Friday, May 31st 2024

AMD Shuffles Feature-sets of its 800-series Chipset, X870 is B650E Successor

AMD is debuting its Ryzen 9000 series "Granite Ridge" desktop processors powered by the "Zen 5" microarchitecture later this year. These chips are compatible with existing AMD 600-series chipset motherboards with a simple UEFI firmware update, but the company is also taking the opportunity to launch the AMD 800-series chipset family alongside these chips. The lineup will be led by the AMD X870E, followed by the X870. These two chipsets should launch immediately alongside the new processors, but will later be joined by the AMD B850 and B840. There's no entry-level chipset planned, the AMD A620 will continue to hold the fort for AMD here. There is an interesting new mix of product differentiation, according to a leaked GIGABYTE slide scored by VideoCardz.

If you recall, the X670E and X670 were differentiated by a lack of Gen 5 PCI-Express x16 PEG slots on the X670, which instead was limited to Gen 4 on the PEG slot. The X670 still had Gen 5 NVMe slots attached to the CPU, and had practically the same I/O features as the X670E, including the same counts of downstream PCIe Gen 4 general purpose lanes. Both the X670E and X670 are 2-chip solutions, in that the second chip is a connected to the general purpose PCIe lanes of the first chip, which in turn is connected to the processor. Things are going to change with the 800-series. The top-spec X870E will be a 2-chip solution, with PCIe Gen 4 general purpose lane counts resembling the X670E; but the X870 is a single-chip solution that more closely resembles the B650E in I/O. The X870 (non-E) now gives you Gen 5 PCI-Express x16 PEG, just like the X870E and the B650E, and at least one Gen 5 x4 NVMe slot attached to the CPU, but has fewer downstream Gen 4 general purpose PCIe lanes than the X670. Both the X870E and X870 assure USB4 connectivity, and support CPU overclocking. Things get very interesting in the mid-range.
The AMD B850 is very similar to the X870 in terms of downstream general purpose PCIe Gen 4 lanes. What's more, it even assures a Gen 5 x16 PEG slot, much like the X870. Where it differs from the X870 is its CPU-attached NVMe slots. Gen 5 is made optional here (motherboard vendors can provide Gen 5 if they want, but are perfectly free to offer Gen 4 instead). Unlike the X870E and X870, the B850 doesn't mandate USB4 connectivity, however, motherboard vendors will be free to offer discrete USB4 controllers on their boards. Much like the B650 and its predecessors dating all the way back to the B350, the B850 supports CPU overclocking.

The B840 is a new introduction with this generation, there's no real predecessor to it, although it closely resembles the A620A, which in turn resembles the B550. It completely removes all forms of Gen 5 PCIe from the platform—the x16 slots are limited to Gen 4, as are the M.2 NVMe slots attached to the CPU. This chipset also lacks CPU overclocking support. It does retain memory overclocking, and B840 motherboards should support AMD EXPO, as well as manual memory overclocking. What sets the B840 apart from the A620 is its Gen 4 PCIe connectivity both along the PEG and general purpose PCIe.
Source: VideoCardz
Add your own comment

74 Comments on AMD Shuffles Feature-sets of its 800-series Chipset, X870 is B650E Successor

#51
TheLostSwede
News Editor
regswww.usb.org/document-library/usb4r-specification-v20
USB4 Connection Manager Guide Ver. 2.0 .pdf, page 60

Yes, USB4 can operate at slower speed, but the MINIMUM speed for a host controller is 20 Gbps and so far, ALL USB4 host controllers are 40 Gbps and ALL hubs have to be 40 Gbps. Devices can operate at slower speeds. If you're going to read the specs, please try to understand them.
You know, I spent a month meeting with companies that were developing USB4 hardware and doing research before I wrote the USB4 article, I didn't just pull some shit off the internet, even if it's from the USB-IF, since you clearly didn't understand what you read.
Posted on Reply
#52
Wirko
Tek-CheckGen5 chipset link will be expensive. And such chipset does not even exist at the moment.
Maybe not very expensive. The Samsung 990 Evo SSD is at least an indication that Gen5 can be had at down to earth prices.

