• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Will Ryzen 4000 CPUs have more PCIE lanes?

The Asus X570 WS Pro-Ace is wired as follows "Three PCIe 4.0 x16 slots with optimized lane arrangement of 3-way x8/x8/x8 to accelerate an increasingly diverse array of workloads"


From the specifications page:

3rd Gen AMD Ryzen™ Processors
2 x PCIe 4.0 x16 (x16 or dual x8)
2nd Gen AMD Ryzen™ Processors
2 x PCIe 3.0 x16 (x16 or dual x8)
2nd and 1st Gen AMD Ryzen™ with Radeon™ Vega Graphics Processors
1 x PCIe 3.0 x16 (x8 mode)
AMD X570 chipset
1 x PCIe 4.0 x16 (x8 mode)
1 x PCIe 4.0 x1

They're bifurcating the x16 GPU slot and using 8 channels from the PCH.
 
And? I looked at that (p. 1-6) before responding to you.

Completely confused, what are you even saying then? You've contradicted yourself, and backup what I was saying about the board layouts currently available. I'm also still waiting for the RX 5700 being bottlenecked by 16x PCI-E 3.0 revelation.
 
Completely confused, what are you even saying then? You've contradicted yourself, and backup what I was saying about the board layouts currently available. I'm also still waiting for the RX 5700 being bottlenecked by 16x PCI-E 3.0 revelation.

It's the 5500XT that gets bottlenecked if its 4GB of VRAM gets maxed out. I'm presuming PCIE 4.0 allows it to swap out data faster. Hardware Unboxed tested it. It's all a bit academic but it does look interesting in view of what we know about the PS5 and the use of fast storage. Is that why AMD went PCIE 4.0 so early?
 
It's the 5500XT that gets bottlenecked if its 4GB of VRAM gets maxed out. I'm presuming PCIE 4.0 allows it to swap out data faster. Hardware Unboxed tested it. It's all a bit academic but it does look interesting in view of what we know about the PS5 and the use of fast storage. Is that why AMD went PCIE 4.0 so early?

Yes fully aware of that card, hence why I am confused about the mention of the 5700. It's also worth noting that the 5500XT is wired only for 8x PCI-E lanes electrically, which is the sensible approach for cards that don't need the equivalent of PCI-E 3.0 with 32 lanes e.g. a PCI-E 4.0 x16 slot. It also allows the rest of the lanes to be used elsewhere to better effect, which makes a great deal of what the OP wanted possible since it adds 8x lanes to your options.

As for the fast storage comment, I can only speculate that having a fixed hardware design with what will be considered limited resources e.g. VRAM/RAM over the lifespan of the PS5 means that this new developer led approach to using the storage as a RAM buffer in effect will mean that you can start playing games faster. A bit like how you can play some games before they finish downloading fully from the internet, and it is done in the background while you are enjoying yourself. :)

AMD went PCI-E 4.0 because of the advantage it offers in the DC/HPC area, and the timing was right, I'd assume that had Intel managed to get their new silicon sorted at the time it too would have had PCI-E 4.0.
 
It is something I find annoying with CPUs these days - hence why I went with X79/4820K on my last build for the 40 PCI-e lanes.

(I'm currently using 28 for the record).

EDIT: Don't forget in many cases you get an extra 16-20 lanes provided by the PCH or PLX, etc. on many boards though that isn't necessarily a good solution depending on what you want.
Same here with 5960X/X99. Currently using 36.
 
Right now Ryzen 3000 CPUs support 24 PCIE lanes - 16 for the GPU, 4 for a NVME drive, and 4 for everything else. That seems too low for for the enthusiast sector, so I'm wondering if the Ryzen 4000 CPUs will support more? A boost to 36 (16 for the GPU, 8 for two NVME drives, 4 for 10 Gb ethernet, and 8 for everything else) would work wonders. This would still leave Threadripper as the premium HEDT product with 88 (never mind Epyc's 128 lanes). This would also allow better product differentiation with the B650 boards supporting the basic 24 lanes and the X670 boards enabling 36.

