Wednesday, October 28th 2020
AMD Radeon RX 6000 "Big Navi" RDNA2 Graphics Card Launch Liveblog
After thoroughly appetizing us with its "Where Gaming Begins: Episode 1" event announcing the Ryzen 5000 series "Zen 3" processors that offer up to 19% IPC gains, in the second Episode, we see the company announcing its next-generation Radeon RX 6000 "Big Navi" graphics cards based on the RDNA2 graphics architecture that introduce full DirectX 12 Ultimate readiness, including real-time raytracing hardware. In the run up to the RX 6000, NVIDIA is already reportedly preparing product-stack updates. In this liveblog, we uncover what has NVIDIA riled up, and whether AMD can pull off better pricing and availbility than the RTX 30-series.
Update 15:59 UTC: It is time! Welcome to the Radeon RX 6000 Series live blog.
Update 16:01 UTC: AMD CEO Dr Lisa Su takes centerstage, fresh off a good quarterly results announcement, and that big Xilinx acquisition announcement.
Update 16:03 UTC: FarCry 6 seems like an AMD optimized title.
Update 16:03 UTC: Possibly the flagship product.
Update 16:04 UTC: AMD is where gaming begins because next-gen consoles trust it. - Dr Su
Update 16:05 UTC: 50% generational improvement in perf/W.
Update 16:06 UTC: 26.7 billion transistors, almost as big as GA102
Update 16:07 UTC: RDNA2 has a breakthrough high-speed design.
Update 16:08 UTC: AMD RDNA2 compute unit 30% more energy efficient.
Update 16:09 UTC: InfinityCache works to significantly improve memory bandwidth, based on the Zen L3 cache. More than 2.17x bandwidth gain despite 256-bit.
Update 16:10 UTC: 30% higher frequencies on the same 7 nm node.
Update 16:10 UTC: DirectX 12 Ultimate and DirectStorage support.
Update 16:11 UTC: Over 2X performance gain over RX 5700 XT.
Update 16:11 UTC: The RX 6800 XT!
Update 16:12 UTC: 4K Gaming from AMD is here! Matches RTX 3080!!!
Update 16:13 UTC: Hello competition!
Update 16:14 UTC: Welcome back ATI Rage, as the AMD Rage Mode.
Update 16:15 UTC: When paired with Ryzen, you get a gaming performance boost, 13% perf increase.
Update 16:15 UTC: AMD introduces its take on the NVIDIA Reflex, announcing latest Radeon Lag and Radeon Boost. No special API needed.
Update 16:16 UTC: Full DX12 Ultimate support, including ray-tracing. Woirking on a DLSS-rival.
Update 16:18 UTC: AMD is leveraging studios working on Xbox Series X / PS5 devs to integrate its Radeon features on the PC.
Update 16:18 UTC: FarCry 6, DiRT 5, WoW Shadowlands (which gets raytracing), RiftBreaker, are AMD-optimized.
Update 16:21 UTC: Available Nov 18, starts, $649
Update 16:21 UTC: Also announcing RX 6800 : faster than RTX 2080 Ti, 4K + raytracing ready
Update 16:22 UTC: $579, November 18 availability for the RX 6800.
Update 16:22 UTC: Lisa gets ready for a "one more thing"
Update 16:23 UTC: Radeon RX 6900 XT: 80 CUs, 65% perf/Watt over 5700 XT.
Update 16:25 UTC: RTX 3090 BEATEN!!!!
Update 16:26 UTC: December 8 for the RX 6900 XT, priced $999.
Update 16:26 UTC: OOF, we cannot wait to test these cards out!
Update 17:08 UTC: The press-release can be found here.
Update 17:19 UTC: The complete AMD slide deck follows.
Update 16:03 UTC: Possibly the flagship product.
Update 16:05 UTC: 50% generational improvement in perf/W.
