[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vr / w / wg] [i / ic] [r9k / s4s / vip / qa] [cm / hm / lgbt / y] [3 / aco / adv / an / asp / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / qst / sci / soc / sp / tg / toy / trv / tv / vp / wsg / wsr / x] [Settings] [Search] [Home]
Board
Settings Home
/g/ - Technology


Thread archived.
You cannot reply anymore.



File: radeon VII.jpg (186 KB, 1920x1080)
186 KB
186 KB JPG
>>
>>69750835
pls delet this kind sri sir pls
>>
>in one game from 2017 that never got popular
>>
Does anyone truely care when playing a game that it's at 75 fps or 98? Just lower the settings man.
>>
the only thing that garbage destroys is your electricity bill
>>
File: r1367464_19263895.jpg (102 KB, 700x394)
102 KB
102 KB JPG
>>
>>69750835
>>69750889
>>69750964
>>69750984
The raw computational horsepower of VEGA 7 is there to compete with the 2080 ti but 2 problems currently exist in vega:

1.) ROPs cannot efficiently handle AA methods like FXAA/MSAA. Going to 8X MSAA essentially cuts performance in half in some cases.

2.) FP32 compute utilization and efficiency was supposed to be high on vega with primitive shaders and full hw async compute. However ever since vega launch devs have been reluctant to add it because game devs must incorporate it into DX12/vulkan APIs and doesn't magically "just werk". This means optimization for thousands of shaders in the game devs jobs and must be meticulously debugged to find performance drops throughout the entire game.

Overall we COULD see vega 7 get pretty close to RTX 2080ti performance but it would rely on primitive shaders being turned on through drivers, game devs implementing them through DX12/vulkan, AND using AA like CMAA/TXAA instead of standard MSAA/FXAA.

For now using vega without AA is as good as it will get. It's also possible to receive more performance boost straight from driver updates but given how AMD managed to get 15-20% from 2017 to 2018 there's not much more juice to squeeze out.
>>
File: 5% slower at 1440p.png (546 KB, 1920x1080)
546 KB
546 KB PNG
>>
File: 0a1.jpg (80 KB, 643x820)
80 KB
80 KB JPG
>>69751081
>>
>>69751121
This graph is not relevant
>>
>>69751162
yeah because it's a generation old card and not the new one it's actually supposed to compete with
>>
File: nividiot kid.jpg (218 KB, 512x800)
218 KB
218 KB JPG
>>69751121
>fortnite -25%
muh gayme
>>
File: amdfag.jpg (2.57 MB, 8624x3896)
2.57 MB
2.57 MB JPG
>>69751186
>>
wow, amazing!
i can't believe these arbitrary measurements, way to go [company]!
it's totally different than [last year's model], that's for sure.
please, take all my money, you deserve it more than me.
i love [company], i just don't understand how anyone could love [other company] more.
i wish i had children to share this with...
>>
>>69750835
>housefire temps
>WHUIRRRRRRR
>destroyed
ok bro
>>
File: 1549831423662.png (494 KB, 1920x1080)
494 KB
494 KB PNG
>>69750835
>>
>>69751183

>>69751311
>>
>>69751121
>>69751162
>>69751183
>>69751186
>>69751225
>>69751272
>>69751285
>>69751311
Yup unfortunately MSAA/FXAA will always weigh VEGA down on top of inefficient use of FP32 compute. The former can easily be fixed by using better AA methods like CMAA or no AA at all but the latter is more trickier to fix without game dev support for additional DX12/vulkan optimizations.

I think someone should make a fair 4K benchmark where AA is disabled on a 2080ti and Vega 7 to better asses what the FP32 compute defficiency in vega is. Also like many anons noted here it doesn't even make sense to use AA on a 4K monitor that surpasses the 150 PPI at 2 feet away limit most people have for discerning individual pixels on a screen.
>>
>>69751353
>please tune the benchmarks in a way that makes me regret my purchase less
>>
>>69751311
Overall worse than 2080, but funny how it beats it on Battlefield V and Shadow of the Tomb Raider, both of the big ticket RTX/DLSS games.
>>
>>69751353
It's amazing how visible jagged edges and shimmers are, even on a 1440p phone with AA off.
>>
>>69751414
DLSS is disable in the bench, Vega could be worst
>>
>>69751186
t. Battlefield V playing AMDrone
>>
>>69751353
>he sits 2 feet or more from his 4k monitor
How are you supposed to see every blemish on your waifu's skin?
>>
>>69751398
I didn't buy a vega 7 and still currently hold a gtx 980 in my rig. I'm just saying that such a benchmark would show overall how bad the FP32 compute inneficiency of vega 7 is if compared to the 2080 ti since BOTH have similar FP32 compute performance.

Going back to polaris for a bit, AMD cutting the ROPs in half from hawaii has made a long lasting damage in performance when AA is on even at 2X MSAA. Vega added the missing rops again but looks like that was not enough.