I see where you're going with this and i agree with you. Whether Zen3 will support this or not remains to be seen but AMD have left a pretty big segmentation gap WRT usable PICe lanes on AM4 so they will have a market if there is an option. Of course they'll charge for it.

Despite all the other (fixed) benefits, AM4 has a woeful amount of flexible PCIe lanes and when taken in isolation the X-series chipsets are a backwards step from the 990FXA chipset in this regard. All slots/connectors can't be used all the time and anyone who reads motherboards manuals will know they are chock full of caveats and 'if, then' type language because not all slots/connectors can be used all at the same time and it's mind-boggling. This is not enthusiast-class stuff, even on the desktop.

PCIe 4 bandwidth is not a substitute for lanes in all cases yet this is what is being argued, a lot. Sharing lanes and compromising other add-in board performance is not a substitute for lanes and this is also being argued, a lot. What is this Stockholm syndrome of arguing for being short-changed? Trade-offs are not benefits.

Since AMD are forcing everyone upwards in price i'm done working around the X-series PICe limitations and cheap TRX40 is my next step, which i will hold onto longer as it costs more. I'm in the target market you're talking about here.
 
Since AMD are forcing everyone upwards in price i'm done working around the X-series PICe limitations and cheap TRX40 is my next step, which i will hold onto longer as it costs more. I'm in the target market you're talking about here.

What devices do you currently use in the system you have, or what are you wanting to put in it? I love the TR4/TRX40 platform for the sheer flexibility it offers, even in the TR4 guise before the introduction of PCI-E 4.0 it was a revelation for WS builds, and not with 4.0 it's like they've done it all over again, if you have the need for the bandwidth not just the lanes.
 
In my OP I suggest that this be a differentiator between B650 and X670 motherboards.

It doesn't work like that. The CPU would have to be designed for 36 or however many lanes. It wouldn't make sense for different boards with the same socket to have access to more or less of those lanes. Or at the least, it wouldn't be good as a consumer to have that artificial limitation.
And you didn't mention quad channel ram.

What I'd want is a TRX4 equivalent to the £400ish 1900 and some cheaper entry level boards.
Maybe Intel if they could use a ring bus on lower CC CPUs and not like the just fail gimped Kaby X CPUs.
 
Or at the least, it wouldn't be good as a consumer to have that artificial limitation.

Indeed, look at the cluster bomb that was the original X299 launch with the i5 7640, and i7 7740, and so on. You couldn't use half the features of the boards, unless you paid more for the CPU that had them, but it meant all the board were more expensive. Great if you want something to grow into, and I though it would have worked but apparently it was an utter flop hence it no longer exists.

The issue we really have is there is too few devices currently that have been moved to take advantage of PCI-E 4.0, unless you are in the enterprise space with nice shiny 200Gb fibre cards. I still can't find a dual 10GbE card with a 4x PCI-E 4.0 interface, so I am forced to use the 8x cards from Intel which run at 3.0, meaning I lose lanes that could be used elsewhere, and this is why I love TR4/TRX40, you have spare capacity, but don't think the AM4 platform needs this at all, just more boards with a better layout that people can grow into as devices become available.
 
Indeed, look at the cluster bomb that was the original X299 launch with the i5 7640, and i7 7740, and so on. You couldn't use half the features of the boards, unless you paid more for the CPU that had them, but it meant all the board were more expensive. Great if you want something to grow into, and I though it would have worked but apparently it was an utter flop hence it no longer exists.

The issue we really have is there is too few devices currently that have been moved to take advantage of PCI-E 4.0, unless you are in the enterprise space with nice shiny 200Gb fibre cards. I still can't find a dual 10GbE card with a 4x PCI-E 4.0 interface, so I am forced to use the 8x cards from Intel which run at 3.0, meaning I lose lanes that could be used elsewhere, and this is why I love TR4/TRX40, you have spare capacity, but don't think the AM4 platform needs this at all, just more boards with a better layout that people can grow into as devices become available.