Update 16:15 UTC: AMD introduces its take on the NVIDIA Reflex, announcing latest Radeon Lag and Radeon Boost. No special API needed.
Update 16:18 UTC: FarCry 6, DiRT 5, WoW Shadowlands (which gets raytracing), RiftBreaker, are AMD-optimized.
Update 16:21 UTC: Available Nov 18, starts, $649
Update 16:21 UTC: Also announcing RX 6800 : faster than RTX 2080 Ti, 4K + raytracing ready
Update 16:22 UTC: Lisa gets ready for a "one more thing"
Update 16:23 UTC: Radeon RX 6900 XT: 80 CUs, 65% perf/Watt over 5700 XT.
Update 17:08 UTC: The press-release can be found here.
Update 17:19 UTC: The complete AMD slide deck follows.
351 Comments on AMD Radeon RX 6000 "Big Navi" RDNA2 Graphics Card Launch Liveblog
They probably have contracts for production.
They are trying to sell out of what inventory they have.
I mean ... they released water cooled Vega with 375 W TBP, why would they hold back now, especially when the competition went crazy with the power consumption?
And then there is this:
Maybe 30% frequency increase over RDNA was for Navi 22, maybe we misunderstood how they calculated this, or maybe they have another ace up their sleeves (maybe Navi RAGE Edition?) ... ;)
The question is, when NV release it, when we can have it in the market shelves? 3080 and 3090, where are live are scheduled for January btw. 3070 at the end of November.
If AMD delivers November 18th, it is a huge win for AMD.
Although those graphs was with the 6800 the SmartAccessMemory on. If you cut that off it may be just 15% faster or around that.The 6800 probaly is the 3070Ti competitor as its above the 2080Ti/3070
---------------------------------------------------------------
I can see there is a lot of confusion about the new feature AMD is calling "Smart Access Memory" and how it works. My 0.02 on the subject.
According to the presentation the SAM feature can be enabled only in 500series boards with a ZEN3 CPU installed. My assumption is that they use PCI-E 4.0 capabilities for this, but I'll get back to that.
The SAM feature has nothing to do with InfinityCache. IC is used to compensate the 256bit bandwithd between the GPU and VRAM. Thats it, end of story. And according to AMD this is equivalent of a 833bit bus. Again, this has nothing to do with SAM. IC is in the GPU and works for all systems the same way. They didnt say you have to do anything to "get it" to work. If it works with the same effectiveness with all games we will have to see.
Smart Access Memory
They use SAM to have CPU access to VRAM and probably speed up things a little on the CPU side. Thats it. They said it in the presentation, and they showed it also...
And they probably can get this done because of PCI-E 4.0 speed capability. If true thats why no 400series support.
They also said that this feature may be better in future than it is today, once game developers optimize their games for it.
I think AMD just made PCI-E 4.0 (on their own platform) more relevant than it was until now!
Full CPU access to GPU memory:
-------------------------------------------------------------------------------
My opinion on the 6000series is mixed.
They probably dont have as good implementation of RTRT and DLSS equivalent than nVIdia has, probably a 2080Ti perf for RTRT at least and cant comment for the Fidelity thing because I dont know.
But that is not why the mixed taste the 6000series had left me. RTRT is the future yes but today or the next year is still too restricted and not so important to most. For some is. For variable render resolution we will also see how it goes.
The 6800XT seem to me the only card priced well at 649$ according to rasterization performance against the 3080. With no rage mode or SAM. Given the fact that with those on it may beat the 3080 depending the OC headroom and the 3080's inability to clock more than marginal it makes 6800XT more appealing.
The 6900XT is not a 3090 rival the way 6800XT is for 3080. Yes it had the about the same chart differences with the below pair (6800XT vs 3080) but only with rage mode and SAM on. Which means that without them, out of the box on 300W will be probably half way between 3080 and 3090. Really tight gap but still. That makes it about 5~6% better than 6800XT? more or less. For +54% price? Yeah...
