff it with an ALU block that they expanded for higher IPC and higher frequency. Even polaris is bottlenecked just like Fury, just like the 290s. They haven't fixed a fucking thing. The 13TF didn't magically disappear, if they feed it that's exactly what they'll get and in ALU stress tests the performance begins to show but not in games and not in actual compute with varied memory ops
Now that's what i call shitposting.

>>61836926
yeah its so shit that the compute benches shows a total destruction of the p6000 and on games pretty much between 1070 and 1080
its a dead giveaway that amd created a bios for the card to be compute specific
and is also a dead give away that they are going to release a new bios as we all know

>>61836948
I sure love comparing pre-alpha emulators on a console that used an NVidia GPU in the first place.

Kill yourself.

File: 19702370_1043809299083752_7431183469714441190_n.jpg (20KB, 450x374px) Image search: [Google]

20KB, 450x374px

>>61836948
you cant be that retarded

>>61836967
>>61836971
>i-it's alpha software
nvidia doesn't have this problem ;^)

>>61836997
(You)

>>61837007
you sure showed me

>>61836990
He's a novidya shill, it's pretty much expected.

>>61834291
this, so much this

i am what you call a amd fanboy, and i sold my shares this week

amd is done

>>61836970
Show me the compute benches

>>61837048
Look, boys and girls.
Concern shilling!

>>61836997
tegra is calling

>>61837007
https://www.youtube.com/watch?v=YrIZHxrczAY
https://www.youtube.com/watch?v=HwZTyhmFgJo

;^)

>>61834267
Repeat after me:
1080 performance levels

>>61837055
http://www.tomshardware.com/reviews/amd-radeon-vega-frontier-edition-16gb,5128.html

have fun

>>61837074
That's what everyone expected anyway. Well, everyone except fanboys and shills.

>>61836997
>using the wink smiley with the carat nose.

File: Vega_FE_Memory_Bandwidth.png (129KB, 800x574px) Image search: [Google]

129KB, 800x574px

>>61834267
Vega 64 is beating a factory oc'd AIB 1080 at least. Uptick vs. Vega FE still isn't enough to suggest they've fixed the memory bandwidth regression vs Fiji yet, though.

>>61837125
no one is going to bench it against a stock 1080
literally no one

File: 1490864481682.jpg (853KB, 1920x1280px) Image search: [Google]

853KB, 1920x1280px

>>61837048
UMA

>>61837099
>literally no compute tests
>just viewport rendering in 2d and 3d design productivity apps which is literally the graphics pipeline of a game but with a fraction of the i/o but a whole lot of ALU work, the ideal scenario

>>61837048
Shit that didn't happen: The Epyc motion picture
Starring: Retarded no-caps typing shill

File: 9y5R5Pc.png (124KB, 900x775px) Image search: [Google]

124KB, 900x775px

>>61837154
If RTG can fix the memory bandwidth issues it won't matter which version of the 1080 they compare the blower Vega 64 to.

>>61837297
i think the reason we heard about the "new bios" after the benchmarks is that..
unless the memory is suppose to work like this when hbcc is disabled

>>61837297
>has the memory bandwith of 1080
>performs exactly like 1080
Rly makes me think.

File: 7BSGeR9.png (62KB, 1016x1203px) Image search: [Google]

62KB, 1016x1203px

>>61837321
Got a link for the new BIOS stuff? I've been scratching my head about this since Vega FE came out.

Wait someone did B3D suite testing?

>>61837367
just that a new bios will come with the rx no specifics on it

File: AMD-Radeon-RX-Vega-64-With-NVIDIAs-TITAN-Xp.jpg (146KB, 1600x1032px) Image search: [Google]

146KB, 1600x1032px

Garbage, fuck all the hype, vega is garbage

>>61837673
Calm down jacketman.

File: 1451171492445.jpg (9KB, 262x263px) Image search: [Google]

9KB, 262x263px

WHERE IS 56

>>61837900
Early September I guess.
And no, you'll never get one.

AMD GPUs have become like luxury limited edition goods, lmao

I wonder if they can somehow fix the effective bandwith issue for RX launch.