Yep, exactly. They were just the mainstream 1151 CPUs modified to fit X299. Still had only 16 lanes, still could only use dual channel ram. Upgrading from them on X299 was a bad idea as well for gaming or mixed usage because the higher CC CPUs used mesh rather than the ring bus.
 
It doesn't work like that. The CPU would have to be designed for 36 or however many lanes. It wouldn't make sense for different boards with the same socket to have access to more or less of those lanes. Or at the least, it wouldn't be good as a consumer to have that artificial limitation.
And you didn't mention quad channel ram.

What I'd want is a TRX4 equivalent to the £400ish 1900 and some cheaper entry level boards.
Maybe Intel if they could use a ring bus on lower CC CPUs and not like the just fail gimped Kaby X CPUs.

I'm with you, I didn't need TR CPU but wanted lanes for shifting data about nice and quick and because really big NVMe were silly money, not only that NVMes are cool :D so I had lots of 1 and 2TB ones that were cost effective, I would have upgraded to TRX40 if the 3950x was on it but its not so I sit here waiting for something to fill the gap much like I did when I moved from z77, couldn't find a board to run all my gear due to lanes back then as no one was doing PLX anymore after Avago hiked prices and did not want to buy another board that would go out of date no sooner than I started using it, then TR came alone, I haz all the lanes :D TR3 too pricey for me.

I have been considering x570 with the Asus WS and a x8 4 NVMe adaptor but its ~£300 addition and has some performance limitations, I will wait for TR4 and a price drop on TR3 :D TRX40 boards are actually real good value, just need a cheaper chip.
 
Indeed, look at the cluster bomb that was the original X299 launch with the i5 7640, and i7 7740, and so on. You couldn't use half the features of the boards, unless you paid more for the CPU that had them, but it meant all the board were more expensive. Great if you want something to grow into, and I though it would have worked but apparently it was an utter flop hence it no longer exists.

The issue we really have is there is too few devices currently that have been moved to take advantage of PCI-E 4.0, unless you are in the enterprise space with nice shiny 200Gb fibre cards. I still can't find a dual 10GbE card with a 4x PCI-E 4.0 interface, so I am forced to use the 8x cards from Intel which run at 3.0, meaning I lose lanes that could be used elsewhere, and this is why I love TR4/TRX40, you have spare capacity, but don't think the AM4 platform needs this at all, just more boards with a better layout that people can grow into as devices become available.

Yeah I miss a X79 like platform for this generation - TR is the closest I guess when you balance all the factors but a foot in the door is far more expensive. I'd be far more likely to have upgraded by now otherwise though I don't really feel the need for anything faster just yet - most games still run great on an overclocked 4820K and other stuff I either spread over 2x i7 systems or just chuck things like video encoding on the 2nd system and forget about it until done.
 
What devices do you currently use in the system you have, or what are you wanting to put in it? I love the TR4/TRX40 platform for the sheer flexibility it offers, even in the TR4 guise before the introduction of PCI-E 4.0 it was a revelation for WS builds, and not with 4.0 it's like they've done it all over again, if you have the need for the bandwidth not just the lanes.

Most of the PCIe add-in cards are PCIe3.0 or lower and they require lanes to work. Bandwidth or divisions of bandwidth on PCIe4.0 make no differences to whether these cards work correctly or not.

This is where i am at with my cards and anything greater than a x1 has to go into the secondary GPU slot that then knocks the GPU down to x8. Unless i forego my games NVME drive and use the bottom PCIe slot instead. I have already substituted HDD into where NVME have replaced SSD so i'd need to effectively go backwards to use the PCIe slots. This still leaves 'stuff' provided on the motherboard unused as they are unusable, not because i chose to not use them.

Instead of me being able to adopt new hardware and use it i'm being forced to substitute it with my other, established hardware. This is not progress. Add me an NVME slot for 'free'? Why, thank you very much and have my hard earned pounds. There has to be a value equation at work and i'm struggling to see it with motherboards at the moment. I'm not sure where the benefit is to me that currently i have to endure trade-offs and pay for the privilege.

