Sunday, February 4th 2024
AMD Readies X870E Chipset to Launch Alongside First Ryzen 9000 "Granite Ridge" CPUs
AMD is readying the new 800-series motherboard chipset to launch alongside its next-generation Ryzen 9000 series "Granite Ridge" desktop processors that implement the "Zen 5" microarchitecture. The chipset family will be led by the AMD X870E, a successor to the current X670E. Since AMD isn't changing the CPU socket, and this is very much the same Socket AM5, the 800-series chipset will support not just "Granite Ridge" at launch, but also the Ryzen 7000 series "Raphael," and Ryzen 8000 series "Hawk Point." Moore's Law is Dead goes into the details of what sets the X870E apart from the current X670E, and it all has to do with USB4.
Apparently, motherboard manufacturers will be mandated to include 40 Gbps USB4 connectivity with AMD X870E, which essentially makes the chipset a 3-chip solution—two Promontory 21 bridge chips, and a discrete ASMedia ASM4242 USB4 host controller; although it's possible that AMD's QVL will allow other brands of USB4 controllers as they become available. The Ryzen 9000 series "Granite Ridge" are chiplet based processors just like the Ryzen 7000 "Raphael," and while the 4 nm "Zen 5" CCDs are new, the 6 nm client I/O die (cIOD) is largely carried over from "Raphael," with a few updates to its memory controller. DDR5-6400 will be the new AMD-recommended "sweetspot" speed; although AMD might get its motherboard vendors to support DDR5-8000 EXPO profiles with an FCLK of 2400 MHz, and a divider.
The Ryzen 9000 series "Granite Ridge" will launch alongside a new wave of AMD X870E motherboards, although these processors very much will be supported on AMD 600-series chipset motherboards with BIOS updates. The vast majority of Socket AM5 motherboards feature USB BIOS Flashback, and so you could even pick up a 600-series chipset motherboard with a Ryzen 9000 series processor in combos. The company might expand the 800-series with other chipset models, such as the X870, B850, and the new B840 in the entry level.
Sources:
Moore's Law is Dead (YouTube), Tweaktown
Apparently, motherboard manufacturers will be mandated to include 40 Gbps USB4 connectivity with AMD X870E, which essentially makes the chipset a 3-chip solution—two Promontory 21 bridge chips, and a discrete ASMedia ASM4242 USB4 host controller; although it's possible that AMD's QVL will allow other brands of USB4 controllers as they become available. The Ryzen 9000 series "Granite Ridge" are chiplet based processors just like the Ryzen 7000 "Raphael," and while the 4 nm "Zen 5" CCDs are new, the 6 nm client I/O die (cIOD) is largely carried over from "Raphael," with a few updates to its memory controller. DDR5-6400 will be the new AMD-recommended "sweetspot" speed; although AMD might get its motherboard vendors to support DDR5-8000 EXPO profiles with an FCLK of 2400 MHz, and a divider.
220 Comments on AMD Readies X870E Chipset to Launch Alongside First Ryzen 9000 "Granite Ridge" CPUs
The orange line is 4800 MHz, and the brown line is 6000 MHz. One could wonder why the result with 6000 MHz is lower than 4800 MHz, but I'd say it's within margin of error.
This is the same difference that I see in games (that is: none).
If you're using this to try and objectively prove some point about gaming, it's entirely irrelevant.
Do a gaming comparison if you want to talk about gaming, or a benchmark that doesn't fit in cache.
I know that Cinebench doesn't scale well with RAM speed, but neither does my gaming experience using a 7800 XT at 1440 UW.
But you cannot notice this trough Cinebench 23, you need something like Geekbench or other program that can meager it.
Horizon Zero Down did ~10-16k Read / 3-4k Write.
Shadow of the Tomb Rider did 6-12k Read / 2-3k Write.
So this is far from theoretical 70-100GBs that Aida64 showed.
Having both is nice though. I.e. Raptor Lake with AIDA 120/120/120 R/W/C and sub 50 ns latency at 8000 MT+ DDR5.
Whether the price for fine tuned Intel is worth it , is another topic.Ryzen x3D are under 7ns to the 3D cache. Of course, when there is what the CPU is looking for :D
In the meantime, here are some Cyberpunk 2077 results with maxed out graphics, 3440x1440, no RT, no upscaling:
4800 MHz RAM:
6000 MHz RAM:
Trying to prove your point about "never" being CPU limited by striving your hardest to find a game and settings combo that pushes your GPU as hard as it can? These representative of your daily settings?
Again, it's 6000 MT, not MHz.
