AMD Ryzen 8000 CPUs likely to get same number of cores, but a lot faster
Click here to post a comment for AMD Ryzen 8000 CPUs likely to get same number of cores, but a lot faster on our message forum
schmidtbag
I too can see that kind of software coming soon, but that's more likely to be done by GPUs and AI processors rather than CPUs.
Agonist
This could be enough for me to ditch my 5800x3d setup.
CPC_RedDawn
Not really, look at Threadripper and Epyc CPU's. They can add more cores but for desktop/home use but there is no point at this time. It would just increase costs for consumers with, like you said, no real software to take advantage of this outside of prosumer or synthetic workloads.
4 cores in the 2010's was because Intel got greedy and had zero competition and their egotistical and selfish way of spinning their own fabs turned on them with 10nm which is why we kept getting 14nm++++++++++++++++++++++++++++++++++++
AMD does not advertise 6000MT DDR5
Max Memory Speed
2x1R
DDR5-5200
2x2R
DDR5-5200
4x1R
DDR5-3600
4x2R
DDR5-3600
This is the max that AMD recommend. As for 6000MT that is on the RAM and Motherboard makers.
https://www.amd.com/en/products/cpu/amd-ryzen-9-7950x
Ricepudding
Honestly this is fine, most consumers are on 6 or 8 cores right now, and most software is still on a few cores. Believe this is why on many programs we see no difference between 8 and 16 cores assuming you set the core clocks the same as its just not making use of it
cucaulay malkin
so was every other intel/amd refresh
wavetrex
Not even close.
Intel's newer 4 cores "upgrades" were not just the same number of cores as previous gen, but also only 5%.. maybe 10% faster than previous generation, IF LUCKY, with pretty much stagnation on power efficiency (+5% faster with +5% more power consumption... ewww)
Right now the number of cores doesn't seem to increase much, indeed, but each new generation is considerably faster than previous, and with introduction of huge caches, way more power efficient (cache doesn't consume much, and reduces expensive memory access which is a much longer travel needing more electrons).
Zen5 is also expected to be 25+% faster than Zen4, most coming from IPC, not from clocks.
Of course, even 25% is not really enough for anyone already having Zen4, but for those using older CPUs that's going to be a massive boost (albeit expensive, with new mobo+ram)
JamesSneed
I have a 5800x and probably will skip the 8000 series and upgrade on the 9000 series.
wavetrex
Well, that depends what you do... I edit videos as part of my work and while I'm not really in a hurry with them, it's starting to feel sluggish and rendering project at 4K takes a damn long time.
Luckily I only have one such project per week (so far), so not a problem leaving it encoding overnight... but if the activity increases, my current CPU will become a bottleneck.
Might even jump to the full 16-core with next upgrade if the price is not out of this world.
cucaulay malkin
Cant use GPU acceleration to help with rendering ? I thought it helps
https://www.purepc.pl/akceleracja-sprzetowa-w-programach-do-renderingu-i-obrobki-materialow-video-test-wydajnosci-kart-graficznych?page=0,6
wavetrex
It does, but hardware encoder has lower quality for the same bitrate.
I do a quick test render in hardware, goes pretty fast actually, if the video is alright and I haven't missed anything, I do it again with x264 (CPU only) with some very slow profile, takes 20 times longer, but retains much more fine detail, less blocky, smoother color transitions, etc, etc.
GPU encoding is basically "fixed function", with some profiles. Nvidia's encoder, despite them bragging a lot about it, is not any better. It might be even faster, but quality wise, it's pretty much the same.
None of them compare with the fine tuning that x264 has
https://en.wikipedia.org/wiki/X264
"x264 implements a large number of features compared to other H.264 encoders"
"x264 contains some psychovisual enhancements which aim to increase the subjective video quality of the encoded video"
However, you pay for it with lots of CPU cycles.
BlindBison
Personally I have no issue with games that are extremely demanding and due to being visually heavy run at lower framerates on modern hardware -- "IF" those games:
1) Actually justify that performance with their presentation (e.g. Post patch Cyberpunk is imo an example of a game which justifies its performance profile -- perhaps the single most impressive game in terms of visuals I have ever seen and doesn't seem to stutter as well). Some games run like garbage and have ludicrous memory requirements while looking virtually no better than games from a few years ago. As nice as some of the modeling work is in Jedi Survivor it really should run much better than it does for example.