Probably AIBs may exploit more out of the 6900XT GPU and match the 3090 or even pass it, but I hardly believe that those will cost less than 1100~1200$.
The 6900XT will probably be a 3080Ti competitor on perf and price(?).
I was expecting a little more distinguishable products at the top end from AMD and not follow the nVidia stupidity. Why AMD didnt give +10% power draw headroom for the "FE" to match or even pass the 3090... I dont know. Maybe its the cooler design that max out at 300W. As I said AIBs (if will exist for 6900XT) probably will do that, but not for 999$.
The 6800 at 579$... a little steep... It will be more clearer when the 3070Ti is out, but comparing it to the 6800XT and its perf difference its should've been at 549$ max. That would make more sense.
+10% more price for +15% more performance of the 3070nonTi and a nice 3070Ti rival which I see the 600$ price tag on it.
I guess reviews will clear things better, but I think wont make any significant differences on my initial thoughts.
Edit(s):
typo(s)
As for TSMC, their contracts are most likely the same, you place an order for X amount of capacity and you can make whatever products you've taped up with them, using up that allocation.
The key thing here is that you pay a fee to tape out your products, it's not little money, something small like a USB 3.0 device controller is in the region of US$500,000 on a not cutting edge node. This is one of the big costs when it comes to making an IC, so ideally you don't want more than one tape out, but we know that this isn't always the case. Once taped out on the fabs node, you can always go back and make more of the same thing and as we know, over time the fab tends to improve both yields and the quality of the chips they make.
So yes, as long as it's on a node that AMD has a contract to manufacture on, they can switch in whatever product they want. It would obviously take a bit of time to do this, a week or two I would say, as this isn't quite like baking cookies. I mean, just remember the issues that happened not to long ago for Toshiba/WD where a 13- minute power outage wasted 6-9 exabytes of flash and you realise how sensitive these fabs are. I have been to a few fabs over the years and although it's not the most exciting thing in the world, they are really quite amazing considering the tolerances and cleanliness that's required to make chips.
The short answer is, yes, AMD can swap in and manufacture whatever they want, as long as they have a contract to manufacture on said node and have enough capacity left.
As for your second question, there are different types of supply contracts. In most cases there's an EOL notice, you must've seen some of the ones Intel send out to notify it's customers that a certain product won't be produced past a certain date and this and this date is the final order and shipment dates. The other is like when AMD makes the console SoCs where they have a contract to supply X amount of chips per months for a certain period of time. I very much doubt AMD has such a contract with any of the graphics card makers. Quite often there's an allocation based per what each partner is planning on manufacturing for the next six months or so. I doubt many of their partners would still be interested in an older product now, so there isn't going to be any issues for AMD. Normally there's also a certain amount available in rolling stock, something you can quite easily see if you go to someone like NXP or Microchip and try to order some parts. Popular products are available in stock, whereas more niche products you might have to place an order for and then there's a lead time of 12-16 weeks at a minimum as an example.The super was "only" $499 at launch.
en.m.wikipedia.org/wiki/GeForce_20_seriesIt's apparently based on the Zen L3 cache design, so that might be tricky.
The RX 6800 XT looks like a good competitor for the 3080 (short listed) but I wonder how the AMD will perform on my monitor (Acer Predator XB1 XB271HU [G-sync]). Cause I don't want to buy a new one.
GPU market going to be most interesting since 2015 and will be interesting to see if AMD go for market share with a price war next year if they can produce mountains of these cards.
Sense makes it none.
Why would yet another GA102 based NV GPU be less of a clusterf*ck than 3080?It helps to remember that that "way" is 10% of perf.
And if you go with a Ryzen 5000 (most new buyers will) performance should be faster than the 3080.
There is a huge gap between 5700XT $399 and 6800 $579 and I doubt anything but perhaps mild price drop will happen here until low end 6000 series arrive.