Why is AMD so incompetent?
>>55415822
3 . 5 G I G S
.
5
G
I
G
S
>>55415835
Funny how, somehow, 3.gigs from nvidia is beating 8gigs from AMD
>>55415858
only with an AMD cpu
>>55415822
As much as I want to hate AMD for being incompetent fucks for their drivers and cucking my hd7950/i5 750, I just want to say at least they haven't killed any GPUs.
I honestly can't tell who's the worse choice anymore...
>>55416278
And an Intel i5
>>55416279
There are no good choices. You either get gouged and fucked by the Nvidia kikes or put up with AMD shooting themselves and then you in the foot every five minutes.
SAVE US BASED INTEL
>>55415822
Don't worry amdrones will tell you that dx11 driver overhead and driver multi-thread is a meme.
I must say that I'm an amd user but i can't fucking defend things like this.
Just look at this pic 16.6.2 (latest driver) negative scaling using multi-thread on dx11.
>>55416330
I5 2500+HD7970 here, I'm actually getting slightly worse performance than my friend with an i7 860 and gtx670.
I'll probably upgrade to zen + gtx1070 if prices are alright, I want off this radeon meme
>>55416357
Absolute bullshit. Even with the dank driver overhead maymay, there's no scenario where a 670 even comes close to a 7970. Either you're a lying shill, your card is faulty or your system is fucked.
>>55416288
ah, didnt notice that
>>55416357
Amd is fine but you need as much single thread performance as you can get.
Then it'll perform great...
i'll like review sites to use more than just the highest end cpus to test the cards so amd will be forced to fix the driver.
>>55416373
Gta v, he gets almost 5 more fps than me on same settings. I get average low 50s, with some dips into low 40s, he gets mid to high 50s, with some dips into high 50s.
Battlefield 4 is roughly the same if I use mantle, or 3-4fps slower if I use dx11.
I'm pretty disappointed, but I try to enjoy the performance I'm getting because honestly, it's enough.
But like I said, I'd like to try nvidia next time because his 670 was quite a bit cheaper.
>>55416405
Also this. I've seen reviews where they tested shit like r7 260x or 270x with an i7 4770k and shit.
Wtf.
>>55415822
DELETE
>>55416373
Driver overhead + cpu bound game and you'll get that problem.
I've more fps on arma 3 with a 7950 + intel (oc) than a friend with a 290x + 8350 (stock).
I love AMD. I bought 5000 shares last year at $1.80/shr and sold recently at $4.75/shr. AMD is brilliant.
>>55416432
Some dips into high 40s*
It makes me sad though, knowing that if we both had an i7 4770k or 6700k that my card would destroy his.
>mfw i3 3220+r7 260x
Did I fuck up?
>>55416657
Kek
>>55416455
If they were so brilliant you wouldn't have sold your shares
>>55416657
Afaik only 970 has that memory problem, but i'm starting to think that some cheap vram for the window manager could be useful on certain configurations/circumstances.
>>55416776
Yeah like r7 240 4gb ddr3 from sapphire is amazing value for gaming pc, it's 4gb!
>>55416887
Well pinch my nipples they actually make this shit?
Why?
>>55415822
>calling amd shit when Nvidia tries to kill your car at least twice per year
NVIDIA CARD DAMAGING DRIVERS OVER LAST 5 YEARS:
364.72 WHQL
364.67 BETA
364.51 WHQL
364.47 WHQL
320.18 WHQL
267.52 WHQL
196.75 WHQL
AMD CARD DAMAGING DRIVERS OVER LAST 5 YEARS:
15.11 WHQL
>>55417452
>all those 364
jesus fuck why didnt they just scrap the drivers and build it from the ground up again
>>55417452
So? not op and I'm an amd user still i want this fixed.
Posting that is like posting 3.5 or amd pcie specs
>>55417525
I would say issues on nvidia drivers are not an issue for me, amd driver overhead is (And fortunately 15.11 doesn't caused any issues for me)
Well as i said we're not discussing 3.5 fiasco, faulty killing drivers, pci-e out of spec problems, 2d acceleration bugs, Opencl compiler bugs, gameworks, vulkan, dx12...
we're discussing driver overhead >>55415822 and related >>55416330 negative mt scaling under dx11.