>>61837938
They don't have to, they're gonna be sold out anyway.

>>61837958
Why?

>>61837970
cryptokikes

>>61837970
good or not on games miners will buy them out
but if they are good on games well amd will get fucked harder

>>61837981
Cryptochinks need bandwith issue sorted out for dem sweet sweet megahashes.

>>61838003
we dont know how hbcc affects mining... if the number of 70 hases is even remotely true amd is going to have to stock millions of gpu's

>>61838032
How fast is anything GP102 on mining?
1080ti, titans, anything?
Also HBCC is for large datasets. ETH is memory bandwith & latency sensitive.

>>61837048
stop it brian, you're not paid to shitpost on 4chan

>>61838057
He's not competent enough to do anything else but shitpost on 4chan.

>>61834291
threadripper isnt a flop in fact why do you think they are releasing it next to vega? to overshadow the piece of shit those gpus are going to be in cost/perf compared to nvidia

>>61837379
For Vega FE, yes: http://www.pcgameshardware.de/Vega-Codename-265481/Tests/Benchmark-Preis-Release-AMD-Radeon-Frontier-Edition-1232684/3/

It shows that Vega FE has ~20-30% regressions vs. Fiji in raw memory bandwidth, effective texture bandwidth and texel fill rate. The real question is whether RTG can fix these regressions via BIOS updates, because Vega is currently heavily memory bandwidth bottlenecked in gaming (and ETH hashrate).

If it turns out that RTG is able to fix it and you see a significant performance boost from it, you can then go on to deduce yourself what Vega20 with double the memory bandwidth might be capable of in memory bound workloads.

>>61838079
They are releasing Threadripper in August to shit on Intel's table, since HCC Sklylakes-eggs launch in September.

>>61838053
nvidia is generally shit on mining
and no we know that hbcc affects EVERYTHING the question is how much

>>61838092
yes but they also timed it with vega

>>61838104
HBCC is basically CPU-like memory management in GPUs. ETH memory footprint is too small for HBCC to actually affect it.
>>61838084
>Vega20 with 1/2 FP64 and HBCC
Now that's what I'd call a proper HPC monster.
Anyway why there's only ONE reviewer that benched FE with B3D suite?
What the fuck happened to tech press?

https://us.msi.com/Graphics-card/Radeon-RX-Vega-64-8G.html#hero-specification
>Minimum Recommended PSU
>750 W

https://us.msi.com/Graphics-card/Radeon-RX-Vega-64-WAVE-8G.html#hero-specification
>Minimum Recommended PSU
>1000 W

>>61838173
Thank you Rajesh!
Yours, leather jacket man.

>>61838173
>I have a 550w fanless PSU
Bring it on.

>>61838209
Don't worry, AMD always rates their cards for worst case scenario (i.e. diablotek shit).

>>61838173
>card uses 50W more
>bump up requirements by 250W

Based AMD

>>61834323
Vega is for professionals, not gamers.

>>61838142
How about the fact that virtually no one was able to deduce the right conclusion from looking at the B3D suite numbers, then looking at Vega FE's performance in geometry bound workloads ( http://www.tomshardware.com/reviews/amd-radeon-vega-frontier-edition-16gb,5128-6.html ) and Vega FE's clock-for-clock performance vs Fiji in games ( http://www.gamersnexus.net/guides/2977-vega-fe-vs-fury-x-at-same-clocks-ipc ).

You tell me what other conclusion is possible than a massive memory bandwidth bottleneck? And yet barely anyone was talking about the bandwidth issues until I started raising a stink about it on various forums recently.

>>61838286
It is, it's intended to work with large datasets first and foremost, BUT it's gaymen performance is totally lackluster given uarch improvements.

>>61838294
I want my 2003 back when everyone tested everything in GPUs, even fucking aniso quality.
Modern tech "press" is so fucking shit.

>>61838228
I know, all PSU calculations are bullshit. (People recommending 750w+ for anything but SLI/Crossfire)

>>61835889
>It's an SSD attached to a GPU, nothing will ever change that

Yeah, and do notice that it is this that makes it fucking good.

At least digi is surprisingly happy.

