Tuesday, January 14th 2020
RX 5950 XT, RX 5950, and RX 5800 XT: New AMD Radeon SKUs Reach Regulators
Confirmation of four new Radeon RX 5000-series SKUs came to light as AIB partner AFOX pushed them to regulators at the Eurasian Economic Commission. EEC filings have been a reliable early-sign of upcoming PC hardware. All thee new SKUs are positioned above the Radeon RX 5700 XT launched last year. These include the Radeon RX 5800 XT, the RX 5900 XT, the RX 5950, and the RX 5950 XT. Going by AMD's convention of two SKUs per resolution serving up to differentiated experiences, the RX 5800 XT could be a step up from the RX 5700 XT in offering 1440p high frame-rate AAA performance. This could possibly put it in direct competition with the GeForce RTX 2070 Super. AMD took a similar 2-pronged approach to 1080p, with the RX 5500 XT serving up 1080p at up to 60 fps, while the RX 5600 XT topping it up with a 40-50 percent performance uplift.
The Radeon RX 5950-series is completely new. This could very well be a new large "Navi" silicon, since dual-GPU is dead. Just as AMD carved out the RX 5700 XT, the RX 5700, and the RX 5600 XT, it could carve out the three new SKUs from this silicon. AMD CEO Dr Lisa Su already confirmed that her company is working to upscale the RX 5000-series "Navi" family. The RX 5900-series could be competition for the likes of the RTX 2080 or even RTX 2080 Super. The RX 5950-series could target premium 4K gaming (RTX 2080 Ti). It remains to be seen if the three new SKUs are based on the existing RDNA architecture or the new RDNA2 architecture designed for 7 nm EUV, featuring variable-rate shading.
Sources:
Eurasian Economic Commission, Komachi Ensaka
The Radeon RX 5950-series is completely new. This could very well be a new large "Navi" silicon, since dual-GPU is dead. Just as AMD carved out the RX 5700 XT, the RX 5700, and the RX 5600 XT, it could carve out the three new SKUs from this silicon. AMD CEO Dr Lisa Su already confirmed that her company is working to upscale the RX 5000-series "Navi" family. The RX 5900-series could be competition for the likes of the RTX 2080 or even RTX 2080 Super. The RX 5950-series could target premium 4K gaming (RTX 2080 Ti). It remains to be seen if the three new SKUs are based on the existing RDNA architecture or the new RDNA2 architecture designed for 7 nm EUV, featuring variable-rate shading.
105 Comments on RX 5950 XT, RX 5950, and RX 5800 XT: New AMD Radeon SKUs Reach Regulators
I really wonder how AMD going in with Ray Tracing. The wording on Xbox Series X launch has a weird framing, something along the line that the Ray Tracing is done externally... I forgot where I read or the exact sentence, but it's around there.
On CPUs, AMD caught an Intel sleeping for almost a decade, with marginal performance increases every year. As if that were not enough, they were too optimistic for their 10nm density and the shot backfired. And to make things even worse, at a time when AMD had access to a better manufacturing process. That is, without taking AMD's merit, but basically everything aligned to AMD almost perfectly, probably not even they expected everything to go that well.
In the GPU front, the story is very different. Not only has Nvidia significantly increased performance every year, introducing new features and always improving the efficiency of its products, unlike CPUs, the 7nm advantage does not apply. Nvidia is still better with the old 12nm and they will change to 7nm in a few months.
These are very different market situations, and unless all the stars line up for AMD and Nvidia starts to fail, I don't see how anything similar to CPUs could happen. I don't doubt that AMD could have something competitive, but doing just like what Ryzen have done on CPUs is pretty much a daydream.
Look at Nvidia cards and AMD equivalents, in most of them AMD not only has less performance, but is less efficient, despite already using 7nm.
Perhaps the only exception is the RX5700 non-XT.
I'm more of a believer we get +30% in the stack compared to the Supers and then a mild price bump too. That will still leave AMD gasping for air. Nvidia is going to want to keep their first 7nm iteration as small as possible, so they can follow up on it within the same architecture. AMD is approaching it the other way around now, with process before arch. Also, Nv needs die space for RT, so that's going to limit their raw perf die space too. You miss the point where Nvidia was still on an old, easy node and AMD is using valuable 7nm capacity. Nvidia can make that huge chip a tad more easily, and they can still scale down to repeat the trick again once 7nm is matured. This is how architecture efficiency translates into economics.
I do get what you're saying about the 5700 and same architecture, but in practice, we see every time AMD is 'forced' to clock the snot out of its line up to keep it competitive within the same SKU. And that kills efficiency and shows the flaws of GCN/RDNA. Bigger dies won't suffer that so much, but then... what's next? Its not like they have 5nm on the shelf so its going to have to come from RDNA2.
You're also taking for granted that Ampere will not improve upon Turing efficiency wise.... I beg to differ. Look at what Pascal did... that wasn't just a shrink, they used that shrink to push 25% higher clocks and maintain efficiency, due to much finer grained power control.
Also, 1500 for all you care; its not like the 2080ti's present a major market share... even Nvidia has trouble selling at that price; and they don't even want or need to, its still a hard chip to make. You don't just poop out big chips with good quality like that. Take special note of the bad batches of 2080ti we have seen.
You are not blind, are you. Nvidia has been pushing smaller dies for a looong time, their efficiency push has always translated into profit there. Smaller memory bus, smaller chip. They departed from that with Turing, but the timing is fantastic, because now they can time their 7nm more comfortably and get comfy with RT. They used Turing to buy time. AMD used the node to buy time to work on architecture... so again, RDNA2 had better deliver, and they had better have something after that too.
AMD can and will price their products above Nvidia's if the opportunity arises, just like they did with Ryzen.
And first to market... it matters. It mattered with Gsync; it matters because it enforces mindshare and the idea of leader not follower. It also reinforces the idea that as a customer you're not missing out. And that is a BIG one, despite what you may think. It translates into Nvidia being able to charge premium. Even something as silly as RT in Turing, while it has very little to show to us, what it did show was that with Nvidia you are again on the first row in tech progress.
So yes, Nvidia can, because they are not AMD. You may not agree with the rationale but that doesn't make it less of a reality.
An example to drive this home. Pascal was released in 2016. Here is Nvidia doing 2000 mhz clocked GPUs
Here is AMD in 2019 doing 2000 mhz clocked GPUs.
They wouldn't have been able to copy them in the past even if Nvidia would have handed them over all their IP, it wasn't possible. OK, enjoy being ripped off by one company in particular because they were first or something along those lines. AMD should go home and stop trying to do anything because there is no point.
RX 5800 series could have RX 5800 and RX 5800 XT.
And then you top it off by saying 7nm is not going to be an efficiency jump when AMD showed that it was all by itself with the Radeon VII. Dafuq?!
Start connecting the dots, mate. Also this really says it all. I can separate my personal buying decisions from the analysis of how companies perform and how its products work.
You apparently seem to have trouble doing that.