3DMark doesn't seem to have a way to search for API overheard results to compare against that I can find. However I did use Google to find some stuff hidden away on the site. The image to the left is my overclocked R9 290. I found a result for a 480 on a slightly lower spec CPU and I also found one for a 970.
480
http://www.3dmark.com/3dm/12838370
970
http://www.3dmark.com/aot/117403
Notice the jump in DX11 performance for the 970.
The fact my overclocked R9 290 beats the 480 out in DX12 is also interesting. But it is a lower spec CPU so that obviously has an impact.
>>55417768
Oh and here is one for a 980 that beats my score but only by a small margin in DX12. I would need to see more 480 API overhead scores to make a fair judgement however.
>>55417796
Shit I forgot the link
http://www.3dmark.com/3dm/8225435?
>>55417452
>364.72 WHQL
>364.67 BETA
>364.51 WHQL
>364.47 WHQL
I will never understand how someone who browses /g/ can be this stupid.
Only 364.47 was busted. Windows overwrote 2 of the inf files with a default VGA driver after the nvidia driver fucked up. The other 3 worked fine if you ran DDU to remove windows VGA drivers as well as the nvidia ones.
Welcome to 3 years ago op? We already all knew fx series is garbage.
>>55417796
Is the same shit amd caps at ~1 million draw calls while nvidia does over 2m.
look a 770 http://www.3dmark.com/aot/38872
tip you can use inurl:3dmark.com/aot/ plus the graphic card you want to search for on google ie: inurl:3dmark.com/aot/ 770.
>>55417971
http://www.3dmark.com/aot/122615
fury x + 5820k 1.075.581 on dx11-mt
Aaaand i'm fucking sad now i've heard that there will be some hardware inside the 480 that will reduce overhead on dx11 but seems that it wasn't true :S
i7-5960X + R9 290X = 935.122 mt (1.075.955 st) draw calls on dx11.
https://www.guru3d.com/news-story/quick-test-directx-12-api-overhead-benchmark.html
>>55415822
Actually it's Crysis 3 that's incompetent, shill. Still peddling your driver overhead meme, eh?
>>55417949
>fx series
How about Intel i5? That's getting bottlenecked too in some more cpu intensive titles, more on AMD cards.
>>55418108
http://www.3dmark.com/aot/26856
Fuck this shit I'm canceling that Nitro 480 I had on preorder and getting a MSI Gaming X 1070 (If I can find one cheap enough).
>>55415822
>using a outdated benchmark and drivers from 2 years ago
get fucked also dx11 is dead
>>55418430
>crysis 3 is very demanding on the system
Yes, we know. It's why it's still used for many benchmarks.
>>55418829
>being forced to use windows10
I bet you love being pounded in the ass by Microsoft while AMD strangles your dick
>>55415822
28 FPS CPU bottleneck, holy shit.
>>55419025
It's sad because with AMD drivers, even the i5 is bottlenecking, putting the gimped 970 6 fps ahead of the 390
>>55416764
Hey, no need for a guy to be greedy. I took my profits and got out. So what you want, but the AMD execs did a great job of pumping up the stock price.
>>55417452
I remember when 320.18 killed my GTX 480.
>>55419449
I remember when crimson killed my HD7970
>>55418888
>implying you can even understand my post
You're clearly retarded. TW3 is THE benchmake game. Anyway, no modern game has core loading nearly as bad as Crysis 3. Even Fallout 4 can use hyperthreading.
You're the idiot who uses a 950 at 4k and then calls the situation "demanding."
>>55419916
except it didn't .
using afterburner with crimson was the issue.
crimson alone or afterburner alone did nothing .
>>55416432
>I'd like to try nvidia next time because his 670 was quite a bit cheaper.
Do you live under a rock?
