Zen 5 already has extremely impressive memory bandwidth. A 192 core EPYC CPU has over committed memory bandwidth twice as much as a 16 core AM5. Don't forget, both of those CPU complex can handle twice the number of threads as they have cores. Even if they double RAM bandwidth, they are going to need a ton more lanes and a ton more cache because they are going to have wild latency. Now, imagine a world in which each chiplet has 32GB of RAM directly connected with a fraction of a contemporary memory bus latency. The performance per core would go up dramatically. We would have less CPU cores burning through memory faster. What's more, we would have a lower pin count socket for some cost saving.
According to a couple of large retailers, AMD motherboards are out selling Intel motherboards 10 to 1. Wow. Intel is a dumpster fire. They are in an existential crisis.
The I/O Die is from Zen 4 and should have been updated for Zen 5 to meet the increasing needs of RAM speeds and capacity. It just signals to people that Zen 5 is a stepping stone from Zen 4 to Zen 6. For most PC users, the current offerings are 100% fine. For people who lament the sane priced HEDT era (like me), what is being offered is gimped IO and RAM capacity to push the HEDT crowd into (now) obscenely priced TR setups. Current high end AM5 boards are insanely priced and are a balancing act of trade-offs with the limited lanes. There is nothing that an X870E has over X670E other than new wi-fi and mandated USB 4. In fact, some boards are regressions in expandability due to the USB 4 and limited PCIe lanes. Now what you propose sounds nice, and if there is a push for integrating RAM onto the CPU package with a forward thinking amount and speed, I would be fine with that absolutely. The problem is, most of the time, integrating usually is code for planned obsolescence and e-waste. I do not doubt that at all.
This is not an effective strategy. To a small extent, you can mitigate latency with cache and all CPUs do some of that but what happens when the CPU branches and needs to start loading a bunch of new cells into the program register? Suddenly, the cache is invalidated and we're waiting for a bunch of cycles while memory is loaded into the CPU. The higher the speed, the more important latency becomes. At 1 cycle of 5GHz, electricity can travel roughly 4cm. To lower latency, distance must be reduced.
I do not foresee RAM being integrated anytime soon. If evolving the I/O Die is not the solution due to latency, then what is?
This is because legions of folks will declare it a bad idea without understanding it or even testing it. Sooner or later, someone will notice that making the numbers higher is not helping outside the marketing department. We are approaching a limit of diminishing returns. I'm not saying we're there, just that it is on the medium horizon. The Fujitsu A64fx is at the core of a couple of the most powerful supercomputers in the world and it has a surprisingly low ARM core count with 32GB of on-package HBM2 RAM per part. The A64FX was the highest performing CPU, when released. The next iteration of this package is expected to restore that crown. It makes more sense than it would to continue what we are doing. Show me an application that randomly hits more than 64GB hard. Even database engines don't do that. I can't think of one that hits more than 4GB hard, including postgres. Apps tend to load data in big chunks and blow through it mostly sequentially. I know you're racking your brain, right now, to find a 1 in a million exception that proves I'm wrong. No need for that. PCIe 3 was really fast. PCIe for has twice the bandwidth but with variable aperture DMA, it's way more efficient than PCIe 3 making it more in the range of 3~4x faster. PCIe 5 doubles the bandwidth again. PCIe 5 can transfer 4GB/s per lane. My M.2 SSD is PCIe v3. It's a WD Black from a few years ago and it's really fast. If an SSD were 8x faster than mine, that would be really, really, really fast. Don't forget, M.2 is a connector with 4 lanes of PCIe. We are near the point, it would make a lot of sense to place 32 or 64GB of RAM on the package and scale above that with a paging swap partition. Latency would go down a lot, the vast majority of the time. Sure, the SSD is going to have high latency by RAM standards but don't forget we have column access latencies over 30 cycles for DDR5 RAM, right now. Some is well into the 40 cycle latency. That's quite a bit. We could even leave the RAM slots as a massive page swap and add the on package RAM, as a massive L4 cache. I suspect, at that point, it would make a lot of sense to delete L3 cache and use the RAM chiplet as L3. So, if someone does pursue this architecture in future, they will be able to load a 2TB database into RAM and directly access any cell, exactly like we can do right now with the current architecture. There is no need to panic over not being able to stuff cheap, Chinese, RAM into expansion slots.
I'm seeing a few strange rumors related to AMD. One such rumor is the 9950X3D will have 16 cores on a single die. This rumor seems to have roughly 5% chance of being true. It's interesting to think about, though. It would be a neat way for AMD to enable X3D caching for all cores. It's tough to imagine TSMC would have much yield at double the die size but maybe they are going for performance and will suffer the reduced yield for a niche part? Suffice to say, I'm not holding my breath on this. lol!
I just went for a new AMD build, also. I tried to be strong but was taken down by a Christmas week sale. It hasn't arrived, yet. The 9950X3D is going to be a single v-cache chiplet connected to just one of the 8 core chiplets. The other 8 core chiplet will be forced to cobble along with only 32 megabytes of L3 cache. There will be a flood of new parts coming out of AMD in the next couple of weeks. Even the mindless pro-Intel drones have nearly all stopped parroting Intel's superiority. Only a very few still tout Intel as leading the industry.
