They said it has 60 CU.
It is the cutdown version of Radeon Instinct MI60.
Die Size 331 mm2
So you are paying 699 usd for a slightly bigger than the GTX 1080.
Meanwhile Master Race RTX 2080 has Die Size of 545 mm2.
>B.b.but AMD is the good guy, nvidia is bad guy.
AMDfags BTFO.
Other urls found in this thread:
youtube.com
twitter.com
All I care is open source drivers on linux. AMD has that with their amdgpu driver. AMD did that and I'm happy.
>16GB HBM2
Gee, I wonder why it's so expensive?
16GB of HBM2 VRAM makes this sound nice for Blender OpenCL rendering.
Im going to buy the 2060 now
>for a slightly bigger than the
What did they mean by this?
>Meanwhile Master Race RTX 2080 has Die Size of 545 mm2
>People offering serious replies to this post
I for one commend OP.
do you want low performance ray tracing? just get the 1160
>Blender OpenCL rendering
Except that blender will probably soon have raytrace acceleration using RTX. Which will BTFO radeons.
>what is 7nm vs (((12nm)))
Nvidiots are truly nvidiots.
Radeon Instinct MI50 = 3840 cores
Radeon VII = same specs as the MI50 except clocked a bit further 1800mhz vs 1754mhz
AMD already had this chip released back in 2017 if they put the full 1 TB/sec HBM2 on it. Vega64 kept gaining performance even as you increased the memory bandwidth without increasing core clock.
Has Jow Forums gone full retarded? Can we quit telling people to buy cards that dont exist?
it s even worse. the fact than a 7nm can barely match a 12nm gpu ( and a 14nm 1080ti) it s sad and pathetic.
also their margins must be so fucking low if they priced it 699.. at 699 i just buy a 2080. if it was 499 then it could have been a real threat to nvidia .
The gpu they just put out if they just made Vega 64 with 4 stacks of HBM would've given them more or less the same perf as a 1080 Ti back when it mattered...
>So you are paying 699 usd for a slightly bigger than the GTX 1080.
What you're actually paying for is HBM. Whether using HBM is a mistake or not for such reasons is a good question, but if there's anything VII is going to have, it is tons of bandwidth. Which is going to matter on workloads that are primarily bandwidth-bound.
>raytrace acceleration
Cycles, the rendering engine, uses pathtracing, and I don't think any part of that can be accelerated with Bounding Volume Hierarchy(BVH) acceleration found in the RT cores, because path tracing relies on bouncing the ray from the source until it hits a light source or some bounce limit which then starts to average out all the rays that have been sent out to get a final pixel color. BVH won't provide any help since it is volume based as its name suggests and none of what path tracing does is volume based for anything. If you bound the rays you send out, you'll get holes along with graininess in your image. But in return you get stuff you don't get with traditional ray tracing like soft shadows, caustics and global illumination.
And yes, I know caustics were mentioned in one of the Nvidia demos at CES but they are cheating there too like what is done with traditional lighting. It's still amazing but notice how the caustics were area bounded to the bridge. If the caustics were naturally acquired with path tracing, it would be on the canal walls too around the boat and under the bridge after the caustics reflected bridge which had no caustics at all.
Calm down shill. It requires rewriting cycles from scratch, and will not be done before years. The developers have said it.
This
Heck I'm hackintoshing and nvidia has no more drivers. Looks like I'm an amdrone now
I would totally but this card if it cost 100 less. But apparently this is going to be a limited, low volume part, so I think that it's actually going to cost even more than that.
AMD is winning the CPU share market, but the fucking GPU marketing and sales units should be brought out behind the shed and peppered with bullets.
strictly on a gaymen basis
but watching the video on Forza Horizon 4 the game gets basically 120 fps at 1080p
now i dug out some RTX 2080TI benchmark with Forza and it basically had a minute difference in FPS
now this isn't conclusive at all, since the maps are all different and the possibility that the guy benchmarking himself was a liar but if it's true then the difference between 2080ti and VII could be only 3-5% in some titles
video for reference
youtube.com
Unless Lisa earmarks the same kind of R&D budget for GPUs it's not happening. The fact that RTG wasn't even able to afford more than one or two dies (minus semi-custom, because that's paid by the client) per generation should tell you how shoestring the budgets are.
Yeah. Some people keep blaming Pajeet Koduri, but it really isn't his fault, in fact that's apparently why he left: his Vega team was drastically reduced to work on Navi (for Sony), and Vega ended up being shit.
Can't blame him for moving to Intel, but at the same time it makes sense for AMD to invest on CPUs and custom stuff for consoles for the moment. Let's hope they start being competitive in the GPU space too.
Good move desu navi wasn't ready and rtx will make them compete in 202x with Arcturus
Ah yes, Jow Forums the board of disinformation and hyperbole.Things never change.
I've been on nvdia for years starting with 8500gt. All the way up to my current 970.
I've never had ati/amd card.
AMD not delivering is a huge disappointment. I've been rooting for then for quite a while in the gpu sector because here's absolutely no competition and nvdia is milking the fuck out of us.
Remember when high end cards cost 400 bux?
You fucking retards are cheering on your own demise. No one wins here except nvdia. Cheering a corporation on like its your favourite football team is the most retarded shit ever.
Are you all eternally 12 except from console wars you moved onto corporations?
Jesus fucking christ
We laugh at deluded AMDfags but cry at the actual market
Nvidia is irrelevant on Jow Forums, until they step up with their open source Linux drivers.
protip: they won't.
Go, and BTFO AMDfags on /v/.
>Cycles, the rendering engine, uses pathtracing
Still they could use the tensor cores for denoising.