A bigger issue for AMD might be wafer allocation. The chipset would have to be made at 7nm/6nm in order to not overheat. AMD need lots of 7nm and 6nm wafers for several other chiplettes, so a completely new Promontory will have to wait some more.
Posted on Reply
#53
Tek-Check
WirkoMaybe not very expensive. The Samsung 990 Evo SSD is at least an indication that Gen5 can be had at down to earth prices.
You saw how motherboard prices went up with introduction of Gen5. Traces needed higher components to maintain signal integrity.
Even more such traces would be needed for Gen5 chipset, plus more retimers.
WirkoA bigger issue for AMD might be wafer allocation. The chipset would have to be made at 7nm/6nm in order to not overheat. AMD need lots of 7nm and 6nm wafers for several other chiplettes, so a completely new Promontory will have to wait some more.
It's not an issue as Gen5 chipset will not be introduced for a few good years. It's just not needed by anyone. Desktop platforms have even struggled to make any meaningful use of Gen5 lanes on CPU since late 2021... Hardly a few enthusiasts use Gen5 SSDs, and those giant heat sinks are monstrocities. Only by the end of this year new GPUs will start using Gen5 lanes, 3 years after their introduction on motherboards. Lack of technology syncing there.

There will be even less reasons to introduce Gen5 chipset link any time soon, as the increase in PCIe lanes is driven predominantly by CPU. It's enough if a new platform adds another one or two Gen5 x4 link and by 2030 we will have at least 32 Gen5 lanes from CPU alone. Plenty. Modern chipset simply needs to remain cheap enough as to not drive motherboard prices too much and reasonably connected to southern components, mostly USB, SATA and other smaller PCIe devices.
Posted on Reply
#54
kapone32
Here is the issue for AMD. X670E boards are expensive. X670E boards provide more 5.0 support than most people have. My board has 3 USB C ports at the back but it is not like an external GPU is cheap in any way. This is not the days of 2 slot HDD enclosures with E-SATA ports USB 4 is not enough of a hook to get a current AM5 user to spend another $500+ to get X870E. I plan on getting whatever the next 12 core X3D chip to plop in my X670E board and use a A620 to add my 7900X3D to the mining farm. No DGPU needed. The memory controller is in the CPU anyway and these 5.0 boards are built to a very high electrical standard.

I will add more. X670E gave us up to 7 M2 support and we don't really need any more than that but USB 4. So I might purchase some faster RAM to bring me up from 5200 MT/s.
Posted on Reply
#56
Polyh3dron
Is a X870E motherboard really going to be capable of running DDR5 at a faster speed than a comparable X670E motherboard with the same CPU and same RAM?
Posted on Reply
#57
regs
TheLostSwedeYes, USB4 can operate at slower speed, but the MINIMUM speed for a host controller is 20 Gbps and so far, ALL USB4 host controllers are 40 Gbps and ALL hubs have to be 40 Gbps. Devices can operate at slower speeds. If you're going to read the specs, please try to understand them.
You know, I spent a month meeting with companies that were developing USB4 hardware and doing research before I wrote the USB4 article, I didn't just pull some shit off the internet, even if it's from the USB-IF, since you clearly didn't understand what you read.
Can you point a page in specification? Not a news article.
Posted on Reply
#58
Ruru
S.T.A.R.S.
For a first time in ages, AMD has way more confusing chipset lineup than Intel does.
Posted on Reply
#59
Assimilator
TheLostSwedeWhy you hating on AMD? This is just to keep the motherboard makers happy so they have something "new" to sell. Same as Intel.
Because they can do better, they know they can do better, and they're choosing not to.
Keullo-eFor a first time in ages, AMD has way more confusing chipset lineup than Intel does.
Nah, that was true with the first generation of AM5. They're just making it even more confusing now.
Posted on Reply
#60
JWNoctis
Keullo-eFor a first time in ages, AMD has way more confusing chipset lineup than Intel does.
AssimilatorNah, that was true with the first generation of AM5. They're just making it even more confusing now.
Interestingly, they are doing all of this with a single design of chipset silicon that even X870E-B850 is apparently still using. Maybe they should have cleaned their customer-facing lineup similarly well.

Probably not helped by the fact that many B650 non-E boards had no PCIe 5.0 at all, despite specification otherwise.

First time I've seen a reference to a Promontory 19 chipset. I thought A620 was a cut-down Promontory 21. Maybe that's what the cut-down version is called?

