AMD Navi 23 ‘NVIDIA Killer’ GPU Rumored to Support Hardware Ray Tracing, Coming Next Year

[EastCoastHandle]

The current way that PC's use RAM is a major bottleneck and a re-engineered architecture is needed to fix the mess. Now that the mass market is using mesh interconnected CPU's and other interconnected components are coming - the RAM is becoming more and more of a bottleneck.

I do wonder if by the time we get to Zen 5/6 we'll see AMD try to interconnect system memory straight into the die - losing the upgradability but gaining low latency and very high bandwidth by using GDDR/HBM.

Very possible but a last, very late IMO. Intel did a lot of harm in R

Our present understanding is that AMD is working with at least one memory supplier to establish an on-site R&D lab at the company’s Austin campus, which is being used to research DDR5 and develop the next generation of memory. AMD is trying to beat Intel to market with DDR5, which fits the company’s history of pushing new memory standards earlier. ATi also did this with older revisions of GDDR memory.

Separately, we also learned that AMD is working on what they call “Near Memory,” or HBM being used in conjunction with future CPU components. We’re not clear presently on whether that includes desktop CPUs, but we do know that HBM for CPUs is in active R&D, and given Hades Canyon, that’s not necessarily a big surprise.

https://www.gamersnexus.net/news-pc/3286-hw-news-amd-going-ddr5-hbm-for-cpus-7nm-challenges

The gist of what is rumored is that AMD is working on something that would replace main memory/ddr(4) as we know it (be it ddr4 or ddr5 by the time it's released). It would be some sort of HBM 16GB (stacked) soldered on the motherboard. With DDR as secondary/backup memory.

Although I've not read anything more about it in a while now. But I'm reminded of it with next gen Consoles. Although I didn't read specifically it called "Near Memory" on console...I will keep an keen eye on Sony's press release in a few more hours to see what they are offering though.
(there were other articles about it but I can't find it at the moment).

In the meantime I really don't know how next gen console games are going to be properly ported over to PC unless a lot of time/resources are put into it this time around. Unless there is implementation from the start for PC.

Be that as it may I do hope AMD has something more in mind for Navi users then simply upgrade to the 6000 series. It's becoming real clear that next gen consoles are now a real option for a lot of folk out there.

 

[GAC]

so how exactly is ddr4 holding the pc back in gaming ?

taking the mincreaft demo for example a pc can run it faster than the new xbox, yes this is down to the gpu, but again the way you guys are talking is the pc is crippled compared to it memory wise, dont forget the fastest memory in the xbox is ONLY for the GPU, the system memory while faster then ddr4 is, only amounts to some 6 gigs split 3.5 for games and 2.5 for os/gui.
any pc with a dedicated gpu has fast gddr memory in it some faster than others my gtx1070 for example has at least (nvidias standard spec) 256gb's of bandwidth using 8 gigs of gddr5, a rtx2080ti using gddr6 has 616gb's and thats a 18 month old card, the xbox has 560gbs for its graphical ram and 336gbs for system ram, so yes on the system it pounds the pc but against a 18 month old gpu (yes it was and still is a bloody expensive one) it gets beaten, so again how is ddr4 holding the pc back today ?

now as for down the line sure, having more system memory will come in very handy BUUUUUUUUUUUUUT today say you are running ddr4 3200 (which to be blunt the vast majority of people arnt still) you'l be getting roughly 25.6 GB/s so yes again the console blasts the pc in to the weeds, but again the majority of people with a good ssd will be looking at read speeds of 400-500 megabytes, so no chance of the system memory being saturated, and even if you're rocking the latest and greatest pcie gen 4 drives you're still only going to be hitting 5gigs which again doesn't saturate the system memory, which interestingly on the xbox the throughput will be slower at 2.4GB/s (Raw), 4.8GB/s (Compressed) for the I/O for storage.

the only reason the whole machine has super fast memory i suspect is down to architecture and it was easier to have one type of system memory and one controller rather than having to have two of everything taking up more room on the system board.