>>55419220
Agree with that, they pulled a great bait and switch with the 480
>>55420173
But still, prices will only go up because more products still launch, which is going to be much more in the spotlight than a silly power issue that'll be fixed with drivers
>>55415858
AMD is using a nice trick, put more memory than needed on a 1080p card to market it as a better card
>>55420304
Yup dumb people see 8 gigs and they're like OMG WHAT A DEAL.
Then a 3.5 gig card beats it hilariously.
>>55420304
AMD can't compete at 1080p though, there's more cpu load at 1080p so unless you're going for unlocked i7 or i5 with a high end AMD card then you'll be losing out.
>>55420349
You do realise that the 970 is a 4gb card, don't you? But yeah, all memes aside I feel bad for people who fell for the budget AMD meme since most reviewers use 5960x, 5820k, or 6700k.
>>55420349
Normies (the market of 480) will eat it up. Keep in mind this has been done before
>>55420385
This.
Only the enthusiast tier market even follows review sites and shit, most people just buy what ever
>>55420377
>AMD can't compete at 1080p though, there's more cpu load at 1080p so unless you're going for unlocked i7 or i5 with a high end AMD card then you'll be losing out.
You cherry pick benchmarks and then generalize like a retard.
90% of games show absolutely no evidence of your meme.
Here is a DX12 title, the near future of gaming.
>>55417383
i got a GT610 for my old P4 machine. werks just fine, can run some games with a 30eur card. and another GT610 for the extra display ports on my moms PC.
>>55415822
Good thing the 480 is here now to completely and utterly blow le 3.5 meme card out of the water, at a lower price too.
>>55420942
What cpu?
>dx12 the future of gaming
You sound brainwashed
>>55421232
Idk about US, but gtx970 was still 40 aud cheaper than the rx480. 970 wasn't reference either.
>>55415822
my 7850 2gb card works just fine with my 8350 though
https://www.youtube.com/watch?v=lQzLU4HWw2U
looking at price / performance, AMD is pretty much king. For the top spot Nvidia and AMD are neck and neck.
I buy AMD because they contribute to open source instead of making shitty windows only proprietary shovelware.
1080 is a joke
>>55421810
>I buy AMD because they add nothing of value to the world with their shit contributions
Funny how your little open source king didn't even bother with Vulkan but is all over that closed source DX12.
>>55421278
No one said it doesn't, it'll just work much much better on a cpu with stronger ipc.
Like an overclocked i5 6600k
>>55421846
>supporting nvidia are jews
>>55421810
>looking at price / performance, AMD is pretty much king. For the top spot Nvidia and AMD are neck and neck.
Unrelated topic, unrelated pic, no overhead info no cpu info.
>I buy AMD because they contribute to open source instead of making shitty windows only proprietary shovelware.
Nvidia linux drivers are on par with windows it sucks that they're closed source but there's nothing near their performance under linux.
Amd closed source drivers are a joke, amd open source drivers (blobs required) are pretty good but still far behind windows drivers (that are not that good)
1080 is a joke
>It's expensive as fuck but it's a fucking beast.
>>55421810
What cpu did they use to measure performance?
Don't forget you'll be held back even by a haswell i5
>>55420942
My fellow posters, ROTR is a gameworks title, and that table is missing how super-overclocked the 970 is.
Compare apples with apples once non-reference 480s are out.
>>55421947
That pic is totally off topic too, we're talking about dx11 overhead but people is still trying to talk about dx12, benchmarks done with the highest ends processors only or just trying to turn this into an amd/nvidia battle.
Just got one of these in my rx 480 box wtf is this,m?
>>55422038
Shit, I hope I'm getting one too
>>55422038
>Copyright 2012
>>55416278
>>55415835
Comparing AMDs newest card with a 2 year old one. Nice to see theyre keeping up
>>55415822
Years of practice.
>>55422038
No you didn't.
>>55422126
No you won't.
>>55415822
asking how amd is bad on an nvidia infested game
what is next? asking why a f1 car is so bad on a rally course?