I agree with you TRX40 is extremely flexible and it's why i am looking at it. It will be overkill for what i want though and there is far too big a gap in terms of PCIe between AM4 and TRX40. Maybe this is why AMD are moving all their chipsets upwards and forgetting that it would be a good idea to re-name them to demonstrate. They still haven't got a x16/x16 GPU even on x570. A x16/x16 that had the option of x16/x8/x8 would be great (doesn't affect primary GPU bandwidth) and of course i already mentioned NVME drives consuming the bottom slot on most motherboards, give us that too AMD.

B550 is clearly an X470 replacement and not a B450 replacement. Read between the lines of the latest 'reviews' and the techtubers are clearly telling us this (covertly, without breaking NDA) and we're probably being softened up for A520 being a B450 replacement. The tech press is probably under the same NDA for A520 that they were for B550 (from the 3100/3300 launch) and know what's coming. I'm sure ScottieB knows as well but he's being cute about it in the B550 thread, although he's getting tetchy about it. AMD messed their strategy and marketing up bad this time round.
 
It doesn't work like that. The CPU would have to be designed for 36 or however many lanes.

Yes and no with AMD CPUs....a change to the IO die is required now and it would probably have been a good idea to do just that when the IO die was introduced with Zen2 as that was a big design change anyway. This might have been one of the improvements AMD keeps talking about that don't make it into some CPUs so they can keep us on the upgrade hamster-wheel.

AMD already has form about re-designing in PCIe uplift; the Renoir APUs will have x16 'GPU' lanes this time round instead of the shonky x8 found on Raven Ridge and Picasso so it's not inconceivable they could do it any time they wanted.

It would have made X570 a more compelling upgrade if the software lockouts (and most of the board/chipset differentiators are just AGESA level software lockouts) were employed to have boards with earlier chipsets than x570 run at x8/x8 while X570 runs at x16/16. Earlier boards wouldn't run at x16/x16 anyway as the second slot is x16 physically but x8 electrically so that really is a reasonable technical reason. I probably would have jumped on that myself.

There's loads of options. the 3950X could have had x16/x16 but all others not. Then the XT versions could have been released as now having x16/x16 as a reason to buy. Or, the 3950X/3900X/3800X/3700X could all have been x16/x16 with non-x versions being x8/x8. That is really getting into differentiation right there and pricing tiers overtly justified by features.

Either way, AMD fumbled as it's a regression from previous AMD platforms not having x16/x16 on the desktop. As i said before, the other PCIe lanes available are not a replacement as they have specific uses, are conditional, and in case of the GPU, degrade performance.
 
Instead of me being able to adopt new hardware and use it i'm being forced to substitute it with my other, established hardware. This is not progress. Add me an NVME slot for 'free'? Why, thank you very much and have my hard earned pounds. There has to be a value equation at work and i'm struggling to see it with motherboards at the moment. I'm not sure where the benefit is to me that currently i have to endure trade-offs and pay for the privilege.

I get where you are coming from, but since you've not said what you are using that forces these artificial limitations then it's hard to fully understand.

I used to use the Asus X99-E-10G WS for a client build/design that called for a huge number of expansion cards, and it wasn't ideal as it utilised PLX chips to get the 7x 16x PCI-E slots, but had dual on-board 10GbE NIC's which was great. Then bam, Threadripper came along and made the 40 lanes native to the CPU on X99 look amateurish. I still struggled as at least two of the 16x PCI-E slots we filled with 4x NVMe -> 16x adapter cards using bifurcation, but still needed 32x lanes for 2 GPU's that pushed the design up to 64x lanes which meant that one of the GPU's had to be run at 8x if we wanted to achieve the full potential of the NVMe drives. With PCI-E 4.0 that is now gone, as the extra bandwidth it offers to the disk subsystem is immense, and we are currently doing a re-design with Samsung 980 Pro's in place, which mean we can drop the number of drives by 50% while keeping the capacity and bandwidth, and allowing us to add a further GPU in to the mix, or a 40Gbps/100Gbps fibre card, depending on where the system will be used.

