This article is saying that GPU speed isn't the main reason for upgrading to Radeon VII plus a few more interesting articles at this link: AMD's speedy Radeon VII GPU is proving Nvidia's point | VentureBeat
AMD announced its new Radeon VII video card today. It’s a $700 GPU that may have similar performance to Nvidia’s RTX 2080, which ranges in price from $700 to $800. This finally makes AMD competitive with Nvidia again on the high-end (if you ignore that $1,200 RTX 2080 Ti). But the Radeon VII may actually help Nvidia make its case about the future of 3D graphics.
From about 2006 to 2016, graphics-hardware engineers strived to meet the rendering demands of developers. The goal was to run 1080p games at 60 frames per second. That struggle came to an end about three years ago when AMD and Nvidia both launched $250 GPUs that can almost any new release at 1080p60.
Modern CPUs and GPUs are so powerful and fast that they’ve essentially conquered 1080p. Developers have so much headroom at this resolution that they are unlikely to need anything more than a $300 GPU.
If we’ve conquered 1080p60, then why do we need more powerful video cards? Well, even if rendering budgets are going to stay within the confines of a Radeon RX 480, display technologies are continuing to progress.
Gaming fans especially are looking to higher resolution monitors that are capable of more than 60 frames per second. This includes 4K panels, but it really means 1440p monitors with refresh rates of 120 or more. On the Steam hardware survey that tracks what hardware people are using, 2560-by-1440 displays are the third most popular resolution (3.89 percent) after 1920-by-1080 (60.72 percent) and 1366-by-768 (14.04 percent).
But running a game at 4K or 1440p144 does require a lot more power than you can get from a $250 card from 2016. That’s where something like the Radeon VII begins to make sense.
The problem for AMD is that 1440p isn’t going to require an astronomical leap in power. And while more speed is great, it’s maybe time to start asking what else a new generation of graphics hardware can do beyond run the same kinds of games 25 to 30 percent faster than before.
Since introducing its Turing generation of RTX graphics cards, Nvidia has tried to shift the conversation away from speed. The company has spent the last several generations focusing on power, but now it is talking about new technologies like ray tracing, deep learning, and variable-rate shading.
Meanwhile, AMD’s approach is a stark contrast that feels like the company is missing an opportunity.
And to be clear, I think now is an obvious opportunity to do more than speed.
Yes, running games at 1440p144 and 4K is awesome, but it’s not something most people need. Again, 1440p monitors make up under 5 percent even if you include ultrawide 1440p displays. And 4K? It only has a 1.42 percent market share on Steam, and that fell 0.03 percent in December.
This doesn’t mean that 1440p isn’t growing. It is. It’s the only resolution category above 768p besides “other” that saw a month-over-month increase in December. But it’s going to take years to fulfill its destiny as the new display standard.
Even for people who have a 1440p-or-higher display, the Radeon VII is in a bit of a tough spot when it comes to price. For 1440p, you can get by with a more affordable RTX 2070 at $500. That’s especially true because the games that benefit most from super high framerates are typically better optimized with fewer visual effects. You don’t need an RTX 2080 or Radeon VII for Rainbow Six: Siege, Overwatch, or Counter-Strike, for example.
And if you want to play in 4K, that means you can probably afford to step up to the RTX 2080 Ti for $1,200.
I’m not trying to claim that Radeon VII isn’t going to have a great bang-for-the-bunch. I bet it will.
Faster video cards are not that exciting. It’s great that AMD is creating some competition for Nvidia. I am hopeful that Radeon VII will drive prices down for RTX cards. But that’s about it.
The point is that at current render budgets, faster GPUs are going to get noticeably diminishing returns. Speed just doesn’t matter as much. And if you are making a GPU in this moment in time, why wouldn’t you take this opportunity to do something bold and new?
Well, Nvidia is making that attempt. Real-time ray tracing is a potentially massive leap forward for visuals and development. The tech renders light closer to the way it works in the physical world, and the results are stunning. Other RTX tech is increasing game performance in creative ways that involve specialized compute cores. It is looking beyond the horsepower race.
AMD, however, is not. At least, not yet. Radeon VII does not have variable-rate shading that enables games to render the same frame at different resolutions. It does not have an equivalent to deep-learning super sampling that can reconstruct a frame more efficiently than traditional methods. And it does not have ray tracing.
Now, maybe AMD is betting that Nvidia is wrong and those technologies won’t matter. That’s completely fair. No one knows yet if they will (OK — variable-rate shading is probably going to matter). But this isn’t about which company is correct in the end.
It’s about Nvidia’s approach making more sense right now. Nvidia’s argument is that it’s time to try a bold leap forward while we are in something of a graphical plateau. And the Radeon VII is evidence of that.
I think it is good to see Radeon VII for compute and content creation.
It should be ~ on par with a low end 2 slot wide RTX2080 OC based on the performance numbers I see and the price is similar.
I thgink it is ~ dead in the watyer if only considered as a gaming only GPU.
I own both RX Vega 64 Liquid and RTX 2080 OC now.
Based on 3Dmark testing, the Nvidia GPU is about 30% faster than RX Vega 64 Liquid running Adrenalin 18.12.1 driver in DX12, and surprisingly runs only 3-5% faster in DX11 on my system.
The RTX2080 OC performs better in practice in DX11 games I looked at.
I am interested in the Radeon VII but I would like to see DX11 Crossfire Support for Radeon VII with the RX Vega 64 Liquid and some work to make sure you can run the cards in DX12 MultiGPU.
I have not tested DLSS or Ray Tracing on the RTX2080OC yet - I have been too busy. They are interesting technology but there is not much support yet.
A PCIe 4.0 and later PCIe5.0 interface may allow fast communnication between main primary GPU for rendering and secondary GPU for DLSS/AI/Raytracing calculations.
This may be interesting: GPUDirect and DirectGMA - direct GPU-GPU communication via RDMA - StreamHPC