>proprietary and as such unverifiable benchmarks
>video games
>>55422190
Got any proof with amd infested games?
>>55416330
getting off dx12 hardware to dx11 hardware
ya dun goofed
>>55417452
the funny part with this is those started right when leather drag queen of nvidia ceo said they are working on the async problem..
till today the drivers are a complete lottery for many people
>>55422202
show me something amd has remotely close to gameworks
i dare you i double dare you
If more reviews would have included weaker CPUs on their benchmarks maybe amd would have fixed this already.
>>55422209
So what you're trying to say is that amd has no dx11 hardware?
i mean 6000, 7000, 200, 300, 400 series have near the same draw calls on dx11 and negative mt scaling (you can't fucking defend this).
because they're a company who market all their products to poorfags. seriously, all their marketing is always something like
"hey look at these benchmarks, we beat nvidia with our new new gpu and it's cheaper!" just look at the fury x marketing.
it's cheaper for a reason. also they're like 2bn in debt. i doubt they have the funds to employ a massive dedicated day one driver team like nvidia. nvidia are just better straight up if you have the money.
>>55422218
You mean mantle, don't you? Which is completely incompatible with nvidia GPUs, and any preGCN AMD gpu.
At least gameworks works on older nvidia GPUs.
>>55422279
Did you know that mantle is broken since 16.5.2 (16.5.3, 16.6.1 and 16.6.2) crashed every time with mantle amd said to me to stop using mantle and use dx11 instead.
Anyway we are going out off the topic again...
>>55422150
the 390 is the 290
it was released in 2013
>>55422318
Didn't they say they were going to use a hardware fix for the 480? What happened to that?
>>55422218
gameworks is cancer and should die
>>55422343
What? No. They're releasing a driver fix today. It was only a software/firmware issue.
>>55422343
¯\_(ツ)_/¯
Doesn't seems to increase performance nor draw calls on dx11 titles.
>>55422279
mantle evolved into vulkan, and it totally works with nvidia, you stupid little midget
>>55422356
Vulkan is lot more than just mantle.
>>55422362
so? mantle is deprecated. it's all about vulkan now. it's how vulkan started
>>55422343
the 480 was rushed out onto the market with some batches having shitty asic quality so they had to overvolt all chips to get as many units out to the public as possible and so all those gpu's work as intended. they're fixing it with software by reducing the core voltage. some cards will die though.
>>55422350
Not the power draw, the overhead issue.
>>55422380
no. vulkan was being developed under a newer form of opencl well before amd donated some of the mantle code. the vulkan project was started with mantle code but the stuff from the project before was also merged into vulkan so it's nothing like mantle anymore, it's evolved a lot.
who cares anyway, devs have no incentive to develop for vulkan over dx12 anyway. it was the same with opencl. go windows or go home.
>>55422350
They said they're going to make an announcement today, not a driver. Who knows when they will fix this fiasco.
>>55422424
Vulkan is supported on way more devices (steam machines, arm based devices, ...) and it works on windows too so it could be incentive enough.
And we're again out off the topic (dx11 overhead).
>>55418829
>dx11 is dead
in what universe you dumb amdrone
>>55422352
That's a shame.
>>55422458
so was open cl. that still wasn't any incentive.
>>55415822
best AMD CPU's are still based off of the Vishera lineup.... 4 years old or therabouts
>>55422523
well, i almost prefer it this way cause if they fix it using drivers it'll be fixed for all cards and not just 480.
>>55422539
That's true, but I'd rather they acknowledge and fix it instead of ignoring it and pretending it doesn't exist
>>55422529
OpenCL was way too bugged on amd to be of use (now it's better but amd opencl compiler is utterly shit)
DX11 came out in 2008. Companies started using it approximately in 2011 and there still were games using DX9.
Do not except that DX12 will take over market over night. We have 3-4 years untill real implementation instead of those "DX12 patches" adding few cherrypicked features for marketing purposes.
Also Vulcan > DX12
>>55422544
The only way for this to happen is lots of reviews using low tier i5 or med-high i3 on benchmarks.