Why not buy a 60Hz ultrawide rather than a 144 Hz one then and save some money for a better GPU? Surely it will make no differences to you if you're happy to (and normally) play at 45 FPS.
But I doubt this is actually the case.
If you genuinely are happy to play at 45 FPS, you massively overspent on your CPU and massively underspent on your GPU.
While consuming less power too.
And with the option of DLAA to bring better than native image quality.
1. 60 Hz ultrawides either don't exist, or they're so rare that I've never seen one, and
2. The Freesync range on my monitor goes from 48 to 144 Hz, so any performance within my expected range gives me a smooth, tear-free experience. It also has LFC, which basically doubles or triples my refresh rate relative to game performance, should it drop below 48 FPS (although I aim for it not to). I love my monitor for more things than its max refresh rate.I'm with you on that. The thing is that hardware is my hobby. I tend to massively overspend on stuff that I'm curious about, but have no need for. I would never recommend anyone else to do the same.
If I only ever bought what I needed, I'd still be on my Rocket Lake platform with a 24" 1080p monitor.
Again, I am ok with buying DDR5-6000 CL30 or even slightly faster RAM that works out of the box, which is what I did. I am ok with buying cheap RAM and OCing it to DDR5-6x00 CL30 if you have the time, which I don't have.
But unless I see perceptible gains in several games, in low fps for example, in 1440p or 3440x1440, with anything you can't buy for a few bucks above what DDR5-6000 costs, I'm not convinced it's woth my time. Even then, since I don't wan't to spend 1.000€+ on a GPU and thus will never have a 4090 or even 4080, I still will always be GPU-bound in 3440x1440.And what is the sensible limit of the IF? Did I read right that CPUs with only one CCD are limited in max bandwith, too?That's one problem I see with memory benchmarks. People only use games they know will scale with memory, even though theyre old and it's totally useless in some cases (who still plays Tomb Raider and what difference does it make if it's 120fps or 200fps minimum?). If they just used the most played contemporal games, to many wouldn't benefit from fast RAM.He didn't, his is a very common setup. And if this shows that RAM-speed doesn't make any difference to fps, minimum as well as average, in 3440x1440, if you don't have a 4090, then nobody that hasn't should care about memory speed. And there sure as hell won't be any tangible difference above DDR5-6000 CL30 out of the box that would be worth the trouble.I will admit that any Radeon isn't ideal for Cyberpunk with RT but your argument is only half valid. Here in Germany a 7800X3D costs about 70€ more than a 7700X, which is 50€ above a 7700, which is only 25€ above a 7600X. The cheapes AM5-CPU, 7500F costs 190€ less than a 7800X3D.
The difference between the cheapest 7800XT and 7900XT is 210€, between the cheapest 4070S and 4070TiS 240€. Yes, a 7900GRE and 4070S would be only 50€ and 60€ above 7800XT and 4070 respectively, so I would at least go for those at the moment, but you can't buy that much more GPU-power by skimping on the CPU, even if you chose the cheapes CPU available. A 7800X3D is not a very expensive CPU, but anything from 7900XT and 4070Ti up is a very expensive GPU. If every € counts, you should skimp on every other component, including RAM, to buy the most potent GPU, it gaming is your goal, but the 7800X3D is worth it in most games. It's the only viable reason to choose AM5 for gaming. Anything below has a faster, cheaper Intel-counterpart, if not as efficient. Of course, there are other reasons to consider that are less ovjective.
The point of a fast CPU is to provide more frames than the GPU can render, making you GPU limited.
If you're GPU limited all the time (99% usage while gaming), it's a sign of a well optimized PC. But there's a limit to this. Zero point building a super high performance CPU PC designed to push 2,3,400 frames consistently, if it's paired with a GPU that can't even consistently push 100 FPS at the settings/resolution you're playing at.
I'm a strong proponent against the 7900/X3D because of it's compromise 6+6 design, but if you're GPU limited to a significant extent (as a 7800X3D+7800XT GPU build is, especially at ultrawide), there's practically zero reason to pick the similarly priced 8 core 7800X3D, which gives up multithreaded performance and therefore productivity performance, for a gaming focused design. You might as well get the strong productivity CPU that gives up a little at the upper end of framerates, but is still well above the ~100FPS that the GPU can deliver). Although to be clear, we're talking about a build that gets ~45 FPS here. And the owner of that build is happy with that (I won't argue if this is right or wrong, but this emphasises the fact that CPU is of very little importance at these low frame rates, paired with a 7800XT, you could get an identical gaming experience with a 10400f, a $100 Skylake CPU).