>>69751438
I'm talking about 4K NOT 1440p. Unless you're referring to 32"+ "monitors" you don't need AA for 4K displays unless you put your face right up against them and burn your retinas.
>>
>>69751150
tl;dr
>vega is borked
>fixing it is hard but doable
>muh DX12/vulkan
>>
>>69751225
>TNshit
lol dumb kid
>>
File: trash.jpg (465 KB, 595x842)
465 KB
465 KB JPG
Why are people obsessed with this abortion of a product?
It's slow, hot, and unavailable.

Go back to the "Wait for Navi" refrain. Something might actually come from that. This is a done deal. You can't buy one and you wouldn't want to. Move on.
>>
POO IN LOO
>>
>>69751611
Because it's an interesting GPU architecture that ALMOST made it but it just didn't have enough steam in the end. Maybe we appreciate things like that because it reminds of us in a way.

A lot of us ALMOST made it as programmers, admins, security analysts, techs, ect but we just ran out of gas or something and we fell through the cracks in the floor and now here we are almost like we're in technological purgatory. But deep inside we know this isn't over until we literally die.

That's how my autistic demotivated ass see's it anyway. Zen was a breath of fresh air when you consider had intel NOT released 5 GHz nuclear reactors they would have bled more than they already did. AMD did good.
>>
>>69751714
>released 5 GHz nuclear reactors
how ironic considering radeon vii sounds like a fucking airport under load
>>
>>69751714
GCN's inability to scale was known from the very beginning and AMD literally said it would take a lot of work to fix it, implying it's not short of overhauling the architecture.
>>
Who cares? AMDRONES and NVIDIOTS are going to be RAPED to DEATH by the summer once Intel release their new, stand alone GPU with 32gb of GDDR... 9 memory and crypto miners are being press ganged into subsidising its r&d costs so it's going to retail at about 3.50.

Screencap this.
>>
>>69751780

So does the 2080/2080Ti (when tensor cores are used) you retard. ~300W of thermal dissipation is not easy to handle especially if you want to do it with relatively little noise output.
>>
>>69750984
>DIRT4
>this image
AMD fanboi here but I kek'd
>>
>>69751353
>CMAA
>better than MSAA
CMAA is post AA, slightly better than FXAA but worse than post only-SMAA. MSAA is spatial and is superior overall to any post method. The only post methods that get near spatial AA have temporal components (SMAA T2x, TAA, TXAA, DLSS) and those often still have a MSAA 2x or 4x component.

How does FXAA weigh VEGA or any modern GPU down. Even the 1.3TFlop Xbone can perform FXAA or MLAA on a 4k image in a few ms. Post AA take less than 1ms on modern GPUs, the FPS of post-AA is negligible. MSAA on the other has a significant effect, though if AMD had anywhere near the compression tech of Nvidia it would significantly outperform the Turing or even Pascal because of the raw memory bandwidth advantage from HBM. AMD is still probably behind Maxwell when it comes to compression, hopefully Navi will improve things.

>4k does not need AA
This is ignorant bullshit, as someone that uses both a 5k Imac and a PC with a 4k 27" monitor I can tell you that AA is still needed. The 5k screen gets close to the point where jaggies are not noticeable most of the time, but they can still be seen in high contrast scenes. Shading artifacts, especially things like HDR highlights are still noticeable. At 5k one can probably get away with a good post or temporal method, at 4k a spatial method is needed in the AA pipeline.
>>
File: 1549733344495.jpg (62 KB, 1280x960)
62 KB
62 KB JPG
Who cares, did they take SR-IOV out or what?
>>
File: Gorrila.jpg (69 KB, 1294x478)
69 KB
69 KB JPG
>>69751150
Nigger retard kill yourself
>>
File: hmmm.png (11 KB, 470x454)
11 KB
11 KB PNG
I am quite perplexed as to why only 1440p results are being posted on /g/ - sure it is logical to also display 4k results given the horsepower of the 2080 and vega 7 which is makes both acceptable cards to use at 4k.
>>
>>69751225
She's got really big boobs for a kid
>>
>>69751081
why would you even use AA
>>
>>69750835
WWWWWWWWWWWWWZZZZZZZZZZZZZZZ
https://www.youtube.com/watch?v=BcL5kbLFBdA&feature=youtu.be&t=46
>>
>>69750889
gcn is always a fucking force to reckon with when compute shaders are in play because they can utilise the sram properly
>>
>>69754919
He's probably one of those retarded faggots that believe goldplating on audio cables makes the sound quality better and swears they can tell the difference.
>>
>>69753838
Yeah where the Fuck are the 4k results.
>>
>>69754919
How the fuck can you not use AA? Even at 1440p you can see plenty of shimmering, and modern shader techniques make shit way worse.
>>
>>69755825
not at native res
>>
>>69751353
>pls make a benchmark that plays to AMD's strengths instead of how you would actually use the game in reality pls sirs do the needful
Why the fuck would anyone do this? People want to know what performance they actually get, they don't give a fuck if it could perhaps maybe perform much better in some stupidly specific setup they will not actually play the games with.
>>
>7nm
>16GB HMB2
>trailing last gen (and current gen) 14nm Nvidia cards
How finished is AMD when Nvidia adopts 7nm?
>>
>>69750835
whats up with all the shit AA slowing AMD cards down man.
>>
>>69751225
Doesn't look like a faggot.
Just a regular american kid
And judging from the amount of fat on him, he's one of the healthier kids.
>>
>>69755905
>request the resolution/settings majority of people who would buy a card that price would actually be playing at
>hurdur that's not the performance you would get from it
Try and convince me that people buying a $700 graphics card don't plan to play on a 4k screen, and then if you cant try and explain to me how hitting pixel density of your eye doesn't nullify the need of AA?
>>
>>69756007
>most people don't use AA at 4k
Keep dreaming AMDrone, maybe you have shit eyes. In any case:
https://youtu.be/n9rhtSwAVfI
Have fun, 4k benchmarks are there, aside from some pretty good 1% framerates, it doesn't impress anyone at that price.
>>
>>69756007
I'd sure hope most people aren't stupid enough to spend $700 on a card to play at 60Hz.
>>
>>69756007
>$700 graphics card don't plan to play on a 4k screen
People who aren't complete cattle normalfags who understand that 1440p@120Hz>>>>4k@60Hz?
>>
File: 1543434568623.jpg (41 KB, 324x419)
41 KB
41 KB JPG
>>69751225
>I have that exact same card
What a chad
>>
>>69755899
Not the guy you were talking to, but you are one blind dude.
>>
Aaaah it's such a flawed card... but I still want one. I don't even play modern games or do anything compute intensive. Am I the AMDrone everyone keeps talking about?
>>
File: 1548833945764.png (77 KB, 750x780)
77 KB
77 KB PNG
>>69751225