 

[LoadsaMoney]

Are you auditioning for the job of train driver?

You're right - it's not here, yet. All claims are unverified.

Plus we don't know if it's an apples-to-apples comparison.

Ultimately for me, I just don't expect AMD with a smaller R&D budget to suddenly leap-frog nV. Hats off to them if they do, but for now I'm not considering it likely.

And I have no bias against consoles; I'm still planning to pick up a next-gen offering (probably PS5).

Xbox Series X could only run 1080p 30fps in minecraft ray tracing yesterday
2080ti does 60fps capped while only using 80% of it's core, so more in the tank spare

Keep telling us how the xbox is faster than the 2080ti, im waiting

Please don't take my posts seriously

 

[FoxEye]

Very possible but a last, very late IMO. Intel did a lot of harm in R


https://www.gamersnexus.net/news-pc/3286-hw-news-amd-going-ddr5-hbm-for-cpus-7nm-challenges

The gist of what is rumored is that AMD is working on something that would replace main memory/ddr(4) as we know it (be it ddr4 or ddr5 by the time it's released). It would be some sort of HBM 16GB (stacked) soldered on the motherboard. With DDR as secondary/backup memory.

Although I've not read anything more about it in a while now. But I'm reminded of it with next gen Consoles. Although I didn't read specifically it called "Near Memory" on console...I will keep an keen eye on Sony's press release in a few more hours to see what they are offering though.
(there were other articles about it but I can't find it at the moment).

In the meantime I really don't know how next gen console games are going to be properly ported over to PC unless a lot of time/resources are put into it this time around. Unless there is implementation from the start for PC.

Be that as it may I do hope AMD has something more in mind for Navi users then simply upgrade to the 6000 series. It's becoming real clear that next gen consoles are now a real option for a lot of folk out there.

Shame we still don't have Storage Class Memory.

That seems to be another tech (or rather techs) that's always "just around the corner", whenever you read about it.

But seems to be just as unproven in 2020 as it was years ago.

 

[EastCoastHandle]

so how exactly is ddr4 holding the pc back in gaming ?

The conventional approach of moving data to the CPU for computation has become a significant performance bottleneck for emerging scale-out data-intensive applications due to their limited data reuse. At the same time, the advancement in 3D integration technologies has made the decade-old concept of coupling compute units close to the memory --- called near-memory computing (NMC) --- more viable. Processing right at the "home" of data can significantly diminish the data movement problem of data-intensive applications.

https://arxiv.org/abs/1908.02640

As gaming developers move to next gen consoles and become more creative, less restrictive and opportunistic in their game engines it's very likely games will become more bandwidth sensitive.

For example, 1st person shooter games have typically been on "rails" with a lot of developer "tricks" to make it look open world. Once this becomes mainstream it will be open world.
Watch this video that Sony Demo'd

No 'pop in' no HD thrashing, no pausing, stutter, hiccups and no break in the immersion. A good portion of an open world made available without streaming textures, etc. I'm telling you this is just the tip of the iceberg, IMO. I'm only showing you just 1 example. Imagine being able to use Adobe on console now? And render quicker then on some PC's (I'm guessing here). The application use are only limited to one's own imagination.

But the gist is that games on this generation on console are going to provide a better experience then on PC. Because PCMR forgot to "innovate" and to bring technology forward even though we had no need for it at the time (not providing it this was never an acceptable excuse). Think about it for a minute. Intel hasn't even intro PCIe4 yet as mainstream!!!!!!!! Think of how far behind we are yet paying over $1k just for video cards and cpus that have never been worth the asking price.

I'm sure Intel will "make a splash" in the PC space however if things don't get better for them in the near future Intel can pull an "IBM" and still be profitable.

But let see what Sony show/demo today though.