What's the deal with that first image in the article, though? That diagram had so much compression artifact that it almost look AI-generated - or maybe scrambled, by someone who really shouldn't be taking that screenshot, off what looked like a remote desktop interface from the watermarks. ;)
Posted on Reply
#61
P4-630
btarunrAMD 800-series chipset family
So childish Lisa....
Posted on Reply
#62
Neo_Morpheus
TheLostSwedeSorry, but you clearly haven't read up on USB4.
Please read this before making uneducated comments.
And for someone that is soo educated, please read my comment again, since clearly reading comprehension failed you.
Posted on Reply
#63
TheLostSwede
News Editor
regsCan you point a page in specification? Not a news article.
That wasn't a news article, it was a USB4 primer, but clearly I'm not good enough as a source, despite having worked most of my adult life in the tech industry and lived for over 14 years in Taiwan and have friends at the companies making the USB4 devices, who I interviewed for the piece I wrote, some of which are C-leve executives at said companies. Then I guess you can go ahead and believe whatever you want to believe.
AssimilatorBecause they can do better, they know they can do better, and they're choosing not to.
What's missing that would change the world for anyone in a truly useful way?
JWNoctisFirst time I've seen a reference to a Promontory 19 chipset. I thought A620 was a cut-down Promontory 21. Maybe that's what the cut-down version is called?
Yes, this is correct. However, the initial A620 boards used the Prom21 due to the Prom19 being late from ASMedia.
JWNoctisWhat's the deal with that first image in the article, though? That diagram had so much compression artifact that it almost look AI-generated - or maybe scrambled, by someone who really shouldn't be taking that screenshot, off what looked like a remote desktop interface from the watermarks. ;)
It's a leak and someone has taken to releasing these kind of images as leaks online recently and then it seems like someone else tries to fix them with "AI".
Neo_MorpheusAnd for someone that is soo educated, please read my comment again, since clearly reading comprehension failed you.
Again, HOW is USB4 half-assed compared to Thunderbolt?
The fact that the ASM4242 is two years late is hardly AMD's fault, as it got stuck in certification hell at the USB-IF due to a certain company that "donated" the specs to the USB-IF didn't explain how a lot of things worked, but assumed that anyone who was going to make a USB4 host controller knew how that company did things internally.
The ASM4242 was always supposed to be connected directly to the CPU this way. This info leaked way before the AM5 platform launched.
Posted on Reply
#64
regs
TheLostSwedeThat wasn't a news article, it was a USB4 primer, but clearly I'm not good enough as a source, despite having worked most of my adult life in the tech industry and lived for over 14 years in Taiwan and have friends at the companies making the USB4 devices, who I interviewed for the piece I wrote, some of which are C-leve executives at said companies. Then I guess you can go ahead and believe whatever you want to believe.
You still fail to link a page in specification
Posted on Reply
#65
TheLostSwede
News Editor
regsYou still fail to link a page in specification
Since I don't have to. I know I'm correct and you're free to go and look it up.
Posted on Reply
#66
regs
Specs that are only matter
Posted on Reply
#67
TheLostSwede
News Editor
regsSpecs that are only matter
Happy now? No mention of 10 Gbps.

Posted on Reply
#68
Tek-Check
regsCan you point a page in specification? Not a news article.
@TheLostSwede is right and he gave you the spec. The host MUST support minimum 20 Gbps.
regsYou still fail to link a page in specification
Why don't you read the spec on your own?
TheLostSwedeHappy now? No mention of 10 Gbps.
The only case where 10 Gbps is bare minimum would be tunnelled USB 3.2 from the host to hub and onto a peripheral device. So, in indirect USB connections where other data packets are present, such as DP and/or PCIe data.

Obviously, the host still must provide more than 10 Gbps in order for USB tunnelling to take place. And that's the whole point of bare minimum requirement of 20 Gbps for the host. Practically, every host has 40 Gbps speed due to Microsoft requirements for PCIe data from x4 PHY.
Posted on Reply
#69
regs
TheLostSwedeHappy now? No mention of 10 Gbps.

Operation means duplex - both ways - Gen 2 (minimum of 10+10), Gen 3 (minimum 20+20) and Gen 4 (minimum of 40+40). It doesn't mean that USB4 shouldn't support single link, it doesn't mean it shouldn't support dual link. It's a minimum operation for each Gen.