>>55422564
But they won't because they don't want to end up like HardOCP
>VIDEO GAMES
>>55422844
So you're surprised that we're talking about games on a dx11 overhead thread?
>>55422560
>Also Vulcan > DX12
This.
Vulkan is supported by all recent hardware and operating systems. Why would anyone give a fuck about DX12?
>>55422844
Why are you even here?
Not sure, but I recently just got my 480 and it's been a blast.
GTA 5 on very high barely ever dips below 60 FPS on 1440p , DOOM fluctuates between 45-70 FPS but the average is 60 (on ultra 1440p).
On 1080p both these games never dip below 60.
>>55422979
What CPU?
>>55422979
I'm not sure if driver overhead is a meme cause im running my i5 4690k at 4.4ghz.
>>55422988
>>55422979
Adding to this, I've came from a 770 2GB card and the performance has been nearly twice as good.
>>55423005
well, you wouldn't erally suffer overhead on an overclocked haswell, but someone with an rx470 or rx460 and i3 6100/i5 2500 will definitely suffer from it unfortunately.
Especially the older CPUs like i5 3470 or i5 2500 with the weaker ipc
>>55422988
as mentioned above (>>55416405)
>Amd is fine but you need as much single thread performance as you can get.
>Then it'll perform great.
The problem comes with weak st performance cpus
>>55415822
>mfw people actually support the green jew
>mfw people believe this
How many shekels do you get paid for this?
>>55423064
>mfw people still confuses dx11 driver overhead with OPENGL
>mfw people still confuses this thread as an AMD vs NVIDIA therad.
How many shekels do you get for derailing threads?
>>55423064
>recent submissions submitted by Anonymous
totally representative of the performance of the card
Keep dreaming shill, better dead than red
>>55423064
the only outlier there is your shitty photoshopped score, faggot.
>>55423064
http://www.ozone3d.net/benchmarks/furmark_score_190.php?id=118123
>mfw this was supposed to be reliable in anyway
>>55423137
>mfw 7970 has more than double performance than your 390 clock/clock
>>55423064
>mfw 7700 is the new god
not shopped check yourself
http://www.ozone3d.net/gpudb/scores-geeks3d-furmark.php
>>55423195
So now that it has been demonstrated that this "as a benchmark" has the same validity as generating random numbers and putting them on a graph, have you anything more to add?
>>55423246
was for
>>55423064
>>55421251
The RX 480 goes for 270-300€ here in my Country. Last week i bought a GTX 970 for 250€.
>>55423261
RX480 is ~450aud, I can get a gtx970 for 410aud, non reference.
Right now there's no reason to buy an rx480 here, until prices drop or we get similar priced aftermarket ones with better cooling and power delivery.
>>55422961
Neo-/g/ everyone
>>55423305
Why are you even here?
>>55423261
I'll never buy a 970 nvidia doesn't deserve to sell that gpu after the 3.5 fiasco, and the 480 was launched in such a hurry that it looks like a bad joke.
I'll like to buy a mid-range card with the best dx11 and dx12 performance but right now you have to choose just one of them.
>>55415835
>>55415858
>>55417509
>>55417593
>>55420349
Friendly reminder the 3.5 VRAM meme is just a goalpost used by AMD shills and it has little to no effect on gameplay. Webm very fucking related.
RX 480 will get fixed by third parties, it doesn't matter what the reference board has because you shouldn't be buying reference boards.
The fact that so much attention was brought to this means it will be fixed.
Definitely won't be buying an RX 480 to replace my R9 290 reference though. I paid $182 for it nearly 3 years ago and it's still faster...
>>55423275
I wanted to buy one, but then i saw the Benchmarks which stated that the RX 480 is only better in Hitman and DX12 titles. Why should i buy a card which performs worse with a higher price. Thats a fucking joke. I was always an AMD fag. I loved their cards had an R9 270x and then upgraded to an R9 290. The 290 overheated by up to 100°C and it broke. Amazon refunded and i stopped using AMD cards. Their CPUs are still good and my FX-8320 performs like a champ, but im staying away from their CPUs. My GTX 970 doen't even get near 100°C. Full load on a hot day is around 60°C.
