AMD Radeon RX 6700 XT could be slower than GeForce RTX 3060 Ti
Click here to post a comment for AMD Radeon RX 6700 XT could be slower than GeForce RTX 3060 Ti on our message forum
cucaulay malkin
that would be disappointing.
amd says it's going against 3070.other sources say it's 3060ti/2080S.Usual Radeon stuff,huge differences across game engines and api's.
e.g. 5700xt - take best 10 games and it's same as 2070 Super or better.Take worst 10and it's barely catching up with 2060 Super.
with Radeon cards nothing is white or black.
On average this will be between 060ti and 070.
I'm more excited about the dlss equivalent they promised.
the card too,I'd buy instantly were it available.It won't be tho.
illusiveman
Yeah it's slower in that specific workload, no surprises.
If you search in geekbench openCL database a RTX 2070 has 82914 points and a 5700XT has 68390 points. Does that reflect real world gaming performance in any way ? Nope....
kapu
Thats BS benchmark . Lets wait for real gaming performance , that only matters unless you play benchmarks 🙂
mikeysg
I usually ignore benchmark scores from Geekbench and Usuerbenchmark, they do NOT reflect real world gaming performance. Now, IF that article had 3DMark scores in it, or (Heavens forbid) actual gaming performance benchmarks, I'd sit up and take notice.
Bitey
Well It started with the 5700xt, AMD started designing gamer specific cards and the compute component performance dropped markedly... if you want compute from AMD its the "other" cards.
Fox2232
Why do you think you have baseline set correctly?
What if AMD's performance is actually what you get by following standard implementation according to DX/Vulkan/... And therefore AMD's performance is baseline.
And nVidia's is all over the place where they over-optimize some games to point where 2060S matches 5700 XT. And not optimize some properly, so 2070S performs worse than 5700 XT.
I hope you see where your line of thinking went wrong. Apparently, neither AMD's nor nVidia's performance is the baseline. Each pushes optimizations for different games/engines differently. And both use more or fewer hacks in process.
- - - -
As for availability. Those are smaller chips (much more per wafer as smaller defect rate does wonders). And if lucky 2GB GDDR6 @16Gbps have sufficient availability too.
Fact that they still did not announce 6700 is saying something.
- - - -
And to actual performance... value given is for OpenCL. It is like joke territory. If anything, one should be comparing RDNA1 to RDNA2 at known cards (5700 XT to 6800/6900) level 1st to get scaling. And then try to properly extrapolate actual gaming performance.
rl66
If slower and aviable... then why not? it's a win for AMD...
But i am not too convinced by the aviability...
cucaulay malkin
except that only happens when 5700xt performs badly due to amd card's well known flaws like dx11 performance and such.when ported to dx12/vulkan they get a performance increase.So NO it's not nvidia performing great,it's nvidia performing more solid where radeon is lacking.
i have everything set correctly,don't worry.it's you who's trying to spin everything.
I wonder if I asked you the same about 980Ti in early dx12 titles.
well,it's doing fine,you just don't have the baseline set correctly for maxwell.it certainly doesn't lack proper dx12 support 😛
this must be the baseline performance for 5700 then
[SPOILER]
https://i.imgur.com/jdloM8U.jpeg[/SPOILER]
taking extremities as the baseline is wrong conceptually.
baseline for a GPU is average of all engines and api's,and resolutions.that's why I think pcgh's method is the most objective way to answer the question "how fast is X compared to Y".it might not be true for an individual,but it's true as a concept.
20 games,all api's,most prominent game engines,four resolutions averaged
https://www.pcgameshardware.de/Grafikkarten-Grafikkarte-97980/Specials/Rangliste-GPU-Grafikchip-Benchmark-1174201/
take baseline for navi and turing,then you can compare two.
kapu
[youtube=JLEIJhunaW8]
It seems Amd has better drivers this round especially in cpu limited scenario.
cucaulay malkin
hw unboxed,always there to tell you the part of the story that you want to hear without having to worry about the one you don't
https://www.purepc.pl/test-ryzen-7-5800x-vs-core-i7-10700kf-na-rtx-3080-i-rx-6800-xt?page=0,13
kapu
Thats different test . And HU is trusted with methodology so , there is nothing wrong with their results , bit of a suprise but a good one tho 🙂
cucaulay malkin
oh yeah the're objective and definitely not there for amd fanbase clicks 🙂
https://i.imgur.com/rPQ0qII.jpg
as for the problem,I think I've seen it on pascal a few years ago already.same situation but worse cause as gpus keep getting faster,gap keeps growing.nvidia fixes some of those games,sometimes sooner,often times not so sooner.
hu picked two games ( 🙄) one of which is horizon that's completely broken on nvidia and just very recently got patched by the studio.
great methodolgy and everything like such as ......
difference is,in case of ampere and dx12,an adequate cpu fixes the problem.with amd driver overhead,it doesn't.all power goes to waste.
kapu
I think you are missing the point 🙂 nVidia has bigger overhead in DX12/Vulkan and AMD has problem with DX11 ( not that i care for DX11 since it will become irrelevant very soon).
cucaulay malkin
no I did not.
how did you get that from what I wrote ?
read that again
it's a fair point,but pairing a 1600/2600/10100 for a test with a 3090 and testing two games..... too limited to make too broad generalizations.
but the problem is there,like I said,I've seen it with pascal already a couple of years ago.
itpro
Leave HU alone. They are doing great job. They do not lick Nvidia/Intel like most of tech people.
kapu
I got that from purepc test you linked and it actually shows similiar situation to HU in same games . nVidia has problem, in my opinion not a small one .
cucaulay malkin
which is exactly their strategy.every YT channel has to cater for their fanbase.I don't blame them.It's what you do on YT.they're doing it very well.
their methodolgy is misleading.
in general,YT reviews are pretty worthless.
if you want objective tests go to an actual tech site.
people are too lazy these days tho.they want to be fed headlines,opinion pieces,and not do the work.
fine by me.
itpro
It is quite easier for people to accept this: " Intel is perfect, Nvidia is flawless ", "AMD is the cheaper alternative, ATI has some serious issues". I get it. That's how people grew up or got paid by the industry, I cannot blame them. I am far too many years in IT business, I've seen lots of things.
kapu
They test they did is valid and shows actual problem, there is alot people sitting on 1600-2600X Ryzens or other weaker CPU and think on GPU upgrade , there is significant performance difference with CPU bottleneck in DX12/Vulkan titles .
cucaulay malkin
im saying the way they do things is so broad they always leave themselves so much wiggle space for whatever opinions they wanna present
look at this one
they took two games for a whole review.does that mean the problem is 2 games from the possible hundreds of dx12 titles ? Or it is just the two they wanna base their whole opinion on ?
paired three $100 budget cpus with a $1000/1400 cards.there is no in-between cpu like 3600,10400 or 3700x.Why the hell test four cpus when three of them are the same entry level ?
how do you judge the extent of the problem from this ?
with HU you're fed whatever you want to be fed,you said it yourself - being anti intel and nvidia is their main thing.first thing you said when you referred to the video.
not I like HU cause they have good testing methodology.I like them cause they talk smack about nvidia and intel.