In 90%+ of use cases, excepting CPU limited games such as simulation games (minecraft, racing games etc) or esports games (CS2, overwatch etc), the user is going to be GPU limited. Therefore GPU should be the main priority for budget. Spending $700+ on a platform (7800X3D + AM5 Motherboard + RAM), and $500 on a GPU is silly, objectively. You could make the argument "future proof platform" but that only makes sense when the whole build is high end. If you're doing a mid range build, you don't waste money on "future proof" and frivolities.
I agree, the RTX 4070 Ti Super is expensive at $800. But I guarantee for the budget this build cost, I could make something with a 4070 Ti Super that didn't do any stupid compromises like a quad core i3 or anything. -
While it's true everyone has different requirements for gaming. I.e. some people are happy at 45 FPS, others want to be above 200 all the time. In a discussion about RAM performance, it's completely irrelevant to talk about the 45 FPS scenario, because a literal DDR3 platform would happily chug away at those frame rates in modern games. The criteria for discussing RAM, is the high FPS arena. It's also important to mention that RAM tuning is literally free. It costs nothing.
You can make the efficiency argument for the 7800X3D, but again, just buy a non X3D chip and lower the voltage/frequency, you'll get the same results. After all it's the same silicon just with a different voltage curve that the cache can tolerate.This encapsulates my point. I have no issue with AusWolf, because he's clear that he's made these purchasing decisions based on hobbyist interests, rather than logical routes to performance within a set budget.
This is what I'm talking about.
You could literally be using a Zen 2 platform and getting ~95 FPS averages, more than twice the 45 FPS AusWolf is getting.
This is the 2024 test bench, so these averages are based off new games too, not the ones released when Zen 2 was introduced.
This base would give a 125 FPS average in modern games at 1440p, while costing 2/3 the price of just the 7800X3D CPU only.
The rest of the saved money could go straight to GPU.
That's buying new too, I'm sure there's plenty of 10th gen combos kicking around on the second hand market for ~$150.
Even if you want to be safe and base your CPU off Minimum FPS, Skylake/Zen 2 still provide well over 60 FPS.
The dual CCD X3D chips also post better synthetic RAM benchmarks, because those benchmarks make use of the frequency CCD (which doesn't have VCache). Higher frequency CPUs operate with lower latency to the system memory. However, because the VCache CCD is practically identically clocked (5.05GHz on 7800X3D, 5.15 on 7900 (but six cores not eight), and 5.25 Ghz max boost on 7950X3D for the VCache CCD), in gaming, this is practically meaningless.
Sure, if you're on really limited budget then it might be to expensive, as might AM5 altogether. But I would even skimp more on RAM than on the CPU if that would get me the 7800X3D, which is less dependent on fast RAM, as you said yourself.
Your argument is still to absolute. I am talking about investing time and/or money into RAM faster than DDR5-6000/6200/6400 CL30/32/34 with EXPO, which you can buy for not much more than any old DDR5-5x00 and which works out of the box. There is next to zero benefit to going faster on AM5 as of now and it's not worth the time needed to get it stable.
It's hard to say what to invest into when you're on a thight budget. From your benchmark it would seem the difference between all AM5-CPUs without X3D is neglegible (7fps average) and it could be better to get a 7600(X) with fast RAM than a 7700X with very cheap RAM (but that you could OC yourself if you time but no money).I did the same, this time. I chose my mainboard with rather unlikely future upgrades in mind (two PCIe x4-slots for 10GbE and TB4/USB4), chose the 7800X3D for above reasons, chose RAM without high rising heat sink for CPU-cooler compatiblity, DDR5-6000 with EXPO because it was called the sweet spot then and even paid 10€ extra for CL30 instead of CL32 (certainly not worth it). I chose a 1.000W PSU because I had not yet decided on a GPU and was considering used 3090.
Now I had to move, have far less space on my desktop for the time being an am back to my Dell U2410 with 1920x1200@60Hz and use the RTX 2070 from my old PC. I don't play any new, demanding games, so everything is at 60fps constantly.
For my 34" 2440x1440@144-165Hz, when I have the space for it again, I probably will buy a 5070 or something in that range.
It's not unusual to be able to halve TRFC and, combined with other timings/subtiming tunes and general BIOS optimization, get 25-30% lower latency.
What I currently run. tRFC in ns is typically 300+ stock XMP/EXPO.
So you're still advocating for a 7800X3D which will show zero benefits with a mid range GPU over a cheaper CPU/platform, but don't see the dissonance by arguing improving RAM (and therefore CPU FPS performance), for free, is pointless?
Here's an example of what I'm talking about. tRFC (which affects game performance significantly, more so than CL tbh) is practically double the latency in ns with stock EXPO.
"Sweet Spot"
@ir_cow if you want to weigh in.