and i thought i was fat as a kid ... most probably shit, utter shit diet :/
>>
>>69751225
team red? more like team fat
>>
File: 1547906253954.png (380 KB, 420x420)
380 KB
380 KB PNG
>>69756560
>>
>>69756427
and you're a pretentious faggot
>>
>>69756560
you mean team fed?
>>
>>69754871
This. Just wanted to note, bigger than mine.
>>
Performance of the jet engine housefire is basically irrelevant.
>>
>>69755129

Sounds like any air-cooled ~300W video card (2080/2080Ti qualify in RTX mode).
>>
>>69752795
my accelero extreme (which is probably cheaper to manufacture than that FE vapor chamber abomination with its 9999999 screws) would like to have a word with you
>>
>>69755930
So finished that they're still raking in more money than in the last ~10 years
>>
Serious question. Which reviewers do you trust? Looking and comparing benchmarks from different sites just shows discrepancies where there shouldn't be any.
>>
>>69750835
>vega 64 has better minimums than 2080ti
nvidiots will defend this
>>
>>69757312
Techpowerup is definitely the best.

Modern games, relative performance across various resolutions, and overall overclocking performance. One of their developers made afterburner so they provide the most realistic scenario for overclocking performance.

I also like how they take pictures of the unboxing, various pictures of the card, and disassembing it.
>>
>>69757341
Stutters don't matter
>>
Wait for better drivers and for more btfo.

>muh 20 fps ray tracing experience
>>
>>69750835
Bro nobody plays Dirt, it's literally a benchmark app like 3D Mark
>>
>>69757853
I play Dirt 4

Reminder that buying AMD contributs to pushing tech forward.
>>
>>69750959
Yeah. There is a difference especially if you have a 144hz monitor. Also it just feels nice to have the best GPU out there.
>>
>>69757894
is there a charity organisation for AMD?
I'd like to contribute the margin of one card, but i don't want to waste the world's resources on actually making the card.
Also I can contribute even more if I know I can write part of it off on tax
>>
>>69750889
>got popular

Go back to twitch, zoomer.
>>
>>69751081
>msaa
How many modern games even use it? Most games opt for blurry garbage nowadays.
>>
>>69750835
So, like what is the point of these ultra high end cards when video games are lagging behind by about ten years?
>>
>>69758009
Race to 90+ fps 4K VR. The biggest problem in VR right now is textures are too low res and pop in and out of environments as you move in them so 16GB of 1TB/s vRAM is a good start.
>>
>>69758009
Because some games are unoptimised pieces of shit that barely run at medium settings on a 5 year old high end card.
>>
>>69758081
But doesn't VR need like double the power because 2 eyes=2 diffenrent cameras?

Like, if your game runs at 90 fps, does this framerate transfer equally to VR?
>>
>>69751121
>fortnite
xDDDDDDDDDDDDDDDDDDDD
>>
>>69751081
>Going to 8X MSAA essentially cuts performance in half in some cases.
TAA exists.
>>
>>69755930
So finished they'll post record Q3 and Q4 this year.
>>
>>69758232
Blurry shit
>>
>>69758140