2.2 USB4 Fabric Architecture
The USB4 Fabric is designed to meet the needs of multiple transport protocols. Its main features are:
Signaling rates that support high throughput interconnects :
- 10 Gbps (for Gen 2), 20 Gbps (for Gen 3), and 40 Gbps (for Gen 4).
Posted on Reply
#70
Speedyblupi
JWNoctisFirst time I've seen a reference to a Promontory 19 chipset. I thought A620 was a cut-down Promontory 21. Maybe that's what the cut-down version is called?
Promontory 19 is B550 and A620A. The functional difference is that the A620 chipset uses a PCIe 4.0 link to the CPU, while A620A/B840 use PCIe 3.0. A620A has fewer lanes enabled than B550, and B840 will presumably have more lanes than A620A (4 PCIe 3.0 + 4 SATA), possibly the same as B550 (10 PCIe 3.0 + 4 SATA or 8 PCIe 3.0 + 6 SATA) or somewhere in between the two.
A620 can be either a cut-down Promontory 21 chip (allowing Prom21 chips with defects to still be sold), or a Promontory 22 chip (which is smaller and cheaper to manufacture).
Posted on Reply
#71
LabRat 891
SpeedyblupiPromontory 19 is B550 and A620A. The functional difference is that the A620 chipset uses a PCIe 4.0 link to the CPU, while A620A/B840 use PCIe 3.0. A620A has fewer lanes enabled than B550, and B840 will presumably have more lanes than A620A (4 PCIe 3.0 + 4 SATA), possibly the same as B550 (10 PCIe 3.0 + 4 SATA or 8 PCIe 3.0 + 6 SATA) or somewhere in between the two.
A620 can be either a cut-down Promontory 21 chip (allowing Prom21 chips with defects to still be sold), or a Promontory 22 chip (which is smaller and cheaper to manufacture).
I just learned earlier today that an 'AM4' chipset (PROM19/B550) found its way to AM5 as the A620A.

Leaves me quite hopeful we might see a PROM21 (or PROM22) AM4 board since, clearly there's no 'restrictions' between platforms. It's just PCI-E linking the 'chipset' to the SoC, after all.
[I *really* want to see a 'new' AM4 board, if only for all the salt it will mine from the "AM4 is ded" crowd]
Posted on Reply
#72
Tek-Check
regsUSB4 superseded USB3. So "USB4" still could mean just USB 3.1 Gen 2 (10 Gbps).
This inital statement of yours was factually incorrect.
USB4 at 10 Gbps is only a mode name of USB4 Gen 2x1, and not "USB 3.1 Gen 2", despite the same speed over one lane.
Both names signify the same speed only, but are coded differently on the electrical layer.
Hence, USB 3.1 Gen2 neither belongs to USB4 specification nor to USB4 product line.
Market name for any product supporting the name 'USB 3.1 Gen2' is SS USB 10 Gbps.
That's why your initial post was dismissed, but you kept going...
regsOperation means duplex - both ways - Gen 2 (minimum of 10+10), Gen 3 (minimum 20+20) and Gen 4 (minimum of 40+40). It doesn't mean that USB4 shouldn't support single link, it doesn't mean it shouldn't support dual link. It's a minimum operation for each Gen.

2.2 USB4 Fabric Architecture
The USB4 Fabric is designed to meet the needs of multiple transport protocols. Its main features are:
Signaling rates that support high throughput interconnects :
- 10 Gbps (for Gen 2), 20 Gbps (for Gen 3), and 40 Gbps (for Gen 4).
Now you are learning finally about USB4.
USB4 spec does support a single link at 10 Gbps, but there is no market name for a USB4 device that would do only USB data at 10 Gbps. Such device or port is called 'SS USB 3.2 Gen 2x1' (old name USB 3.1 Gen2), with logo 'SS 10'. Therefore, "USB4 10 Gbps" device or port do not exist, only a mode of operation within wider data envelope.

Minimal market name for USB4 device or port is 'USB4 20 Gbps', with a logo 'encircled 20'. This is USB4 Gen 2x2. Such device supports both 10 Gbps per single lane and 20 Gbps in total per dual lane, in both directions. And here we are coming to "USB4 10 Gbps" situation whereby USB4 data could operate over one lane while other data, such as PCIe or DP operate over the other lane. If you, however, connect 'SS 10' peripheral, such as flash drive into USB4 host port, the port will not operate as USB4 port, but as USB 3.2 Gen 2x1 port due to different data coding protocol. Therefore, once again, USB4 could NOT mean "just USB 3.1 Gen 2 (10 Gbps)". It never meant that, and it will never do.

Posted on Reply
#73
1969neo
b1k3rdudeSo absolutely no mention of the CPU socket type or what RAM is supported...
Well it's going to be the same since ddr5 r.a.m and the am5 platform haven't been around that long.
Posted on Reply
#74
b1k3rdude
1969neoWell it's going to be the same since ddr5 r.a.m and the am5 platform haven't been around that long.
My question got answered a while ago and back when I asked it, were werent sure if the new CPU's would support AM5.
Posted on Reply
Add your own comment
Nov 23rd, 2024 16:10 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts