As if you forgot, right? Today, the real next generation in gaming begins, with the release of the NVIDIA GeForce RTX 3080 as the first in the desktop Ampere architecture.
Need a reminder of just how ridiculous and powerful the RTX 3080 is? Here's some specs:
| GEFORCE RTX 3080 | |
|---|---|
| NVIDIA CUDA® Cores | 8704 |
| Boost Clock (GHz) | 1.71 |
| Standard Memory Config | 10 GB GDDR6X |
| Memory Interface Width | 320-bit |
| Ray Tracing Cores | 2nd Generation |
| Tensor Cores | 3rd Generation |
| Maximum GPU Temperature (in C) | 93 |
| Graphics Card Power (W) | 320 |
| Recommended System Power (W) (2) | 750 |
| Supplementary Power Connectors | 2x PCIe 8-pin |
Additional details: they will support the latest Vulkan, OpenGL 4.6, HDMI 2.1, DisplayPort 1.4a, HDCP 2.3, PCI Express 4 and support for the AV1 codec.
Stock is expected to be quite limited, especially since they did no pre-ordering and stores will likely sell out quite quickly. Even so, here's a few places where you might be able to grab one. Some of the sites are under quite a heavy load too due to high traffic, so prepare to wait a bit. I've seen plenty of "website not available" issues today while waiting to get links.
UK
USA
Feel free to comment with more and we can add them in.
Driver Support
Along with the release, NVIDIA also put out a brand new Linux driver with 455.23.04. This is a Beta driver, so there may be some rough edges they still need to iron out. It brings in support for the RTX 3080, RTX 3090 and the MX450.
On top of new GPU support, it also has a bunch of fixes and improvements including support for device-local VkMemoryType, which NVIDIA said can boost performance with DiRT Rally 2.0, DOOM: Eternal and World of Warcraft with DXVK and Steam Play. Red Dead Redemption 2 with Steam Play should also see a bug fix that was causing excessive CPU use.
The VDPAU driver also expanded with support for decoding VP9 10- and 12-bit bitstreams, although it doesn't support 10- and 12-bit video surfaces yet. NVIDIA also updated Base Mosaic support on GeForce to allow a maximum of five simultaneous displays, rather than three. For PRIME users, there's also some great sounding fixes included too so you should see a smoother experience there.
Some bits were removed for SLI too like "SFR", "AFR", and "AA" modes but SLI Mosaic, Base Mosaic, GL_NV_gpu_multicast, and GLX_NV_multigpu_context are still supported. There's also plenty of other bug fixes.
What's next?
Today is only the start, with the RTX 3090 going up on September 24 and the RTX 3070 later in October. There's also been a leak (as always) of a RTX 3060 Ti which is also due to arrive in October. Based on the leak the upcoming RTX 3060 Ti will have 4864 CUDA cores, 8GB GDDR6 (no X) memory clocked at 14Gbps with a memory bandwidth of 447Gbps which means even the 3060 is going to kick-butt.
Are you going for Ampere, sticking with what you have or waiting on the upcoming AMD RDNA 2 announcements? Do let us know in the comments.
Article taken from GamingOnLinux.com.
{
"file_format_version" : "1.0.0",
"ICD": {
"library_path": "libGLX_nvidia.so.0",
"api_version" : "1.2.142"
}
}
It would be nice if AMD could finally put something competitive out, but seeing that the 5700XT barely breaks half the performance of the 3080, it's not likely the RDNA2 will suddenly compete with the 3080.
What's much more interesting to me is that the rx5700xt numbers are not all that much behind in the reviews (price/performance ratio) and that I am already seeing those cards in the used market for *very* reasonable prices, which I expect to get even more tempting in the coming few months.
Last edited by emphy on 17 September 2020 at 2:27 pm UTC
Quoting: The_Aquabatbecause synthetic benchmarks are not real world gaming.
100% increment in performance is just a buzzword.
They did 2 things here:
1. compared the 3070 to the 2080ti, then compared the 3080 to the 2080 (non super). They primed your brain so when you hear "twice as fast" you think "twice as fast as the 2080ti"... Even though 25% faster than the 2080ti is exactly what their graph (shown in the article) roughly says.
2. pushed REALLY hard for RTX numbers. Linus Tech Tips says Nvidia reached out and told them "oh no, the double performance thing was for Quake RTX and Minecraft RTX". In reality, the 3080 is only 25-75% (depending on game) faster than the 2080. (which still leaves it faster than a 2080ti by a pretty nice margin).
What really bothers me the most is that this is in fact a new fastest card. it IS faster than the 2080ti, at like half the price. No one buying this is getting ripped off, even at the lower numbers (unless AMD has something even faster for even cheaper coming out in a month, and even then, worst case, it will be comparable). So why even bother to fluff the numbers. The only reason is because most 1080ti owners didnt upgrade to 2000 series. Which is a good thing in a way because of e-waste and what not... But nvidia this time around really wants everyone to feel the need to upgrade, even if they dont need to.
I personally am holding out for some big navi news. It looks like the driver situation is already making some progress.
Quoting: The_AquabatAnd additionally looking at the power consumption ( 320W!!!!!Quoting: EhvisInitial benchmarks on sites dealing with less useful operating systems shows a fairly consistent gain of 25% over the 2080 Tibecause synthetic benchmarks are not real world gaming.
100% increment in performance is just a buzzword.
) seems to me that they simply added more cores / computing units and performance increase of the actual hardware is somewhere around zero.So no reason for me to even think about replacing my 2080Ti. If there'd be a card with same performance & memory and much lower power demands ( -> less heat / noise produced ), I might think about a replacement. But it doesn't look like NVidia can deliver that at the moment.
Quoting: peta77And additionally looking at the power consumption ( 320W!!!!! ) seems to me that they simply added more cores / computing units and performance increase of the actual hardware is somewhere around zero.
The funny thing is, they actually changed things quite drastically compared to previous generations. This can be seen with high number of "cuda cores". Is I understand they have been messing with the balance between fp32 units and other components. Which makes the excessive power requirements all the more surprising. NVIDIA won't be getting away with this for another generation I think. So I'm happy to wait that out while looking to see what AMD is going to do.
someone not waiting for AMD's release (soon) before buying a 30XX must be outright crazy. :D
Patreon
PayPal
See more from me