Some Kickstarter VR has 3840x2160 at 120hz

Refresh rate matters more than fps in that regard
>>
>>69750835
Cool. Now do the drivers.
>>
>>69758140
Vertex processing is doubled (or even less with some cleverness) but pixel shading is still just dependent on the total resolution. With foveated rendering it's even less.
>>
>>69758232
>>69758276
The correct choice is FXAA. Best performance to looks ratio.
>>
>>69758449
>even more blurrier shit
I'd rather not have AA at all and performs better
>>
It's a Navi stopgap compute card. Not really a reliable option for gaming (and it doesn't even exist)
>>
>>69758545
It's not even Navi.
>>
>>69758549
I know lol
>>
>"$499 product sold as $699"
- Guru3D.
>>
What is GCN and why is that bad?
>>
>>69758449
FXAA is the antialiasing version of me taking off my prescription glasses for myopia.
>>
>>69758750
GCN is AMD's very fluid graphics ISA that replaced their older VLIW ISA.
It isn't bad. The current implementations in hardware just have a few longstanding limitations that haven't been addressed yet.
>>
>>69759002
>It isn't bad
But it's getting fucking old. Polishing and shrinking the same fucking cores generation after generation is why they never beat anyone.
>>
File deleted.
>>69753472
>in a few ms
>>
>>69759063
Hence why we're still using x86.
And ARM.
>>
>>69759063
>its old
Everything on the market is old, anon. Most arch is 5 years old by the time its put into volume production and reaches shelves.
GCN isn't bad because its old.

Power consumption tends to be high because the structure of 4x16 SIMD lanes are underutilized. They're not gated off, still drawing power, and not used to their full capacity in your typical workload. Nvidia has smaller units with better power gating leading to lower power per op right off the bat.
They need to rework their front end, and improve their total pixel throughput by reworking or dramatically increasing ROPs. All of that can be done in an iterative process. The ISA doesn't limit this in any way.
>>
>>69759117
x86 is an instruction set, same with ARM. Both have cores that have absolutely nothing to do with x86 of old.

>>69759122
>All of that can be done in an iterative process
Then why aren't they? Why are they instead just shrinking their cores, keeping the same amount of ROPs, boosting the clocks here and there, reducing the power consumption.
>>
>>69759133
>x86 is an instruction set, same with ARM
So is GCN.
>Both have cores that have absolutely nothing to do with x86 of old.
And Vega has as much to do with Tahiti as ARMv7 cores with ARMv8 ones.
>>
>>69759150
Enjoy your dieshrink with more HBM cobbled on man. Keep convincing yourself it's not exactly the same shit.
>>
>>69759133
They aren't "just shrinking" anything. There were major changes between the original Southern Islands and Vega.
Vega was supposed to address a front end bottleneck through a driver level feature, but TRG failed to deliver on that under previous leadership.
>>
>>69759133
>Then why aren't they?
bad financies. amd cannot provide budget both cpu and gpu departments. they chose cpu for ryzen and cut the budget for gpu.
with the ryzen money and maybe with sony and microsoft's help, AMD can do something about limitations of gcn with navi. but probably after navi.
>>
>>69759170
3/10 bait, meh.
>>
>>69759185
You don't need R&D money to fix the glaring uArch flaws.
You need them to tape out 3-5 dGPU dies market demands each gen.
>>
>>69759205
>You don't need R&D money to fix the glaring uArch flaws.
yes you do.
>>
>>69759217
You don't.
You need it for each $300 million 7nm tapeout you're going to commit to.
>>
>>69759235
and you think amd has that. ok.
>>
>>69751225
seems like a nice kid. Why the hate?
>>
>>69759251
That they do.
They even has some decent amount of cash right now.
>>
>>69751514
This is a tricky question..
>>
>>69759260
still not enough to iron out the problems.
>>
>>69759334
Money is unrelated to fixing the uArch problems.
It's related to the amount of dies you'll get per lineup and per each launch window.
Navi has 3, split in 2+1.
>>
>>69757933
It doesn't matter with good adaptive sync. My monitor with g sync is silky smooth from 30-120hz for instance.
>>
>>69755129
Why did it not throttle down before reaching 95 C?
>>
>>69756454
No. Just a dude, who likes to collect rare things (5000 in the whole world).
>>
>>69759355
>Money is unrelated to fixing the uArch problems.
but it is. they need to develop a new method to fix the problem and it needs trial and error.
>>
>>69759441
>hey need to develop a new method to fix the problem and it needs trial and error.
Academia already did that for you, just apply the bandaid.
Now taping out 3-4 7nm dGPU dies is yikes.
>>
>>69750964
>Being this poor you can't afford an extra 3 cents per kilowatt.
Jebus Crust you poorfags need jobs.
>>
>>69750835
How about with DLSS

