Flappy Pannus
Veteran
Dude.
Predict: Next gen console tech (10th generation edition) [2028+]
- Thread starter RDGoodla
- Start date
Dude.
Artistry is great, but important details are completely whack. Look closely at the ports, vents, text, etc! It can't even draw straight traces or pins! This is where artificial 'intelligence' proves its lack of actual intelligence - it doesn't understand anything at all and can only approximate stuff people do and have done. It's an advanced copying algorithm. And we can spot it because we actually know what we're supposed to be looking at.
Johnny Awesome
Veteran
For now...
Yea i completely missed those details lol. I was only paying attention the great use of dof and the play on lights.
It is true that the performance gain of PC hardware has slowed down somewhat, but not nearly as much as some here think. Here is a benchmark comparing AMD 6700xt vs 7700xt:
AMD Radeon RX 7700 XT Review Photo Gallery - TechSpot
View 47 detailed pictures from our AMD Radeon RX 7700 XT Review, close-up photos in high resolution.www.techspot.com
A 27% compute increase can therefore be realized. If we start from this, then:
1(current console gen)x1.27x1.27x1.27x1.27= 3.30x (next console gen if 2026 debut)
obsession
1(current console gen)x1.27x1.27x1.27x1.27x1.27= 4.19x (next console gen if 2028 debut)
It is clearly visible that a 3-4x basic raster performance increase is expected. Of course, other technological capabilities are added to this. Thus, the performance of the next gen consoles is expected to be significantly higher than the Nvidia 4900 VGA, even when the console debuts in 2026!
It also depends on the architecture of the graphics chip. Ms could get a larger multiplier if AMD changes RDNA enough (or leaves rdna for something else) that offers a higher performance in ray tracing. Ray tracing seems to be the way foward with all the games.
RDNA2/3 are equally crappy at it. So RDNA 4 or 5 could see significant gains esp with ray tracing enabled
On the cpu side zen 4 is clock for clock much more performant than zen 2 and so games like Dragons Dogma 2 or balders gate 3 that are heavily cpu limited can see massive gains on a console built around zen 4/5 or even 6 depending on when it hits. The new zen chips are rumored to have xillian (or whatever that company amd bought is called) ai cores. That could actually bring about a huge change in game play if the cores could in real time be used to create new content in the games.
Frenetic Pony
Veteran
We shouldn't discount "the largest technical leap you have ever seen.", Xbox 360 NoScope could have "10x the raytracing power and 20x the AI power!!!" Let's not discount public relations here, these people are professionals.
I think there are a lot of ways to make what MS is saying accurate.
1) New zen - AMD is readying zen 5 but MS could go with the rumored zen2/zen5 hybrid big little. They could also go with a zen 5 with ai cores. This could solve significant cpu bottlenecks in the console space Even going with a 8core zen5 with a single quad core zen 2 core set up would allow them to completely off load the OS to the zen 2 and leave the 8 core zen 5 for game code to run and other tasks.
2) new generation of rdna. RDNA 2-3 was a poor increase in performance. RDNA 4 /5 could be a much bigger leap for AMD esp in terms of ray tracing which could give MS a larger advantage ray tracing games which are becoming more and more common as the generation is going along. Even getting to a 4070 ray tracing performance would be drastic compared to the series x and ps5 (even the ps5 pro would be behind)
3) More ram / Faster ram . MS could adopt an exotic ram type. like HBM or go with an infinity cache for massive amounts of bandwidth. They can also go with a split memory pool again like the old days. Go dd5 ram for the cpu and gddr for the gpu. MS has an exotic ram design this generation.
4) Speaking of splitting. MS could go back to a separate cpu/gpu instead of an apu. This would allow them to go with an zen 3d cache part with large amounts of cache and also infinity cache on the gpu which could elevate bandwidth issues
5) They could simply go with a larger console allowing them to cool a higher wattage .
Of course it can be a mixture of any of these things. I think MS having the a big/little console set up also can work into this for them. They can release an xbox series x performance level new console using the new technology and go with a larger console at $500.