>>55423328
3.5gb was blown out of proportion, the GPU can still access and use all 4gb and the 970 performs exactly as it should at that price point.
>>55423362
there was no info from nvidia about this issue previous to it being discovered by an user, that's not a good practice.
>>55423349
>AMD shills remain silent
lmao btfo.
>>55423389
It's not, but it's not like it changed anything about the card.
That's like being upset the gtx670 is a cut down 680, or that the 290 got refreshed as the 390
>>55423351
>Definitely won't be buying an RX 480 to replace my R9 290 reference though. I paid $182 for it nearly 3 years ago and it's still faster...
How on earth did you get a 450$ card for 182 3 years ago?
>>55423437
He probably stole it and the theft cost $180
>>55423362
That's not true at all. It is a fact that if the 970 tries to access the last 0.5 gb of vram, your fps will tank very hard.
>>55423460
Afaik they do something on the driver side so it doesn't drop that hard... but it doesn't excuse them for hiding this from the users.
>>55423494
It still uses 4gb, they didn't lie about the vram at all...
>>55423515
Is not black is not white but it's a pretty dark grey.
>>55423515
They didn't lie, they just omitted certain facts that would've made some people think twice if they had said so in the marketing
> hey guys this card has 4GB VRAM!
> why does my game turn to shit if it uses over 3.5GB?
> oh yeah turns out that if the last 0.5GB is accessed, it craps your game out
> why didn't you tell us this?
> haha come on it's not really that big of a deal you guys
> we fucking paid for a 4GB card
> and you got a 4GB card haha
> we can't use 0.5GB of it
> yeah you can it just runs badly, but it's still usable
> what the fuck nvidia
> come on we didn't do anything wrong its not our fault haha come on guys we didn't lie at all
>>55423437
bitcoin difficulty increased and miners were selling R9 290s
>>55423659
Funny how it runs 4gb games fine, like gta v or even shadows of mordor which is 6gb iirc
>>55423805
You do realise that VRAM usage is variable, right? Just because the Minimum or Recommended specs say 4GB or 6GB, that doesn't mean that's what uses. There's no standard for the Min or Rec specs, so the developers can put whatever they want.
>>55423830
Yeah, I know.
Watch some YouTube videos though of people playing games fine on a 970 going over 3.5gb, it's not as bad as /g/ makes it out to be, although you do see more of a performance drop when compared to a 980.
>>55422279
so when you say mantle you obviously mean dx12 too
good to know what road paid shills are going again lol
>>55423862
I saw someone play Far Cry 4 and it literally become unplayable - like sub-10 FPS - when it reached like 3.6GB
>>55423923
I once saw C-beams glitter in the dark near the Tannhauser gate.
>>55422350
they are not releasing a driver fix today they said they will update the community about their solution
>>55423908
Mantle != DX12
>>55423862
we all know there's literally non issue with going above 3.5gb utilization on the 970 unless you're trying to play a game like gta at 4k ultra settings which will then cause some stuttering because of the reduced bandwith and read/write of that last segment when the game is trying to rapidly move around and load resources into the game. your fps will be like 5fps anyway so it's literally a non issue at playable settings/resolutions.
>>55420942
>ROTR
Nvidia exclusive game
>>55424001
literally nvidia broke the game and they are actually using it as a bench
just like arkham knight
>>55416435
Congratulations! Today it's your turn to man the Team Green Gloryhole!
>>55415822
One title, muh ~10FPS difference.
Only retards take your shitty bait thread seriously.
>>55424073
Why are AMD shills so stupid? Can you please answer that? Is it intentional?
Typical fanboy reaction, instead of discussing and transmitting our concerns to amd is just better to derail the thread turning it on a dx12, mantle, or 3.5 meme thread.