Ryzen 9800X3D sales are really high. This is an expensive part. While income from the 9800X3D will not move the needle in terms of gross revenue, it bodes very well for AMD.
There is a rumor AMD has very early engineering samples of Zen 6. When I first read this rumor, I dismissed it. Neither Zen 6 nor TSMC's 2nm node are scheduled until well into 2025. Then it hit me... lol! I believe Zen 5 is going to have legs through 2025. Zen 6 will be delayed by the fruit of Zen 5, not by technological limitations. That bodes very well for Zen 6 because a product rushed is a product flushed. Just ask Intel. With the very good Zen 5 core and CCX architecture, AMD is poised to perform extremely well through 2025.
I've been meaning to get back to this but life got in the way. I think you and I are on the same page regarding this issue of running into limitations on the bleeding edge. My issue is not where do we go from here, it is the artificial product stack segmentation that AMD is imposing on regular desktop SKUs vs more higher end SKUs. I'm going to just paste what RanceJustice from Hardforum.com stated on this topic, since its pretty much what I want to say anyway: "Old HEDT used to have a significant feature and value uplift for a moderately increased price over the mainstream platform - both Intel and AMD. It was a no compromises mixed use enthusiast machine - increased core count, RAM channels , PCI-E lanes, and other features while maintaining single/few threaded performance, overclock ability , and top of the line gaming experiences - and it seems that both AMD and Intel have moved away from offering this feature set and I think that's part of the problem here. Threadripper still exists, but over the past several generations it has transitioned into a later coming (anyone remember when HEDT platforms used to launch closer to the latest mainstream generation built on the same process, not 6 months or much longer afterward etc? I'm sure we'll never get a Nehalem / Westmere release schedule again but a man can dream) , with massive price increases way beyond the differential that ,in general, we used to expect between a high end mainstream vs a high end HEDT platform. Don't get me wrong I'm glad that AMD decided to allow core counts above 8 on Ryzen, but I do wish that along with such things they also would have created options for quad-core RAM and more PCI-E lanes for instance which, especially the latter have stayed static even in the wake of increasing demand thanks to multiple SSD storage usage; populating more than 1 PCI-E x16 slot means others fall back, and even on a top of the line X670E/X870E mobo populating certain M.2 slots could negatively impact your GPU because of lane sharing! These sorts of issues along with symmertical vcache suggestions tend to come from what seems to be the currently underserved userbase of those who want the "old HEDT" experience. At the moment, it seems both Ryzen and Threadripper do parts of those things, but there are significant compromises to make either way and in the case of choosing the latter a significant premium. I don't know if AMD needs to add more variation for features in the Ryzen lineup, or perhaps continue to iterate on Threadripper given they've already as of the 7000 series split out "Threadripper PRO" WRX80 vs "Threadripper" TRX50 chipsets, but it seems there's room for them to shape another path forward for a modern HEDT serving some of the same user desires that the 'old' version provided." I'm not going to drop thousands on something that was available to us a short while ago just because AMD wants to force people into higher priced SKUs to pad their profit margins. There is a healthy market for HEDT.
You could very well be closer to the mark than I am, RTN. Either way, I appreciate your perspective and hope you don't cave to my opinion. Nice post, BTW. We usually agree 100% but if our views vary a bit, maybe thats a good thing?
The numbers I have been seeing suggest more than doubling of sales at AMD. When this is considered in conjunction with the product shortages AMD is currently under, Q1 should be a barn storming quarter for AMD with momentum into the foreseeable future.
Of course it is. It gets people to re-evaluate their stances to further strengthen them or to facilitate a change of mind. Good. Competition is always a healthy thing for the consumer.
I wonder how much of the volume dynamic is understood by the average person. I would guess none. Perhaps a couple of folks in here have an idea? AMD has literally doubled their sales. Over. They are ramping as hard as they possibly can but they are struggling to keep product in the channel. AMD has a lot of debt but they were making their debt service at less than half current volume. The next few quarters are going to be heavy margin quarters. Meanwhile, Intel is drowning in debt. They were losing money when they were selling 35% more than they currently are. If they couldn't carry the load with 80% market share, it's not going to go well with them in the mid 50% range. Gross may only be down 25% but losses will be a multiple due to leverage. That concludes today's lesson on borrowing responsibly. lol!
I am loving the rather frank wording by AMD : https://www.tomshardware.com/pc-com...-product-is-causing-ryzen-9-9800x3d-shortages
I think that a lot of the great performance from AMD is/was baked into the price...not sure AMD is too much of a bargain right now. Intel feels like it is a little undervalued...
On die RAM, anyone? Lol! https://www.tomshardware.com/tech-i...-mi325x-smiles-for-the-camera-256-gb-of-hbm3e This doesn't prove anything or make any points. I just thought it was amusing, given recent discussions.