Remember when looking at the ps4 pro and the xbox one x , MS was much more agressive. I pointed out that MS could have a great opportunity if Sony goes with a more ps4 pro type upgrade instead of taking advantage of newer technology. We see they stayed with zen 2 with just a higher clock. The gpu is still a bit unknown but Ms could have another 360 type moment by launching a new console and shedding the older console.
But hey we have to wait to see. If they are announcing new hardware we will hear about it in June or maybe at the game awards in the fall like they did with the one x
The next Xbox will be a PC with an Xbox storefront and an optional Windows operating system, with minimally optimized PC games, i.e. a single software version. The advantage of this model is that its pricing will be more moderate compared to pre-assembled PCs, and it will be delivered to customers ready to launch for games.
With this model, the significant jump in performance that was talked about is feasible.
With this model, the significant jump in performance that was talked about is feasible.
Last edited:
Think that "largest technical leap" is going to be the easiest to fudge the numbers with and that's TOPS. Imagine all the marketing with 1000+ TOPS, people have no idea what that actually means, "100x in a single gen", "advanced AI", "AI console" etc plastered everywhere. It's the "easiest" and cheapest thing to improve significantly, builds on current hype and trends and so is easy to sell to superiors because of it. Could even go the FP8 angle, from non-existent to 1000TOPS, undefined increase
82 CUs, 4096 ops/cycle for FP8/Int8 (same as CDNA3), 3GHz = 1007 TOPS. Whether that happens is another thing but it's one of the more likely
82 CUs, 4096 ops/cycle for FP8/Int8 (same as CDNA3), 3GHz = 1007 TOPS. Whether that happens is another thing but it's one of the more likely
That'd be 1 POPS- the first ever console with an added grandparent.
Could be even more simple than that.
12TF to 48TF. It's only a 4x increase and most of that is because of dual issue, but it's also technically a 36TF increase, which could be sold as the biggest ever leap.
But your scenario is quite plausible as well. Especially if they go even lower precision to something like INT4.
Frenetic Pony
Veteran
I forgot about the "double TF" thing, that's definitely gonna come up in PR releases.
Also "int1". Actually scratch that, at this rate the "AI" market is going to reach "fp0" and "zerodivisionerror" flops all by themselves. Trust us bro our startup has used quantum mechanics to invent time travel specifically for machine learning.
Hmm, single bit ops counted individually. A 16 bit ADD becomes 16 OPs. Heck, let's count the carries as an op too. 0111111111111111 + 0000000000000001 becomes 31 ops, 16 bits and 15 carries.
Last edited:
People are not that gullbible though. With comparisons mvailable witha few clicks people can see the difference. it's how DLSS is sold mostly, not by numbers but because people are comparing Nvidia DLSS vs the others inferior solutions.
Most people have no idea (really!) what DLSS is, they just buy Nvidia VGA because it's more powerful and popular...
That's quite possibly not true for those buying discrete GPUs, especially the higher end. There'll certainly be a more informed populace than console buyers responding to console marketing. It's the kind of generalisation that hasn't any supporting data and can't be proven one way or the other.
In short, I think console buyers are more 'gullible' than PC GPU buyers, but I don't think anyone can prove me right or wrong. And it's a non technical discussion anyway- better to just focus on the numbers and how to justify/rationalise them.
In short, I think console buyers are more 'gullible' than PC GPU buyers, but I don't think anyone can prove me right or wrong. And it's a non technical discussion anyway- better to just focus on the numbers and how to justify/rationalise them.
Most GPUs sold are not at the high end. As to pc users being more gullible than console users? Yea I’m not sure about that… We have people purchasing GPUs at up to $2000 to play a game. It’s not generating an additional income for them, they’re not a professional, it’s just to play games. Then in 2 years, they’ll buy the next ridiculously priced top end GPU and repeat the cycle again. Yea let that sink in.
I will save this comment.
