Something that could be rather exciting for AMD enthusiasts, AMD has officially revealed the AMD Radeon VII at CES 2019. On top of that, 3rd generation Ryzen desktop processors are coming.
Getting ahead of the curve a little here, the Radeon VII is built on 7nm which makes it the first consumer-level GPU to be built with it which is interesting. AMD say it's built on an "enhanced second-generation AMD ‘Vega’ architecture" and it seems it will be a decent boost over the current Radeon RX Vega 64.
When compared directly with the RX Vega 64, AMD said it performed up to 27% higher in Blender, up to 27% higher in DaVinci Resolve and they saw up to 62% higher performance in the OpenCL LuxMark compute benchmark.
Some more specs:
- 60 compute units
- 3840 stream processors running at up to 1.8GHz
- 16GB of HBM2 memory (second-generation High-Bandwidth Memory)
- 1 TB/s memory bandwidth
- 4,096-bit memory interface
When it comes to gaming, that was also mentioned as well of course. It's nice to see Vulkan mentioned along side DirectX too! Naturally, they're only going for big Windows games right now but they did say it offered "35 percent higher performance in Battlefield V, and up to 42 percent higher performance in Strange Brigade 1" over the Vega 64 which is quite impressive.
The Radeon VII will be available February 7, 2019 for around $699 USD.
Additionally, they've teamed up with Google to power Project Stream, Googles new cloud gaming service using their Radeon Pro GPUs.
On top of that, 3rd generation Ryzen desktop processors are coming. They will also be built on 7nm tech, based on the Zen 2 core architecture and AMD say it's the "world's first" to support PCIe 4.0 connectivity. Sounds like it's going to be a beast, as they did a preview of it against an Intel i9 9900k where the Ryzen processor came out on top while also using around 30% less power.
They're launching the AMD Ryzen 3000 series sometime in the middle of 2019.
For notebook/laptop users, they also revealed the 2nd Gen AMD Ryzen Mobile processor with Radeon Vega Graphics coming to a range of devices from companies like Acer, Asus, Dell, HP, Huawei, Lenovo and Samsung throughout 2019.
You can see their CES 2019 video here.
Article taken from GamingOnLinux.com.
Quoting: cRaZy-bisCuiT7 nm but only GTX 2080 performance? Why? For the same price?
Nice only because we got a open well performing driver on Linux for AMD. Still I'm not much impressed.
Because that's limitation of GCN architecture. Reduced process allows them to squeeze more, but the limiting factor now is GCN. For really major performance breaktrhough, you'll need their super SIMD, which is post GCN architecture. So another year or more for that. They are working on it, so don't worry.
Last edited by Shmerl on 9 January 2019 at 8:24 pm UTC
cRaZy-bisCuiT Jan 9, 2019
Quoting: ShmerlBecause that's limitation of GCN architecture. Reduced process allows them to squeeze more, but the limiting factor now is GCN. For really major performance breaktrhough, you'll need their super SIMD, which is post GCN architecture. So another year or more for that. They are working on it, so don't worry.Well, just a normal die shrink should yield in better performance usually. That's not a major step-up. I hoped for a new APU for Lenovo A495 in 2019. :/
Quoting: cRaZy-bisCuiTWell, just a normal die shrink should yield in better performance usually. That's not a major step-up. I hoped for a new APU for Lenovo A495 in 2019. :/
Depends on the architecture too. For example original Ryzen wasn't just a die shrink, it was a brand new architecture. That's why it performs so well. AMD didn't do that with GPUs yet. But it's in the pipeline.
Last edited by Shmerl on 9 January 2019 at 8:38 pm UTC
Quoting: cRaZy-bisCuiTI hoped for a new APU for Lenovo A495 in 2019. :/
That will happen, but using 12nm Ryzen 2 (Zen+) + Vega. Mobile AMD is always one year behind desktop one, so nothing unexpected here.
A495 seems like an overkill (using Ryzen Pro). E495 would be more like it with refreshed APU.
Last edited by Shmerl on 9 January 2019 at 8:43 pm UTC
Quoting: Shmerl16GB HBM2 at $700... I hope they'll have 8GB variants. So much VRAM might be good for 3D modeling, but I doubt gaming needs so much yet.
That's because the doubled the memory bus width up to 4096 for an impressive bandwidth of 1 TB/s.
I think this last value is relevant for performance.
Quoting: edmondoThat's because the doubled the memory bus width up to 4096 for an impressive bandwidth of 1 TB/s.
I think this last value is relevant for performance.
Such bandwidth is only needed because of all the RAM (parallel access). With less RAM they'll need less bandwidth.
Last edited by Shmerl on 9 January 2019 at 8:51 pm UTC
heidi.wenger Jan 9, 2019
Quoting: ShmerlQuoting: edmondoThat's because the doubled the memory bus width up to 4096 for an impressive bandwidth of 1 TB/s.
I think this last value is relevant for performance.
Such bandwidth is only needed because of all the RAM (parallel access). With less RAM they'll need less bandwidth.
It's probably more a question of balance between CUs, VRAM and ROPs (and power consumptions, and a lot more) to get the best performance. It seems to me that Vega II with 60 CUs is a more balanced design with double HBM2 memory bandwith than the first Vega, which even in its greatness it always felt to me like somewhere a little bottlenecked and never running at maximal potential.
And then for sure, marketing reasons like 16 GB is more than 11 GB ;)
Quoting: heidi.wengerWhat about that much talked about "Navi"? So this is not it at all?
Navi will come later. Most of us are surprised to see Vega II at all. It seems AMD has seen an opportunity in the market for it, probably strong 7nm process (and Nvidia skyrocketing prices).
cRaZy-bisCuiT Jan 9, 2019
Quoting: edmondoWhere do you see this opportunity? The price is about the same. Just because of vram?Quoting: heidi.wengerWhat about that much talked about "Navi"? So this is not it at all?
Navi will come later. Most of us are surprised to see Vega II at all. It seems AMD has seen an opportunity in the market for it, probably strong 7nm process (and Nvidia skyrocketing prices).
Patreon
PayPal
See more from me