Truth is nvidia invested big on AI generated graphics this generation. It's completely fine if AMD has better throughput. They had the same amount of time and invested solely in gains. AMD can win this battle but Nvidia will win the war.
>>
>>69759584
>How about with shitty upscaling
That's console territory.
>>
>>69751353
>Yup unfortunately MSAA/FXAA will always weigh VEGA down
What are you even talking about? FXAA is basically free on every single graphics card I've used. Its quality is pretty crap too, you either get blur and loss of detail or it doesn't do a very good job against aliasing. MSAA is vastly superior in terms of image quality, but of course much heavier.
>>
>>69756699
It's hard to tell from a video since there's no real way to know exactly how loud you have to have it to match reality, but I'd say that sounds worse than my 2080 Ti with a 380W TDP BIOS when hitting the power limit and an aggressive custom fan curve. Then again, my card actually has a 3-slot cooler so the fans don't have to ramp up to complete batshit levels to keep it from thermal throttling.
>>
Word on the street is VII will now support pro drivers so you can kiss goodbye of any chance of buying one - the compute horsepower VII offers is staggering for the cost.
>>
>>69759455
could you kindly post the paper you spoke of? I'd love to read it.
>>
>>69759900
I'm going to need sources for that Batman
>>
>>69760113

https://www.hardwareluxx.de/index.php/news/software/treiber/48592-radeon-vii-bekommt-radeon-pro-software-unterstuetzung.html
>>
>>69757956
buy AMD stock
>>
>>69760113
Its right on TPU.
Uninstall the Adrenaline consumer drivers and you can install the Pro drivers.
>>
dis score iz gud fer a refer..referre....reefer..stock Vayga 56 alright?
>>
File: vii.jpg (794 KB, 2560x4320)
794 KB
794 KB JPG
>>
>>69760782
Where are the 4k numbers anon?
>>
>>69760816
>My brain hurts
Here. Let me spoonfeed you. The power consumption graph comes from the application named 'Heaven' which is a 3D stress testing tool. Notice how it says 'Heaven' in the third screenshot. Now if you will look further on that image you will see the term '4K'. Look closely now. I know it's difficult using your brain cells.
So what have we learned from this? That one screenshot does not always correlate with another. Dipshit!
>>
>>69760816
Nobody benches GPUs at 4k apparently.
>>
>>69751081
Dude, why would you EVER go through all that effort when Nvidia Gameworks gives you the industry standard.
>>
>>69760990
Not him but
>Nvidia Gimpworks
>>
It looks like Navi could be delayed until Q4 or 2020 in real terms.
>>
>>69760925
>So what have we learned from this?
That power consumption during the synthetic benchmark has no bearing on its in game performance at 4k. If you care about power draw, go buy a 1050ti, not a 700 burger card meant for crushing framerates and crunching numbers.
>>
>>69755129
ive owned an Assus gtx560ti
This is tame in comparison
>>
>>69755930
so finished they supply both consoles (yet again)
>>
>>69750959
>Does anyone truely care when playing a game that it's at 75 fps or 98?
Usually having higher max fps also means higher minimum meaning no discernible skipping or stuttering
>>
>>69753838
>>69755806
4K displays are 60hz max
>>
>>69757312
print publishers: pcghx
else P3Dn
The rest, especially not native Platforms i don't trust at all
>>
>>69761236
Besides that, playing shooters at sub 100 fps feels delayed even with a 60 hz monitor.
>>
>>69761248
No, they aren't.
>>
File: 7649846784678.png (276 KB, 1826x1098)
276 KB
276 KB PNG
So if the R7 gets pro driver support it means I'd be able to use it as an accelerator in Ansys like I'd be able to use a profesional card?
Pic is list of their supported AMD cards, which doesn't include the R7 but besides the possibility of the list being outdated would it be likely to work?
>>
>>69756591
Underrated as fuck
>>
Wait for Navi.
>>
>>69761097
I don't like you.
>>
File: _0.png (187 KB, 919x489)
187 KB
187 KB PNG
>>69760113
gotcha senpai
>>
>>69762424
when they say enterprise virtualization do they by any chance mean multi user GPU support??? If so please feel free to take all my money AMD. This would make Linux finally succeed on my gaymen system. Or do you think having to install the professional driver means, even shittier performance in gaymen?
>>
>>69762530
If it really supports SR-IOV it will be sold out before it hits the shelves and price will skyrocket up to 2500$ mark my words.
>>
>>69761328
>No, they aren't.
Prove me wrong
protip: you can't
>>
>>69761248
did ANUS finally release their newest 4k 144Hz now?
>>
>>69762633
Is the hobbyist workstation market that big? Corporate buyers probably won't opt for something off-label with minimal support.
>>
File: 1549914546946.jpg (130 KB, 858x721)
130 KB
130 KB JPG
>No raytracing
lmao
>>
>>69763235
when will BFV Support DX12 featured gaytracing?
>>
>>69763235
I am yet to own a single game title that supports real time raytracing
>>
>>69763253
>>69763333
Unless you buy a new card every generation you'd be stupid not to get a card that supports raytracing in hardware.
Games will use it more and more as time goes on. In the future it'll be THE method of rendering 3D, period.
>>
>>69750959
>run game in HTC vive
>drops to 80 FPS
>vive force-drops to 45 because you can only have 45fps, 90fps, or a blank screen