Changing the I/O die sounds simple but at the launch of Zen2 the I/O die's are fabbed on 12/14nm so adding more PCI-E lanes would add quite a chunk of extra space which may have limited the availability of keeping the dual die options for the CPU's, and also added extra overhead for the power draw if running the I/O flat out.

Saying they fumbled is a bit harsh given the fact they came from a platform which might have had great boards but the CPU's were terrible, and lets not forget it was PCI-E 2.0 so your argument about the GPU's being held back would stand on that platform too.
 
I get where you are coming from, but since you've not said what you are using that forces these artificial limitations then it's hard to fully understand.

It shouldn't be hard to understand. I need more lanes, without compromise, than is available from the current 'enthusiast' AM4 platform. I've said this more than once.

In reality it doesn't matter what i'm using and it's not me forcing the artificial limitations....the motherboards are. The PCIe bandwidths are not a substitute for lanes in all circumstances although people seem set on constantly conflating them. Your SM/enterprise-class example clearly shows that you gained benefit from PCIe4 because it unlocked the bandwidth you required. This isn't my use case. I require lanes and not bandwidth and PCIe4.0 will not change this.

I mentioned to you i have other >x1 cards and the reasons why it is not optimal to use. I'll humour you by letting you know that, among other things, i use a Perc H710 and as i alluded to i compromise my GPU by using it.

I used to use the Asus X99-E-10G WS for a client build/design that called for a huge number of expansion cards, and it wasn't ideal as it utilised PLX chips to get the 7x 16x PCI-E slots, but had dual on-board 10GbE NIC's which was great. Then bam, Threadripper came along and made the 40 lanes native to the CPU on X99 look amateurish. I still struggled as at least two of the 16x PCI-E slots we filled with 4x NVMe -> 16x adapter cards using bifurcation, but still needed 32x lanes for 2 GPU's that pushed the design up to 64x lanes which meant that one of the GPU's had to be run at 8x if we wanted to achieve the full potential of the NVMe drives. With PCI-E 4.0 that is now gone, as the extra bandwidth it offers to the disk subsystem is immense, and we are currently doing a re-design with Samsung 980 Pro's in place, which mean we can drop the number of drives by 50% while keeping the capacity and bandwidth, and allowing us to add a further GPU in to the mix, or a 40Gbps/100Gbps fibre card, depending on where the system will be used.

You had a bandwidth deficit here. I don't. This is an apples/oranges comparison so while it was interesting for me to read, offered nothing to my situation.

Changing the I/O die sounds simple but at the launch of Zen2 the I/O die's are fabbed on 12/14nm so adding more PCI-E lanes would add quite a chunk of extra space which may have limited the availability of keeping the dual die options for the CPU's, and also added extra overhead for the power draw if running the I/O flat out.

AMD designed a new IO chip, of course they could have added another 8 lanes while they were at it. Unless you have first-hand knowledge of what it took to design the IO die and why 8 lanes couldn't be added i'll file this statement under 'speculation'. What still remains is the fact they missed an opportunity to drive differentiation.

Saying they fumbled is a bit harsh given the fact they came from a platform which might have had great boards but the CPU's were terrible, and lets not forget it was PCI-E 2.0 so your argument about the GPU's being held back would stand on that platform too.

Harsh but true. Again, total bandwidth is not a substitute for available lanes (and vice versa). I also didn't lament having 990FXA for AM4, i lamented the flexibility of the large lane allocation. AMD could easily have done that while implementing PCIe3.0. It's a bit of a nonsense to bring PCIe2.0 up as if that's what i wanted.

At this point, PCIe4.0 is not a reason to buy for most people. Should AMD have brought out PCIE4.0 and more lanes than X370/X470 then i'd have bought, even at PCIe3.0. Not for the bandwidth, for the lanes. If AMD had brought out X470 but with an uplift from X370 in PCIe lanes then i'd have bought (where i didn't). See where i'm going with this? I don't think i'm unique in this regard.
 
Back
Top Bottom