Asus X570 Motherboards.

[mackerel]

What would "higher than X570" offer? If we were to assume such a thing existing, I think it could address one area of overlap with HEDT: PCIe expansion potential. Keep the 4.0 x16 from CPU to 1st slot (or split to two 8x with 2nd slot). Move the x4 lanes from the M.2 to also go to the chipset. So a total 8x 4.0 to chipset is a decent chunk of bandwidth. I'm thinking you could offer 8x 3.0 on 3rd slot, 4x 3.0 to 4th slot, with enough left over for an M.2 and a handful of SATA ports. This would be enough for quad crossfire if you were so inclined.

 

[EarthDog]

Or a different split with lanes/bandwidth for pcie 4.0 m.2 modules.

My mistake, ED. Yes, those built-in fans were for the chipset's, not the VRM sections. I should have read the first few posts in this thread more carefully. So, is the higher wattage on the chipsets mainly due to PCI-e 4.0?

Mostly, yes. I hear the fans will only spin up on heavy loads there so there is hope.

 

[Dolk]

I'm actually skeptical about the rumored x590 chipset. I don't think it will be adding really anything except more PCIe lanes, but that doesn't mean it will be worth it. There is only a x4 Gen4 link between the CPU and chipset. On the x570 there is a x12 FlexIO bus that can be configured for SATA or PCIe. All of that information has to go back to the CPU. The PCIe bridge between the CPU and chipset will already act like a bottle neck in some cases, with supposed additional PCIe lanes we may see an even greater bottle neck. However we are talking about constant access to either 10G NIC cards or NVMe drives that would create such a bottle neck. It may not be noticed by most consumers.

 

[mackerel]

Just wondering, is direct PCIe to PCIe transfer a thing? Especially in consumer workloads? Stuff that might not need to go back to CPU and can be routed locally on the chipset? Yes, I'm grasping at straws here.

 

[Dolk]

Absolutely. The Zen 2 CPUs have at min (1) x4 PCIe Gen 4 port for NVMe. Direct attach PCIe NVMe is designed to be directly attached to the CPU or any other PCIe sources. Direct attach to the CPU will of course be the fastest.


PCIe is also a bit of a general purpose bus. In the sense that you can time share what data is being sent. So in the case of the x570 chipset, we have several end devices down the line. Each of the end devices off the chipset will not always be active. If there is collision on terms of communication, PCIe will time slice the data and get it to the CPU eventually. This is overhead. For consumers we are typically fine and will never see this happen. If the rumors are right on the x590, we will have more end PCIe on the chipset and will create a higher chance of this overhead, but its not guarantee.

Local to local is something i'm not aware of with PCIe and the chipset. It all depends on how the whole PCIe tree is structured. the chipset may not have any understanding on how to do this and would cost more to create a chipset to do this rather than a large hub/bridge.

 

[bassnut]

Ok so came across this..... there is no X590 as reported by PC GAMES.....

https://www.pcgamesn.com/amd/x590-chipset-x570-renamed-ryzen-3000

Hidden code that hinted at a potential AMD X590 chipset may have been a red herring. A reference to the high-end board was spotted with X570 BIOS code last week, portent of a new chipset to come. However, motherboard manufacturers now insist the X590 chipset codename is deprecated, and all its functionality instead shifted onto X570 motherboards.

 

[Alaric]

Saw similar info from Tech Jesus yesterday.

 

[Johan45]

I'm actually skeptical about the rumored x590 chipset. I don't think it will be adding really anything except more PCIe lanes, but that doesn't mean it will be worth it. There is only a x4 Gen4 link between the CPU and chipset. On the x570 there is a x12 FlexIO bus that can be configured for SATA or PCIe. All of that information has to go back to the CPU. The PCIe bridge between the CPU and chipset will already act like a bottle neck in some cases, with supposed additional PCIe lanes we may see an even greater bottle neck. However we are talking about constant access to either 10G NIC cards or NVMe drives that would create such a bottle neck. It may not be noticed by most consumers.

I got the straight dope from Gigabyte on that x590 rumour and where it came from. There is NO x590 chipset planned, let's get tht out of the way up front. The BIOS code that got this started comes from the fact that initially AMD was contemplating a two tiered launch because of some difficulties with PCIe 4.0 so there was going to be X570 PCIe3 and X590 PCIe4. The impression I got was that after some pressure from board partners and concerns that the two tiers were just a bad idea AMD dropped the idea but the partners still weren't sure of the naming of the chipset. That's why it was named 570/590 in their BIOS and I'm sure that won't happen again.

 

[bassnut]

I bet you're right about that. I was pretty skeptical about that rumour as it was. Gosh Dang but the time is crawling just 2 mòre weeks that's providing stock is Vailable on the 7th.

 

[Bustos]

I just don’t understand the 550 using pcie 3.0 lanes, at least rumored to be. Why design a 500 series mobo for non 3000 chips?