it matters
>>
>>69763426
Should have bought a Rift.
>>
>>69763424
except by that time the rtx 4080 will be out and you paid an arm and a leg for a card to beta test the raytracing for a day before you realise it's shit because it gimps your performance.
>>
>>69763424
>In the future it'll be THE method of rendering 3D, period.
Yeah and by that time 2080 is outdated garbage
>>
>>69763477
RTX 2080 Ti can do over 60fps with raytracing enabled at 2560x1440 in BFV.
There were performance increases up to 100%+ between the release implementation of raytracing and the latest update as well, so gimped performance will be solved in software long before the next generation of GPUs.
>>
File: 1495214061240.jpg (19 KB, 400x320)
19 KB
19 KB JPG
>>69758334
>Refresh rate matters more than fps in that regard
wew
>>
File: 5325563f781beff01_w.jpg (44 KB, 600x337)
44 KB
44 KB JPG
>>69763480
>Unless you buy a new card every year
Do you not know how to read? Wouldn't be surprised from /g/ posters.
>>
Let's face it. AMD RTG is at least a year behind the curve against Nvidia. Nvidia is not hampered by consoles so do as they please.
>>
File: henry_cavill_and_plane.jpg (69 KB, 1080x810)
69 KB
69 KB JPG
>>69758334
>Refresh rate matters more than fps in that regard
poor choice of words anon
>>
>>69750835
Oh, look, it wins in a game I've never heard of or seen played outside of this post.

Yep, I'm certainly regretting my buying decision, and the fact that this is the only game out there that the r7 performs better than a 2 year old card in.
>>
Navi has to be GCN because of the nextgen consoles. Sony have said PS5 will backwards compatible with PS4. Once AMD gets past the consolees next year Arcturus may take a different route.
>>
>>69750889
>NOOOOO let's bench GTAV from 2013 again!!!!
>>
>>69751516
wait, AMD cut the numner of ROPS to half? wjy?
>>
>>69751837
i think AMD realized that in 2013-2014, when they were nearly bankrupt
>>
>>69757144
0.1 × 10 is still only 1
>>
File: 1528036283194.png (196 KB, 1024x749)
196 KB
196 KB PNG
>>69763685
>I've never heard of ColinMcRae
>>
>>69759122
wjat is power gating, goy?
>>
>>69759150
what major arch changes was implemented in polaris vis a vis hawaii/ souther islands series
>>
>>69751081
Once again it comes down to drivers, which AMD simply can't make.
>>
>>69763424

That's very unlikely until ray-tracing rendering gets used by gaming consoles (good luck!).
>>
>>69756757

That unit is far more expensive and is massive as fuck. It isn't quiet if you want to kept your high-end GPU from throttling at load either.

The physics are just against you.
>>
>>69756757
Dunno my 2080 ti zotac and it's a cheaper brand usually is around 70c on 1900+ MHz under a hot summer with 60% fans
Level of performance is awesome with 260-270w
>>
>>69763835
You can always make it compatible via driver layer
>>
>>69763835
The point that sony is developing it means it will be a gaymen card.
>>
>>69764208
I don't see why next gen consoles can't use a little bit of ray tracing.
>>
>>69764715
No one wants to pay 1600 dollars for a gaming console
>>
>>69764758
Nibba an RTX 2060 is only 350
>>
RTX is too vramlet to run 3DMark
>>
>>69764907
>only 350
Nigger do you even consoles?
>>
>>69766880
What the fuck did you do to your drivers to make this happen?
>>
File: ffxv4K.jpg (284 KB, 1654x989)
284 KB
284 KB JPG
>>69750835
Radeon VII completely beat the hell out of top of the line NVIDIA MAXWELL!!!!!!
>>
>>69750835
>Oh look, amd drones post another misleading cherrypicked benchmark to sucker some poor souls into buying shit
Not only is it the most favorable game in terms of amd gpus, but they are also using the antialiasing to push the advantage just for experimentation.
So if all you do is play dirt 4, with those specific antialiasing options on then by all means the loud radeon 7 card is for you.
>>
>>69750959
>he doesn’t have 144hz
get out
>>
>>69763835
>Arcturus
this is the name of a chip, not an architecture, confirmed by an AMD employee.
>>
>>69767810
Oh look, that chart where Radeon RX Vega outperforms Radeon(TM) RX Vega, and both are significantly better than NVIDIA Graphics Device.
>>
File: 1485680025167.jpg (171 KB, 960x1280)
171 KB
171 KB JPG
>>69751225
>>
File: hurrdurr1080plol.png (45 KB, 717x909)
45 KB
45 KB PNG
>>69767810
I'm not saying the Radeon VII is even equal to 2080 performance in most games, but this FF benchmark meme needs to die.
>>
>>69764027
cope
>>
>>69761178
They will supply them forever, even when consoles mutate into streaming boxes.
>>
>>69770225
Which won't ever happen until most of the world has near 0 latency and massive infrastructure upgrades.
>>
What is with the AMD shills at the moment? The VII isn't a terrible GPU but it's expensive for what it is and that price is making it compete directly with cards that have an edge in 99% of situations. That's just the fact of the matter.
>>
>>69751081
I thouht the problem was retarded frontend that can't feed all the SPUs, hence why Vega56 performs within 5-10% Vega64 in most cases.
DIRT 4 works around the frontend problem, which is why Radeon performs well in this title and why V64 is actually considerably faster than V56.
>>
>>69770305
They're selling you a badly binned DC product.
>>
>>69768919
I have 144Hz, the higher the framerate the less it matters. There is very little difference above 100-110Hz, especially with G-sync/freesync.
>>
https://www.youtube.com/watch?v=fBX0k7YDQvQ&disable_polymer=1

yet another reviewer that confirms R7 UV is a fucking masterpiece
>>
>>69770305
yes imagine if they sold it at 500 bucks with all of the fp64 power unchecked and the soon to come pro driver support......