However, I believe that as this tech matures and PCMR still doesn't innovate, today, it's going to be prohibitively costly to backport games to PC, IMO. Regardless of how "fast" and robust the GPU is.

 

[GAC]

you keep saying a better experience, explain HOW. and try to be exact because claiming pop in and no hdd thrashing means sod all when most pc games are there now with systems running ssd's and lower grade gpu's

will the xbox be a nice system and possibly the most powerful and best console of all time, a very good chance. will it change how pc games work, probably not very much unless you know things no one else does outside of ms ?

 

[gpuerrilla]

Think about it for a minute. Intel hasn't even intro PCIe4 yet as mainstream!!!!!!!! Think of how far behind we are yet paying over $1k just for video cards and cpus that have never been worth the asking price.

PCIe6 has been specced out already and is on course to be done for next year.

 

[EastCoastHandle]

PCIe6 has been specced out already and is on course to be done for next year.

Ah yes, it should deliver 256 GB's through x16. And I'm all for it. But by the time this comes out we PCMR will already be lagging consoles. Will PCIe 6.0 PC components (storage, video, etc) also be mainstream by then at a decent cost? Who knows. But from what I'm seeing we only have the 'potential' while they (console) will have "actual" benefits.

Here is one example in potential (specific bandwidth benchmarks) vs actual (in game benefits) going from PCie 3.0 vs PCIe 4.0. I can see why they are now on PCIe 6.0. But lets hope by then the landscape for PCMR is a bit more competent for gaming.

 

[GAC]

so thats a no on my question then ?

 

[CAT-THE-FIFTH]

so how exactly is ddr4 holding the pc back in gaming ?

taking the mincreaft demo for example a pc can run it faster than the new xbox, yes this is down to the gpu, but again the way you guys are talking is the pc is crippled compared to it memory wise, dont forget the fastest memory in the xbox is ONLY for the GPU, the system memory while faster then ddr4 is, only amounts to some 6 gigs split 3.5 for games and 2.5 for os/gui.
any pc with a dedicated gpu has fast gddr memory in it some faster than others my gtx1070 for example has at least (nvidias standard spec) 256gb's of bandwidth using 8 gigs of gddr5, a rtx2080ti using gddr6 has 616gb's and thats a 18 month old card, the xbox has 560gbs for its graphical ram and 336gbs for system ram, so yes on the system it pounds the pc but against a 18 month old gpu (yes it was and still is a bloody expensive one) it gets beaten, so again how is ddr4 holding the pc back today ?

now as for down the line sure, having more system memory will come in very handy BUUUUUUUUUUUUUT today say you are running ddr4 3200 (which to be blunt the vast majority of people arnt still) you'l be getting roughly 25.6 GB/s so yes again the console blasts the pc in to the weeds, but again the majority of people with a good ssd will be looking at read speeds of 400-500 megabytes, so no chance of the system memory being saturated, and even if you're rocking the latest and greatest pcie gen 4 drives you're still only going to be hitting 5gigs which again doesn't saturate the system memory, which interestingly on the xbox the throughput will be slower at 2.4GB/s (Raw), 4.8GB/s (Compressed) for the I/O for storage.

the only reason the whole machine has super fast memory i suspect is down to architecture and it was easier to have one type of system memory and one controller rather than having to have two of everything taking up more room on the system board.

The slowest RAM reserved for the system is 336GB/s(which is a few times more than DDR4 can achieve) as opposed to the GPU which has 560GB/s and it was mentioned the developers can use either pool as they see fit,which means they can move chunks of data into and out of main memory very quickly unlike in PC,so the lower amount of RAM is probably not as a big a concern as it appears. DF also said certain things about the SSD controller. The SOC has specific blocks for the SSD bandwidth decompression which hardly impacts CPU performance,and MS said if they hand't implemented this,the decompression itself would takeup a few CPU cores on itself. Hence,MS have implied the SSD can be used to hold certain information on the fly which can be quickly streamed back - apparently this is much harder to do with PCs due to the amount of constant CPU overhead. This is all said in the DF video and article.