Amd needs to take nvidia as an example of dx 11 overhead and multi-thread, not as an example of how to gimp memory nor on how to do async compute on dx12.
I'm an amd user and I'm fucking disgusted that every time someone criticizes amd someone comes to derail the thread using the buh-buh-buh-buh-buht nvidia has 3.5 gb.
>>55424089
Do you know who thinks that all the other people are stupid and he is the only one smart?
Yeah, you got that right, usually the dumbest of people.
>>55424109
You mean people who call others retards because it makes them feel better? Like
>>55424073
>>55424104
>I'm an amd use
Post Speccy with timestamp. Thanks (, nvidiacuck shill).
>>55424104
Thank you. I'm glad not everyone on /g/ is a 12 year old attention whore.
>>55424073
So you failed to realize that the graph is showing that the 390 is better than the 970 but it gets bad results cause driver overhead?
Is not a nvidia's thread, at least for me is an AMD appreciation thread where we can appreciate how great most of the amd hardware is and how much performance is wasted cause the drivers on dx 11.
Just with a software update that graph could be amd at the top almost everytime, but cause the overhead we have to watch a 390oc performing at 40fps on crysis when it should run well above 60.
>>55424123
>>55424123
>Amd fanboys 101
Remember kids everyone who doesn't want to derail a thread where someone shows an amd issue is a nvidiacuck shill.
>>55424104
multithreading on dx11
MULTITHREADING ON DX11
stop it just stop it...you will never know what multithreading is
>>55415822
AMD is cheap
>>55424304
Civ5 buh-buh-buh-buh-buht dx11 can't into multithread.
You can find more info here:http://forums.guru3d.com/showthread.php?s=e317da9a19e92da19ae864005315d7dc&t=398858
>>55424340
post me 20 dx11 games that actually was multithreaded
i the sense they could use almost every single thread of a cpu
no? you cant? meh.. just cherry picked once more
>>55424340
dx11 can't into multithreading, you are right.
Game engines can.
>>55424396
Can you please retpye that and make a sentence that actually makes some sense?
What you typed is garbled
>>55424418
>>55424396
Samefag. You can tell
>>55424396
The overhead test on 3dmark uses dx11 multi-threading, also bf4, hardline, bf1...
>>55424418
I'm not gonna waste more time with you cause you're not even reading what i said, READ THIS:
http://forums.guru3d.com/showthread.php?s=e317da9a19e92da19ae864005315d7dc&t=398858
>We are especially hopeful about a faster shift to DX11 because of the added advantages it will bring even to DX10 hardware. The major benefit I'm talking about here is multi-threading. Yes, eventually everything will need to be drawn, rasterized, and displayed (linearly and synchronously), but DX11 adds multi-threading support that allows applications to simultaneously create resources or manage state and issue draw commands, all from an arbitrary number of threads
Go search some info about deferred contexts and command lists on dx11 and then come here so we can have a civilized discussion.
>>55424467
mantle games and bf1 being dx12 is now dx11
yay
good job everyone
http://www.anandtech.com/show/2716/3
literally a 7 years old article being brought on 2016 as relevant because you know paid shills are being so stupid
>>55424536
I'm so sorry for you not knowing that ALL mantle games has a dx fallback so they're not amd only games.
>>55424536
>calling people shills/stupid
Pot and kettle spring to mind.
>>55424536
>literally a 7 years old article being brought on 2016 as relevant because you know paid shills are being so stupid
Go to the dx11 documentation look for deferred contexts and command lists, also look for multithreading on dx11 on microsoft web I'm sure I've found some samples there a few years back.
>mantle games and bf1 being dx12 is now dx11.
I've reported a bug that since 15.5.2 all mantle games crashed loading, AMD reply was stop using mantle and switch to dx11 instead.
As for Battlefield 1 dx12 is an option too (pic related).
>>55424467
Apparently somebody has been living in a cave because developers have genuinely been complaining about how heavily DX11 relies on a single thread. Engines can definitely split all the work they do internally into multiple threads, but drivers for implicit APIs like DX11 and pre-4.X OpenGL still do most of the work in a single thread.