surely you must be delusional to think that amd will give something that destroys even the recent tesla cards at fp 64 for something so low
>>
>>69769316
that kid fucks
>>
>>69754871
plastic bottles and s o y products are a helluva drug in combination with no exercise and a vidya addiction
>>
So then, what should I do, currently on a 8GB 1070 I have had for quite a while now. I have have 1,500 I can spend on a GPU. But I'm really iffy about the performance coming from the RTX cards, I'm going to sitll be doing shit at 1080p 144hz. But even the 2080 ti being twice the performance while costing so much is still detering me from buying.

Just wait for Navi?
>>
>>69770669
>Just wait for Navi?
Pretty much, but that one isn't pushing the performance up either.
>>
>>69770360
>I have 144Hz, the higher the framerate the less it matters. There is very little difference above 100-110Hz, especially with G-sync/freesync.
But there are relatively few 75 / 90 / 100 / 120hz monitors, meaning if you want more than 60hz you basically got to go all the way to at least 144hz
>>
>>69770777
At least I will pay less than 300USD for a 1080ti level card.
>>
>>69750835
I really wanna buy a Pooga VII, bros.
Should I do it?
>>
>>69750959
>buy $59+ game, full price.
>not ensuring you can run it at all ultra settings without frame drops below 60fps

It’s almost as if you don’t want to fully utilize what you paid for.
>>
>>69770952
The only games that run at 60fps are PC games.
>>
>>69771023
Even in 2019? How will consoles ever recover?
>>
>>69771051
30fps is is good enough for most games
>>
>>69771058
25 is good enough, sweet spot for the cinematic games on consoles
>>
>>69763189
There's a bunch of 4k 144 Hz monitors.
>>69753838
Can't let the one advantage (VRAM) be shown in any reviews or Nvidia will get mad.
>>
>>69757341
Typical
https://youtu.be/tHxXgOTMVLc?t=428
>>
>>69763426
how the fuck does that not have variable refresh?
>>
Vega is a dumpster fire. it's the GPU equivalent of the Bulldozer. AMD just can't get itself together on both fronts, can they?

t. Ryzen user.
>>
>>69773052
>performance 5% under 2080
>lower power draw with undervolt
>bad dumpster fire
okay.
>>
>>69773089
then why would you need to undervolt it? why don't they set the voltage in the factory?
>>
>>69773127
>why don't they set the voltage in the factory?
AMD set high voltages to sell as many functioning cards as possible
>>
>>69773150
that's the real problem. These Radeon Instinct rebrands are too volatile to failure. As Jim said, these cards aren't really meant for the market segment they're targeting.
>>
>>69773170
>these cards aren't really meant for the market segment they're targeting.
sure right. thats why AMD gives users pro drivers soon to sweeten the deal for plebs who don't care about /g/ayman
>>
File: zoomer2.webm (48 KB, 652x562)
48 KB
48 KB WEBM
>>69750835
BUT NO FORTNITE
>>
File: 2600 a shit.jpg (1.67 MB, 3840x2992)
1.67 MB
1.67 MB JPG
>>69772623
uuu nice, that frametime performance from the 2080 reminds me a lot about my testing with AMD 2600
>>
>>69759543
>3 cents per kilowatt.
fucking pg&e charges 30 cents per kwh
>>
File: Most Curious - Eru.jpg (57 KB, 637x685)
57 KB
57 KB JPG
>>69773206
>thats why AMD gives users pro drivers soon
Explain further.
Not like we're getting full features, but what are the VII owners even getting?
>>
>>69773448
https://www.anandtech.com/show/13947/amd-releases-radeon-pro-software-for-enterprise-19q1
>>
>>69773458
Neat.
I was terribly disappointed they didn't throw some virtualization goodies, but anything from the enterprise thrown down to consumers is a nice change.
>>
>>69759063
nvidia just incrementally updates things too

