Thursday, January 18th 2024
AMD Ryzen 8000G APU Memory Sweet Spot is DDR5-6000
During CES, PCWorld had a chat with Donny Woligroski, Technical Marketing Manager at AMD. The new Ryzen 8000G APUs were a large part of what covered in the almost 17 minute long video and PCWorld got some details that weren't covered in the official press materials that AMD released at the launch. The officially supported memory speed listed by AMD is DDR5-5600, which is a step up from the official speed of DDR5-5200 for the Ryzen 7000-series CPUs.
However, we know that the Ryzen 7000-series is more than happy to use faster memory and as before, AMD has an unofficial memory sweet spot and just as with the Ryzen 7000-series, the Ryzen 8000G-series of APUs has a memory sweet spot of DDR5-6000. That said, it's unknown if the Ryzen 8000G-series will support faster memory or will start flaking out above DDR5-6000, like many Ryzen 7000-series CPUs do unless you switch to a 1:2 ratio. Woligroski is also pointing out that dual-channel is a must to get the best performance out of the new APUs, although this shouldn't really surprise anyone. Full video after the break.
However, we know that the Ryzen 7000-series is more than happy to use faster memory and as before, AMD has an unofficial memory sweet spot and just as with the Ryzen 7000-series, the Ryzen 8000G-series of APUs has a memory sweet spot of DDR5-6000. That said, it's unknown if the Ryzen 8000G-series will support faster memory or will start flaking out above DDR5-6000, like many Ryzen 7000-series CPUs do unless you switch to a 1:2 ratio. Woligroski is also pointing out that dual-channel is a must to get the best performance out of the new APUs, although this shouldn't really surprise anyone. Full video after the break.
30 Comments on AMD Ryzen 8000G APU Memory Sweet Spot is DDR5-6000
arstechnica.com/gadgets/2024/01/intel-freshens-up-its-old-laptop-and-desktop-cpus-with-speed-bumps-new-names
Intel refresh training
People just buy what ever they put out.
But, No... why keep a logical and easy-to-understand line. It seems like they want to get straight to the absurdly long numbers (10XXX), or switch to another even more confusing naming scheme.
Yes, the optimal solution would be to have AMD stick with and actual provide effing supply for Phoenix, not kill it off within a year and basically rebrand it as 8000, but here we are already. AMD's new favourite activity ever since their marketing dept dipped their toes into Renoir, and subsequently conjured up the spaghetti monster that is Ryzen 7000 family. I have a feeling this "sweet spot" was decided by marketing dudes with no knowledge of the product on the engineering side, because even if they are fundamentally different tech, I find it hard to believe that Phoenix/Hawk does LPDDR5-7500 with ease yet needs to step down to 1:2 DDR5 on a desktop platform with zero power saving measures for UMC and Fabric. And the CCD => monolithic switch alone usually results in a 1:1 ceiling improvement. I don't expect these marketing dept dudes to be familiar with APU UMC OC, or what Hynix A can already do in 2024 (DDR5-8000+).
But zero improvement over Raphael (6000) would be disappointing to say the least, considering 780M is known to be limited by DDR5-5600 1:2 (ie. in Phoenix -HS laptops).
Also, considering the other comment this AMD rep made was about the "importance" of running dual channel memory, I'll believe shitty UMC when I see it on an official slide or actuual 8700G OC.
Also, they indeed inflating the naming numbers, without any apparent reason. At this pace they will end up the 10K naming shceme as intel did, before switching to even more confusing "Ultra".
This doesn't serve any good purpose, and it's bad, when the buyer needs "decoding ring" in order to buy their products. Except it's for investors to capitalize on unsold stocks of old chips.
IMHO, many companies used to stick with same product naming sceme, if the SKU/product "core" is the same. Eventually it is much easier for the own production itself. I'm layman, but from my research, this seems to be the same 7840-ish, chip. Since, the "Hawk Point" is exactly same thing as "Phoenix Point", but with "AI" chip added. There's seems no incemental difference in the silicon/architecture itself.
Even AMD own official page for 8700G claims it to be "Proenix".
The problem is, people do not buy CPU separately from mobile product, like laptops etc, so the it's less the problem, since it requires a lot more of research for mobile device, to begin with.
But for desktop... why add confusion? Especially Lisa Su herself, during AM5 announcement, something as "APUs on this architecture, will come later..." True. But some marketing doecheb*gs of come X company, doesn't make it "normal" or "correct" behaiour for company Y. This doesn't make it ok, at all.
It makes no sense for anything new to be Phoenix at this point in time as all the signs point to Phoenix being axed as an entire family, Hawk is its direct replacement. AMD slides for Ryzen 8000 describe Phoenix exclusively in the past tense.
I didn't say that Hawk Point is functionally any different from Phoenix aside from it's NPU, it's a low effort rebrand. But it makes no sense to be renaming a product launched as 8000 series, backwards into 7000 series.
Zen5 and 4 CPU cores aren't anywhere near bandwidth limited. If anything, due to the high amount of cache, they're relatively bandwidth insensitive. They're actually more sensitive to latency which is why the focus on overclocking the RAM for Zen4 has mostly focused on getting the timings (tertiary and secondary in particular) down with EXPO once you get to 6000.
The APU's on the other could indeed absolutely use DDR5 8000....for the iGPU! NOT THE CPU!! While it'd be nifty if faster RAM made a big difference it doesn't so its worth bothering with on Zen4 or Zen5. Unless you're trying for some AIDA read/write world record of course.
Being able to be pretty fast with, relatively cheap, DDR5 6000 is a good thing. You can get 2x16GB DDR5 6000 kits for ~$90-ish. Mean while cheapest 2x16GB DDR5 8000 kits run around $200-ish on newegg. Twice the price for minor performance gains, even with the lowest possible graphical presets and low resolution, is not something to get worked up over.
www.techpowerup.com/review/klevv-cras-v-rgb-ddr5-6400-cl32-2x-16-gb/7.html
Much of the time in game the difference was only 1-2fps vs DDR5 6400 vs DDR5 6000 for Zen4 guys. Sometimes it was even less. Less than 1 fps!! For actual production CPU benches the differences were usually just as minor.
This is not at all a bandwidth limited CPU and its bizarre to me that people think it is.
[CPU 7000]
CPU die---(FCLK,XGMI link on CPU PCB)---I/O die(UCLK,iGPU)---Memory(MCLK*2=DDR)
MCLK3200:UCLK3200:FCLK2000 (DDR5-6400, Low Latency)
[APU 8000G]
CPU+I/O(FCLK,UCLK,iGPU)---Memory(MCLK*2=DDR)
MCLK4200:FCLK2400 (DDR5-8400, High Bandwith)
*Reccomended B650/X670(Support PBO/Curve Optimizer) 2Slot DIMM M/B + Hynix A-die or Next Generation hynix chip
*Hynix's next generation DDR5 features ClockDriver to achieve higher clocks. (More than A-die, DDR5-6400 1.1v Native)
[Tuesday, January 30th 2024] AMD Ryzen 7 8700G Loves Memory Overclocking, which Vastly Favors its iGPU Performance
www.techpowerup.com/318446/amd-ryzen-7-8700g-loves-memory-overclocking-which-vastly-favors-its-igpu-performance