2) Consistent smooth performance is a must with ideally zero stuttering behavior. E.g. I don't mind if a game that is visually feature heavy runs at a variable 30 or 40 FPS on my G-Sync display so long as it doesn't stutter. I also will gladly take aggressive asset fade in RAGE 1 style over stuttering if that's the trade off I have to make. Imo as much as possible should be asynchronous where you don't hold up the next frame just because an environment asset isn't ready yet -- just show a "dummy" object in the meantime then fade in the "real" object or surface over the next few frames once the game has it ready (iirc this is sort of how DXVK Async handles shader compilation -- you just don't get the shader immediately but that beats the alternative imo). Something like that should be possible to do I would think. Not saying that's a good thing to have aggressive asset fade like that, but it beats a harsh stutter and freezing the game with a microhitch by a country mile imo. What I find nutty is many recent games have BOTH problems (aggressive asset fade + traversal hitching). Redfall for example, like wtf is happening there (obligatory "I hate UE4" for PC games here).
On the one hand I get that memory management, core/thread scaling, and asynchronous rendering techniques are difficult and time consuming to implement. But on the other hand the games that bother to go to that effort are dramatically better as a result and I'd say it's always going to be worth doing that technical legwork work since you can carry if forward to future projects. Games like DOOM Eternal and Cyberpunk show that it's possible to have your cake and eat it too (proper core/thread scaling, practically zero stuttering, good performance considering the visual and gameplay return, etc). Red Dead 2 is also a great example and imo so are some advanced emulators out there. Good examples exist, but I dunno maybe studios need to hire fewer artists and more programmers. Or, maybe more restrictions need to place on those artists (apparently Id Software has many limitations in place that their artists must work within as per DF's old video on Id tech 7, but that's definitely wound up being a good thing).
Reddoguk
Performance Performance Performance.
nah just sort out better power draw/heat and costs and you may have a winner on your hands like the 5800/X3D was.
Someone show me better game performance with 32 threads over say 16 Threads. It doesn't really exist yet.
5800X3D beat all comers in where it mattered Gaming. All that other stuff like instructions most people have no idea or care about.
We probably know 8XXX will be still on AM5 because they would be shooting themselves in the foot if they needed AM6 mobos to run.
Having a 8xxx chip just being able to drop in any AM5 mobo will make them more popular. Having to buy ram and costly motherboards i bet people didn't get the best possible chips because of that.
Krizby
In AMD own marketing, 6000MT is the sweetspot, they even sent reviewers 6000MT EXPO kit along with Zen4 for reviews
vestibule
@BlindBison
Yeah I hear ya.
Currently playing control and that game is all over the shop with what it wants to give you in eyecandy.
Hmm. Time to try a few tricks.
CPC_RedDawn
They are not marketing that as safe. It's still classed as overclocking much like XMP is.
Krizby
If they advertise something are likely to destroy their product, there should be disclaimers that say it's not safe LOL.
If you see a car advertisement, there are disclaimers that the video are done by professionals, don't do it at home, or something like that
BlindBison
Control was very impressive on a technical level I recall. I will have to retry it once I upgrade my PC with the director's cut mod that got released (DF covered this recently and it has a slew of upgrades).
H83
That might be true but then it's very misleading.
I'm thinking about getting a 7700X with DDR5 6000mhz Ram because it's the sweet spot for the CPU.
But if they offer no guarantee that it's safe then there's something very wrong with the marketing around AM5...
XenthorX
13900k/7950x3d
Pretty much the same, you can actually run 6200/6400/6600+? DDR5 on AM5 as well. Extreme Overclocker for DDR5 seems to prefer Intel right now, but I doubt it means much for 99% user base.
CPC_RedDawn
Hence the word OVERCLOCKING in the slide you posted.
Their sites are filled with small print just like their last slides are always filled with things like "test bench used X, Y, Z to achieve X frame rate" or something like that.
Its much the same as ISP's with their "upto X Mb" speeds.
It can become very misleading I 100% agree on that.
It has been this way for years and years. Intel has had XMP for over a decade now but its still advertised as OVERCLOCKING and their actual officially supported RAM speeds always tend to be much lower. AMD now has the same with DOCP/XMP and now EXPO. Its still classed as overclocking and therefore not 100% guaranteed.