look at their white papers
>>
>>69763424
It's unusable on all the nvidia cards right now
>>
File: 1537211324605.png (427 KB, 513x497)
427 KB
427 KB PNG
>>69750835
>dirt 4
>>
I think I will buy one.
I'm currently using a GTX1070 and am plagued by the lack of free drivers.
The huge amount of memory in the radeon7 (though unnecessary right now) will keep it relevant in the coming years.
>>
>>69773983
lol at least try to be less obvious
>>
>>69756007
i have 1080Ti and playing on 1080p
>>
>>69773848
This literally looked like something else completely
>>
>>69773994
What do you mean by that? I'm completely honest.
>>
>>69774037
even after reading your post i didn't realise what it actually was until i clicked the image. then again, i'm on day 6 of nofap.
>>
>>69751353
Antialiasing is literally the least important setting in any game, except maybe at 1080p but
>1080p
>>
File: 1524816848915.jpg (153 KB, 1024x1024)
153 KB
153 KB JPG
>>69773458
Looks like Vega 7 and old Vega really are a bargain for hobbyists after all.
We're going to be able to push a meme about Radeon users being serious, productive people while Nvidia ones are childish Overwatch players.
>>
>>69760782
>graph
>uses famerate instead of frametime
absolutely retarded
>>
>want a 2080 ti
>performance gain is okay
>cards are failing left and right
Only reason I'm avoiding it. like god damn wood screws all over again.
>>
the madmen at gamers nexus were able to push a vega 56 to rtx 2070 levels with better frametimes... but using a mod that will probably decrease the lifespan of your card
>>
>>69774464
how many NiggaWatts did it pull for that performance?
>>
>>69774431
only read about FoundersEdition failing, not AiB cards but correct me please
>>
>>69774489
I was going through Amazon reviews and pretty much all of the 2080 ti cards were plagued with 1 star ratings for failures.
>>
>>69774554
copy that
>>
>>69774464
The unlimited power mod was a very interesting piece as it highlights soem of the scaling issues old vega has simply because you can't push the HBM hard enough. Vega 7 shows some cosniderable gains at times beyond mere clockspeed due to the insane bandwidth (buildzoid guesstimates the vega cores stop scaling with memory at anywhere from 700-800gb/s). It would be interesting - when possible to downclock vega 7's core to V56 levels (I know its not an identical core config, but close enough) and then ratchet down the HBM until the scaling matches V56.

Purely academic but still interesting.

>>69774476
All of them.
>>
>>69774476
An additional 200 watts
>>
>>69774464
that power mod is retarded, he only managed to achieve 1710mhz with shitton of watts, while my 56 pulse can get actual 1650mhz @ 1.1mv ~200w
>>
Just ordered a Vega 7 - when it arrives I will have much vram as system ram, a rather odd concept to me.
>>
>>69750959
It does matter when there is fps drop and you go from 70 to 30 fps and start to see lagg but when you drop from 100 to 60 it is still playable.
>>
>>69774820
Where did you order from? Its out of stock everywhere I've looked in the US.
>>
Next Vega refresh when? Figured I'd wait more before upgrading.
>>
>>69774878
I'm in bongland so unless I felt like spending an extra £150 for an MSI sticker (I did not) I pre-ordered the sapphire card as its the cheapest so it will ship when the next batch comes in. Given I've ordered from the biggest retailer for computer parts in the UK I doubt I will have to wait long.

This has alsop pushed me to buy a 4k screen (which cost more than the card) but that wil lturn up in a couple days. I'll get to experience glorious 4k on my 290x. While thats actually useless (i'll go down to 1440p for a lot of games I play) the first game I will play in 4k will be need for speed II using my 290x. A game from 1997 just because I fucking can.
>>
File: 1526935894184.png (37 KB, 586x578)
37 KB
37 KB PNG
>>69751150
this is not twater or plebbit

if you cant deal with posts or sentences that are longer than your iq is "high" you should hang yourself
>>
Howad would everybody be if navi was a 7nm refresh of the current cards?
>>
>>69775428
3,5/4
>>
>>69775428
It can't be that bad considering the gains from Vega 56 to Radeon VII. I'd upgrade my 480 easily.
>>
>>69775428
Polaris hitting best part of 2ghz at current power draw would be highly competitive against the 2070 and below - so i'm not really seeing a downside given the target market for such a card.
>>
>>69750835
Yawn
Wake me up when it can do 4k @ 100fps+.

Till then GPUs can stay with the miners I really dk.
>>
>>69775428
they have been working on navi since before gddr5x was a thing, while focusing on gddr6, it just cant be a simple refresh.
>>69775838
>I'd upgrade my 480 easily.
navi is mid-range though
>>
>>69776218
>navi is mid-range though
Yes, and what is Polaris? Low to mid range.
>>
>>69751121
So on all the games that are actually relevant, you know the ones with millions of active players who spend hundreds of hours a week playing the game, it's worse by a wide margin.
>>
It's going to take them 2+ years to figure out how to reduce transistor size
>largest stake in PC industry
>no release plans yet on the roadmap
>gonna have to get used to hearing crying through all quarters
they're gonna go bankrupt on this one
nvidia is finished, it's over. DONE FOR
>>
>>69769316
>>69751225
Why do you hate children so much? Is it because they're happy?
>>
>>69776263
Relevant is on a per user case basis. Fornite is irrelevant to me but RE2 is not, so the high performance there is a selling point, rather than the defeceit on fornite.
>>
>>69767185
3DMark 99 is probably too old to handle W10 and/or modern graphics cards with 6GB VRAM. Back when 3DMark 99 came out most people wouldn't even have a 6GB HDD.
>>
>>69776280
This post made me realize I am so ronery and sad. I want to be 8 again.
>>
>out of stock
Fuck me
>>
>>69776375
>or use 3DMark 99
I thought it was suggesting go get an old ass version of 3DMark because the current one is too big for your shitty GPU.



Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.