This used to not be that big of an issue when you didn't have that many objects on screen at once and they didn't consist of as many draw calls, but now that you have a lot of objects on screen at once and they all consist of several draw calls and this is an issue as frame rates end up becoming bound by single thread performance. DX12, Vulkan and many of the enhancements to OpenGL from 4.0 onwards were specifically created to solve this issue.
>>55424648
Of course Dx11 is worse than dx12 but amd can get at least 3 times the draw calls they've right now.
And we're discussing driver overhead on dx11 and you're just trying to derail the thread and most of the time in a wrongly manner and spreading lies around, I've proven you were wrong and you have proven that you can't read so i'm fucking done with this shit.
And if you can fucking read here >>55417593
>we're not discussing 3.5 fiasco, faulty killing drivers, pci-e out of spec problems, 2d acceleration bugs, Opencl compiler bugs, gameworks, vulkan, dx12...
>we're discussing driver overhead >>55415822 (OP) and related >>55416330 (You) negative mt scaling under dx11.
and if you can read here >>55424172
maybe you can realize that we want better drivers so amd cards can shine.
Talking with you have been a fucking waste of time cause i feel no one has learned anything.
>tfw r9 390 + i5 4460
JUST
>>55425076
yeah it seems you're loosing some gpu power on dx11, pray for dx12 support on every game for now on so you can get full performance.
>>55423195
Do i understand something wrong or did a 750ti beat a 980ti? Is lower better?
Is this just for shit and giggles?
I'm not into this benchmark stuff. Don't care for /g/'s dick comparision and shilling.
>>55425752
Furmark has no validity as a bench some faggot tried to derail the thread using this benchmarks but according to the results the 7700 can score higher than a quad crossfire of furyXs :P
Also it's opengl based and we're discussing dx11 overhead.
>>55425138
Or just, you know, buy from a competent manufacturer next time.
After overclocking my R9 290 to relative 480 perormance (actually better for the most part) I still could not get ROTR to not stutter on very high settings. So I cancelled the Nitro 480 for the 1070.
>>55422011
The pic isn't off topic - it shows that there is no "driver overhead meme" bottlenecking AMD GPUs in RotTR. This is the case in 90% of titles, DX11 included.
>>55426260
>We're discussing dx11 overhead and the pic is showing dx12 results so no, it's not related.
>>55426191
>yfw the 1070 stutters too
>>55426296
>rotr
>dx12 path
choose one..
>>55426477
It could be a half assed dx12 game but it's not a dx11 game (maybe it has the option to use dx11 i don't know cause i haven't tested it myself but sure thing the pic you were referring at it's showing DX12) so it's not related to this thread.
>>55426581
it has vxao over the dx12 it most surely isnt dx12
might use 1-2 things but thats it
>>55426634
The fact that it uses dx12 features is what makes it dx12...
Brand loyalty is disgusting and you should all feel ashamed you consumer whores.
>My God is better, no my God is better"
They both occasionally kill a baby, but one of them won't let you fuck dogystyle and sends the plague when you get old to test your faith.
>>55415835
Designated
>>55415858
why is /g/ so fucking terrible now
this is some preteen level bullshit
>>55423349
>thinks massive frame dips have no effect on gameplay
>>55428770
3.5 is a hardware issue that nvidia synthetically fixed with drivers.
Overhead is a software issue that AMD is refusing to fix.
>>55428809
that AMD is too incompetent to fix*
>>55428809
Did you even see the diagrams showing how the 3.5gb issue is impossible to fix in software?
It's like if you bulldozed a freeway and tried to convince everyone that it has no effect since you can take another freeway that's twice as long to get to the same location.
>>55428851
Gtx970 uses all 4gb of its vram without issue thanks to drivers.
Sure, it might stutter without drivers, but that's not my point. AMD could dominate the market instead of falling down to 20% if they didn't love being so incompetent.
>>55423349
https://youtu.be/Bm5ZCJah-FY?t=32s