Not sure if it's gullible and more don't know any better or don't actually care that much if at all, they hear a cool number that everything and everyone else is repeating so it must be true and that's as far as it goes. The proportion of people who are actually into spec at more than surface level, how stuff works, want to understand it etc is very small
Frenetic Pony
Veteran
So supposedly RDNA4 offers a substantial performance per area gain architecture wise.
Which means I can suddenly see where it is Microsoft's "Giant Leap" is kinda sorta a bit coming from, which they've doubled down on in an internal email. While they had to have suspected an internal email could be leaked, it's also a direct communication to employees 100x more than any sort of PR stunt, they're really trying to push the narrative that it's going to be significant directly to employees.
Given RDNA5 is heavily chiplet based, and that they're already designing next gen hardware, that seems like "2026" is plausible even if "2028" is what the article says, 4 years to design a console is, well that's not unheard of that's like buying AMD to get a whole new GPU arch custom designed almost. 2026 gets the jump on Sony and 2028 though. So given the total assumption of "2026":
Xbox 1080: Handheld, 20CU RDNA5, 2ghz, "10.4"(5.2 more like) teraflop, 4 Zen 6c cores/4 Zen 6e cores, 64bit bus/24gbps GDDR7/18gb clamshell, 1tb UFS 5.0 5gbps, XDNA3 60 TOPS, 1080p 500nit VRR screen, $399. No "too little" ram complaints from Devs, run Series S games with proper settings (Frontiers of Pandora/Dragon's Dogma 2 with Series X/PS5 settings but @1080p).
Xbox 2160: Stationary, 120CU RDNA5 2.8ghz, 84(42) teraflop, 8 Zen 6 cores, 256bit bus/36gbps GDDR7/24gb, 2tb 10gbps NVME, dual XDNA3 chiplets @120 TOPS. Games on the 1080 that run at native 1080p/30 run at 4k/120 (frame gen of course). Can cloud stream @4k 120 HDR AV1 100mpbs to any compatible screen, controller uses Wifi direct to minimize controller latency to almost nothing inside the house even when streaming to other screens. $599. As fast as 5080(ti?) for $599, all the raytracing 120 whatever the hardcore audience could want at a reasonable price.
Which means I can suddenly see where it is Microsoft's "Giant Leap" is kinda sorta a bit coming from, which they've doubled down on in an internal email. While they had to have suspected an internal email could be leaked, it's also a direct communication to employees 100x more than any sort of PR stunt, they're really trying to push the narrative that it's going to be significant directly to employees.
Given RDNA5 is heavily chiplet based, and that they're already designing next gen hardware, that seems like "2026" is plausible even if "2028" is what the article says, 4 years to design a console is, well that's not unheard of that's like buying AMD to get a whole new GPU arch custom designed almost. 2026 gets the jump on Sony and 2028 though. So given the total assumption of "2026":
Xbox 1080: Handheld, 20CU RDNA5, 2ghz, "10.4"(5.2 more like) teraflop, 4 Zen 6c cores/4 Zen 6e cores, 64bit bus/24gbps GDDR7/18gb clamshell, 1tb UFS 5.0 5gbps, XDNA3 60 TOPS, 1080p 500nit VRR screen, $399. No "too little" ram complaints from Devs, run Series S games with proper settings (Frontiers of Pandora/Dragon's Dogma 2 with Series X/PS5 settings but @1080p).
Xbox 2160: Stationary, 120CU RDNA5 2.8ghz, 84(42) teraflop, 8 Zen 6 cores, 256bit bus/36gbps GDDR7/24gb, 2tb 10gbps NVME, dual XDNA3 chiplets @120 TOPS. Games on the 1080 that run at native 1080p/30 run at 4k/120 (frame gen of course). Can cloud stream @4k 120 HDR AV1 100mpbs to any compatible screen, controller uses Wifi direct to minimize controller latency to almost nothing inside the house even when streaming to other screens. $599. As fast as 5080(ti?) for $599, all the raytracing 120 whatever the hardcore audience could want at a reasonable price.
Last edited:
Similar threads