I wonder if AMD will use their fat slab of interposer- and chiplet-related patents in Navi.

>>61838142
>What the fuck happened to tech press?

They've been bought by Intel and Nvidia ages ago. Fuck, I remember when Assassins Creed added DX10.1 support, giving it a huge perf boost on AMD cards, and then NVidia paid them to undo that patch (the GF200 cards at the time only supported DX10, not 10.1).

>>61838424
No, I mean even the ones that are on AMD's train. No one but germans benched FE in ANYTHING relevant despite the card being out for over a fucking month.
This is getting silly.

#MakeSomeNoise
#BetterRed
#PoorVolta

>>61838863
Hi, dear shitposter.
Do you think making 815mm^2 dies is okay?

>>61838353
Vega draw more power than Fiji

>>61838992
While clocking much higher.
But higher clocks do pretty much nothing for Vega since its bottleneck is in memory bandwith.

File: 1500839938783.gif (505KB, 1080x1080px) Image search: [Google]

505KB, 1080x1080px

>>61838863
>#PoorVolta
Gee, I wonder what would happen if RTG made a four HBM2 stack version of RX Vega???

>>61838931
wew lad

>>61838931
AhAHaahahHAHA nvidia idiots basiacally sayd that threadripper is better than intel

how cock can you be

expected from the makers of rypoo

>>61839399
SOPA DE MACACO

File: 1502158411586.jpg (38KB, 630x472px) Image search: [Google]

38KB, 630x472px

meanwhile in India animals eat their stillbirth ass babies

https://www.liveleak.com/view?i=855_1425380286

>>61839463
There's 1.3b of them.

>>61839039
4D CHESS

>>61839039
That's (probably) Vega20 and it's FP64 card you'll never see.

>>61840121
>4D CHESS
No, that would be RTG *magically unfucking* Vega's bandwith issues. Not gonna to happen though.

>>61840121
Stupid as this is going to sound, I actually think it is. I think the capability to introduce four stack RX Vega as a refresh is staying in the back pocket in case they can't finish Navi until 2019. In theory, Vega would gain enough from doubling its memory bandwidth to buy them more time to work on Navi if (when) it becomes necessary.

Now if they could manage to fix the bandwidth issues with 2-stack Vega in the meantime, this would be a whole lot easier to live with.

Nvidia is the absolute superior company compared to AMD and Intel.
The market experts think likewise.

AMD will never catch up to Nvidia again unless the latter makes a huge blunder.

>>61840162
>make shit card only worth for mining
>it's sold everywhere
>nvidia thinks vega is shit
>trying to milk consumers like intel
>vega unleashes
>nvidia BTFO

>>61840180
4 stack Vega refresh on 14+ node?
Sounds reasonable.
>>61840181
AMD already patended everything that will matter in the near future, including chiplets and "one-fits-all" interposer.
Only Intel has EMIB to compete.
>>61840199
No, not exactly.
We don't know what's causing said bandwith issues.
And everyone who knows (or can guess) is probably under the NDA.
Sad things.

>>61834359
yeah, looks like my r9 390 will survive for another year

>>61840311
you and me both

all aboard the navi hype train

Think volta will be another big leap, or was the 1000 series a fairly unique jump in performance?

>>61840416
Pascal was a fucking double node jump.
Volta is just biggur dies with half-node jump to prevent housefires.

>>61840311
my 390 died 4months ago. got the full money back and i'm waiting for vega :(

>>61840512
Just get whatever nvidia card you want.
You'll never get Vega if it's mining perf is any good.

>>61840512
>>61840537
preorder :^)

>>61840558
No preorders (according to Gibbo that is).

>>61840537
no i fucked up. got the 32" Omen monitor. it has 75hz and freesync, but on nvidia cards it only works on 60hz. the 15hz loss would trigger my autism everytime i'm infront of this monitor

>>61840575
Oh, that's bad.

How does one fuck up the gaymen perf so much?

>>61838931
>threadripper
>gaymen

>>61840575
vega 56 might be good, just wait™ for benchmarks and buy the saphire one when it comes out

>>61834359
i want to manhandle that fun sized goddess