 

[mackerel]

I haven't seen a side by side comparison of X470 vs X570 features. Apart from PCIe 4.0 support, does the new gen get you much more?

 

[bassnut]

All I can remember reading is better components like the VRM. The top tier boards are supposed to be server grade the lower grade similar to current top tier. I can't remember where I read that ..... might have been one of dolks post.

 

[mackerel]

https://www.gamersnexus.net/guides/3482-amd-x570-vs-x470-x370-chipset-comparison
Finally looked it up. PCIe 4.0 is still the big one, but outside that it seems you get newer USB 3.x. I'm still looking to get X570 for use with Ryzen 3000. For more cost sensitive areas I guess 400 series is still a good option, if there are no other differences. The link above didn't make mention of the was it XFR2 that was added with 400 series, that allowed CPUs to run faster (without OC) than when used with 300 chipsets. I haven't heard anything similar for needing 570 for 3000.

 

[Johan45]

I do know that Gigabyte has put a lot more effort into the 570 series than they have on any other launch with options available across their lineup from day one even iTX. My opinion is most of the manufacturers are giving this launch more attention because of the increased interest in AMD recently. Other than PCIe4 there aren't too many differences, improved power delivery is one there are changes to the XFR PBO which I'm not sure will be ported to the x470 other than that ???

 

[mackerel]

That was my concern when looking at all the mobos being shown off recently. They all look relatively "high end". How much $$$ do we need to throw at it? Are there even "budget" versions? There is quite a chance X570 would be the most expensive mobo I ever got. More even than X299.

Ooh, wait, XFR PBO, is that the "faster" power wake up they mentioned? Or something else? Basically, if running 3000 CPU stock, would you get different clocks in a sustained load between X470 and X570? If not, I might have to give X470 another look. With Ryzen 2000 it required 400 chipset to get a bit more boost out not possible with 300 series.

 

[Bustos]

It wouldn’t make sense, imho, to go half way. If someone wants a 3000 chip then they should be getting an x570 mobo, along with other pcie 4.0 hardware. Otherwise why pay the premium and not use it, like those RTX cards.

This also goes back to my question about the 550 mobos. What does it offer over the 400 series? Guess we’ll find out whenever they hit the shelves.

 

[Johan45]

That was my concern when looking at all the mobos being shown off recently. They all look relatively "high end". How much $$$ do we need to throw at it? Are there even "budget" versions? There is quite a chance X570 would be the most expensive mobo I ever got. More even than X299.

Ooh, wait, XFR PBO, is that the "faster" power wake up they mentioned? Or something else? Basically, if running 3000 CPU stock, would you get different clocks in a sustained load between X470 and X570? If not, I might have to give X470 another look. With Ryzen 2000 it required 400 chipset to get a bit more boost out not possible with 300 series.

PBO= Performance Boost Overdrive
Basically a BIOS setting that easily allows OC up to your cooling limits

 

[mackerel]

PBO= Performance Boost Overdrive
Basically a BIOS setting that easily allows OC up to your cooling limits

I'm getting confused with AMD terminology. What I was thinking of is apparently Precision Boost 2 being available with 400 chipset, not 300 chipsets. This is supposed to give Ryzen 2000 more turbo more of the time. My 2600 does 3.9 all core, would it do that if I put it on an older mobo? So the parallel I had was, if there was a similar feature in 500 that could give it an advantage over 400.

If it is a kinda OC, then it wasn't the power thing I remember being presented. That was claimed faster than ever response between idle and active states.

 

[Alaric]

X570 would likely have a heftier VRM section due to PCIe 4.0. The Ryzen 2 lineup doesn't seem to have any greater power requirements than Ryzen + until you hit the 12c and 16c chips. I would imagine folks spending the cash on the top tier chips will want to spring for the top chipset boards, but I'm hoping the lower tier boards will still be fine for the majority of users.

 

[Johan45]

I'm getting confused with AMD terminology. What I was thinking of is apparently Precision Boost 2 being available with 400 chipset, not 300 chipsets. This is supposed to give Ryzen 2000 more turbo more of the time. My 2600 does 3.9 all core, would it do that if I put it on an older mobo? So the parallel I had was, if there was a similar feature in 500 that could give it an advantage over 400.

If it is a kinda OC, then it wasn't the power thing I remember being presented. That was claimed faster than ever response between idle and active states.

There are a couple of things that AMD worked out with MS for the scheduler and response times. They're not going to make a very noticeable difference in performance overall but some things did show improvement. When Windows was assigning threads it did so without being aware of the CCX or now the chiplets. Take gameplay as an example it's not a noticeable FPS increase but it smooths things out and helps prevent any lags that would be caused from swapping threads between CCX which causes cache flushes. What the end uses would see would be a slight "hiccup" for lack of a better term. This should help eliminate that problem.