With more and more cores on desktop CPUs,DDR4 is going to run out of bandwidth.

This is from a 2016 review Techspot/Hardware Unboxed did with an overclocked Core i7 6700K with two GTX980 GPUs in SLI.

This is only with SLI GTX980 GPUs with a quad core CPU at 1440p,not 720p.

The RAM cartel for years has put a huge premium on faster RAM,so very few gamers will own 4000MHZ DDR4 or good enough quality RAM which can overclock that much with not crap timings,when higher speed system RAM would help.

 

[CAT-THE-FIFTH]

In terms of how the memory is allocated, games get a total of 13.5GB in total, which encompasses all 10GB of GPU optimal memory and 3.5GB of standard memory. This leaves 2.5GB of GDDR6 memory from the slower pool for the operating system and the front-end shell. From Microsoft's perspective, it is still a unified memory system, even if performance can vary. "In conversations with developers, it's typically easy for games to more than fill up their standard memory quota with CPU, audio data, stack data, and executable data, script data, and developers like such a trade-off when it gives them more potential bandwidth," says Goossen.

The form factor is cute, the 2.4GB/s of guaranteed throughput is impressive, but it's the software APIs and custom hardware built into the SoC that deliver what Microsoft believes to be a revolution - a new way of using storage to augment memory (an area where no platform holder will be able to deliver a more traditional generational leap). The idea, in basic terms at least, is pretty straightforward - the game package that sits on storage essentially becomes extended memory, allowing 100GB of game assets stored on the SSD to be instantly accessible by the developer. It's a system that Microsoft calls the Velocity Architecture and the SSD itself is just one part of the system.

"Our second component is a high-speed hardware decompression block that can deliver over 6GB/s," reveals Andrew Goossen. "This is a dedicated silicon block that offloads decompression work from the CPU and is matched to the SSD so that decompression is never a bottleneck. The decompression hardware supports Zlib for general data and a new compression [system] called BCPack that is tailored to the GPU textures that typically comprise the vast majority of a game's package size."
The final component in the triumvirate is an extension to DirectX - DirectStorage - a necessary upgrade bearing in mind that existing file I/O protocols are knocking on for 30 years old, and in their current form would require two Zen CPU cores simply to cover the overhead, which DirectStorage reduces to just one tenth of single core.

"Plus it has other benefits," enthuses Andrew Goossen. "It's less latent and it saves a ton of CPU. With the best competitive solution, we found doing decompression software to match the SSD rate would have consumed three Zen 2 CPU cores. When you add in the IO CPU overhead, that's another two cores. So the resulting workload would have completely consumed five Zen 2 CPU cores when now it only takes a tenth of a CPU core. So in other words, to equal the performance of a Series X at its full IO rate, you would need to build a PC with 13 Zen 2 cores. That's seven cores dedicated for the game: one for Windows and shell and five for the IO and decompression overhead."

I wish people actually read the DF article fully or watched the video they made on it. The SSD itself with decompression can deliver over 6GB/s.

The PS5 specifications have also been released:
https://www.eurogamer.net/articles/...s-and-tech-that-deliver-sonys-next-gen-vision

It uses a custom 12 channel SSD,which delivers upto 8~9GB/s bandwidth after decompression.

Again like the Xbox it uses custom I/O hardware:

A dedicated DMA controller (equivalent to one or two Zen 2 cores in performance terms) directs data to where it needs to be, while two dedicated, custom processors handle I/O and memory mapping. On top of that, coherency engines operate as housekeepers of sorts.

 

[GAC]

snip

hmm ok, seems like faster ram could be a bonus would be interesting to see how a 2080ti in sli does with a 3950 with a 4000mhz or faster ram scales. also costs as would be interesting to see what 16 gigs of gddr6 is compared to a fast 16 gigs kit.

but again the way eastcoasthandle was going on you'd think pc's wouldnt be able to play new tripple a games in a year or two.

 

[Calin Banc]

The slowest RAM reserved for the system is 336GB/s(which is a few times more than DDR4 can achieve) as opposed to the GPU which has 560GB/s and it was mentioned the developers can use either pool as they see fit,which means they can move chunks of data into and out of main memory very quickly unlike in PC,so the lower amount of RAM is probably not as a big a concern as it appears. DF also said certain things about the SSD controller. The SOC has specific blocks for the SSD bandwidth decompression which hardly impacts CPU performance,and MS said if they hand't implemented this,the decompression itself would takeup a few CPU cores on itself. Hence,MS have implied the SSD can be used to hold certain information on the fly which can be quickly streamed back - apparently this is much harder to do with PCs due to the amount of constant CPU overhead. This is all said in the DF video and article.

With more and more cores on desktop CPUs,DDR4 is going to run out of bandwidth.

This is from a 2016 review Techspot/Hardware Unboxed did with an overclocked Core i7 6700K with two GTX980 GPUs in SLI.

This is only with SLI GTX980 GPUs with a quad core CPU at 1440p,not 720p.

The RAM cartel for years has put a huge premium on faster RAM,so very few gamers will own 4000MHZ DDR4 or good enough quality RAM which can overclock that much with not crap timings,when higher speed system RAM would help.

ArmA 3 and Fallout 4 are some of the worst games out there in terms of performance. I wouldn't give bad engines examples for what something does.

 

[CAT-THE-FIFTH]

ArmA 3 and Fallout 4 are some of the worst games out there in terms of performance. I wouldn't give bad engines examples for what something does.

But they are both openworld games,and so is Witcher 3. It's those kind of games which are more receptive to faster RAM and faster storage due to their scale.The fact is these games have one or two primary threads,ie,such as rendering threads etc which get hammered,so more memory bandwidth is useful. They also prefer SSDs over HDD,again due to the amount of data which needs to be loaded into the memory.

Also remember,this is an Intel test set-up - Ryzen likes faster RAM even more.

hmm ok, seems like faster ram could be a bonus would be interesting to see how a 2080ti in sli does with a 3950 with a 4000mhz or faster ram scales. also costs as would be interesting to see what 16 gigs of gddr6 is compared to a fast 16 gigs kit.

but again the way eastcoasthandle was going on you'd think pc's wouldnt be able to play new tripple a games in a year or two.

The problem here is the RAM cartel who held back adoption of higher speed RAM for years - look when DDR4 was released for desktop,and only now after over 5 years is 3200MHZ~3600MHZ DDR4 RAM is affordable and even then timings can be crap too. Remember,Haswell E was out 5 years ago,so that is how long its taken!

 

[Poneros]

It doesn't matter how fast an SSD is, only how fast it needs to be. It's not so easy to make use of faster storage across the board. We'll have to see if even open world games can fully make use of these new hyper-fast SSDs.

GPUs? Faster is always better, and can always be put to use. No ifs or buts.

Personally, I'm more curious to see if an Optane will finally be properly useable in games going forward. So much speed there, but kept back by archaic software.

 

[GAC]

have to see what amd has up its sleeve for ryzen when it finally moves on from am4, maybe another total step change again. but they would have to have memory makers behind them.

 

[Poneros]

2021

Zen 4 + DDR5 (+Pcie 5.0 maybe?)

Can't wait!

 

[GAC]

depending on costs who knows, they may just say sod it and go gddr5 or 6.

 

[BigBANGtheory]

I'm sure Intel will "make a splash" in the PC space however if things don't get better for them in the near future Intel can pull an "IBM" and still be profitable.

But let see what Sony show/demo today though.

IBM is not in a healthy place, if their machine learning and quantum compute tech doesn’t deliver they will become the next Atari

 

[gpuerrilla]

Another powerhouse that didnt get their strategy right. Proves if your not consistently pushing the boundaries you can easily get left behind.