AMD 300-series and 400-series Motherboards to Lack PCIe Gen 4 with Ryzen 3000

[Midland Dog]

only reason for pcie 4 on consumer is so you can use twice as many pcie 3.0 devices as full speed, no point giving gpus more pcie bandwidth if they dont need it, see gn and there titan v test

 

[lexluthermiester]

see gn and there titan v test

Titan V is not a consumer GPU.

 

[Vya Domus]

Titan V is not a consumer GPU.

I don't know the context of your reply but the Titan V is certainly a consumer product, anyone can buy one straight from Nvidia, requires no special licensing or anything like that. I think what you mean is that it's not a mainstream GPU.

 

[lexluthermiester]

I don't know the context of your reply but the Titan V is certainly a consumer product

The context of my comment is not difficult to understand and your point is incorrect. NVidia themselves have stated that the Titan V is not a gaming card. It is meant for professional work environments, not the consumer market as reflected by the hefty price tag.

anyone can buy one straight from Nvidia, requires no special licensing or anything like that. I think what you mean is that it's not a mainstream GPU.

Most professional products can be purchased by general consumers without "special licensing". That point was pointless.

 

[Sora]

Jokes on Robert, there are already test bios's running the 1st pci-e slot at 16GT/s

Could someone remind us of a single case, when old mainboards would start to support faster PCIe spec, just because some CPU manufacturer released a new CPU?

Sandy E (Gen 2 compliant) to Ivy E (Gen 3 compliant)

motherboards that were overbuilt for 5GT/s but received 8GT/s certification and Gen3 compliancy tested when IB-E came out.

 

[Metroid]

That's always the case with high end.

Yes. However, there were times where your just purchased high end cpu lasted a long time at the top, for example my old i7 920, 4 cores 8 threads, mainstream at 2.6ghz, overclocked to 4.5ghz for $284 in November 2008 and was only surpassed in core count and justified performance wise with the 8700 $320 in October 2017. That was almost 9 years. Well, we had that 1 to 5% performance every process node, funny thing, at that time you could achieve 50% overclock, nowadays trolls are happy if their amazing $500 motherboard can sustain their 100 - 300 mhz increase, 4.9 to 5.2ghz or so for the 9700k, and 100mhz for the 9900k, 5 to 5.1ghz, so overclock only 1 to 2% hehe, you may say that happened because we hit a peak, I say okay if we hit a peak so where are the cores? Intel took 10 years plus and heavy competition x ryzen when it was released to start offering more cores for the mainstream users or even high end for a good price.

See the thing is, I really expected amd to deliver a 16 cores right now and break the paradigm we have had for 10 years, unfortunately they have not, okay they helped to break some of it but not as somebody that want to lead the industry to new heights. AMD wants to milk in just like intel have done to us in the past 10 years or so, okay amd might launch a 16 cores next year but for now are just rumors and even if they launch intel will have had time to counter it with same core count and amd in the end wont have made an impact like intel did with the i7 920 in 2008. I think the i7 920 was the best cpu purchase of all time to date and for just $284.

Do you know when we see everything is F***ed at intel? is when Intel shows that their i7 is not high end anymore. The i7 at the top lasted almost 10 years and it started with the i7 in November 2008. The creation of an i9 series shows that intel's reign is over but does not not mean amd is the new king of the hill, it shows there is a balance at moment.

List of Intel Core i7 processors - Wikipedia

en.wikipedia.org

8W is nothing and can easily be cooled passively or not even need a heatsink at all.

I found this article today 04 june, at anandtech, related to your post.

GIGABYTE’s PCIe 4.0 SSD Uses 77g of Copper Due to 8W TDP, up to 2TB

www.anandtech.com

 

[Keullo-e]

For graphics cards, I think that this isn't a major problem.

 

[kapone32]

This reminds of the change from PCI-E 2.0 to 3.0 I had a 990FX Sabretooth that had PCI-E 3.0 slots. DId I notice a difference in GPU performance vs my 890GX board that was only 2.0 nope. As stated elsewhere the only tangible benefit for PCI-E 4.0 is NVME storage. I looked at the chipset picture for X570 and the chipset could support 2 PCI-Ex16 3.0 from 1 PCI-E 4.0 16. I did have a look at some boards on Vendor websites and I find that the boards I saw (As RocK) are wired the same as X470 x16 ,8 ,4. Perhaps As ROck's top MB with the water cooling block may be wired different. I really don't understand if the CPU has 40 PCI-E lanes why we do not have fully wired x16 slots on X570 other than just 1. I don't need to have 20 USB 3.1,3.2 ports on my machine.

 

[Flaky]

This reminds of the change from PCI-E 2.0 to 3.0 I had a 990FX Sabretooth that had PCI-E 3.0 slots

Kind of bad example, since AM3+ didn't have 3.0. The gimmick on this specific motherboard based on PLX chip, but yet the most important link - between north bridge and PLX - still was running at PCIe 2.0.

I looked at the chipset picture for X570 and the chipset could support 2 PCI-Ex16 3.0 from 1 PCI-E 4.0 16.

Whatever you did see, you did interpret it wrong.
First of all, chipset does not provide x16 links. That wouldn't even make sense with narrower uplink link.
Second, lanes do not magically multiply, just because certain configurations have equivalent speed.
PCIe 3.0 x16 GPU in a 4.0 x8 slot will work at 3.0 x8.
PCIe 4.0 x2 NVMe drive put into 3.0 x4 slot will work at 3.0 x2.

I really don't understand if the CPU has 40 PCI-E lanes

CPU does not have 40 lanes. You're looking at a sum of both cpu's and chipset's lanes.
Check again what puts out what, and the reason why even the high end motherboards come up in "x16 ,8 ,4" setting will be clear

 

[kapone32]

The chipset supports 3 PCI-E 4.0 slots x4 slots.

Kind of bad example, since AM3+ didn't have 3.0. The gimmick on this specific motherboard based on PLX chip, but yet the most important link - between north bridge and PLX - still was running at PCIe 2.0.


Whatever you did see, you did interpret it wrong.
First of all, chipset does not provide x16 links. That wouldn't even make sense with narrower uplink link.
Second, lanes do not magically multiply, just because certain configurations have equivalent speed.
PCIe 3.0 x16 GPU in a 4.0 x8 slot will work at 3.0 x8.
PCIe 4.0 x2 NVMe drive put into 3.0 x4 slot will work at 3.0 x2.

I am not saying that at all I was using the theorertical bandwidth of PCI_E 4.0 64 GB/s looking at what the the PCI _E support is on the As rock X570

AMD Ryzen series CPUs (Matisse)
- 3 x PCI Express 4.0 x16 Slots (PCIE1/PCIE3/PCIE5: single at x16 (PCIE1); dual at x8 (PCIE1) / x8 (PCIE3); triple at x8 (PCIE1) / x8 (PCIE3) / x4 (PCIE5))*
AMD Ryzen series CPUs (Pinnacle Ridge)
- 3 x PCI Express x16 Slots (PCIE1/PCIE3/PCIE5: single at Gen3x16 (PCIE1); dual at Gen3x8 (PCIE1) / Gen3x8 (PCIE3); triple at Gen3x8 (PCIE1) / Gen3x8 (PCIE3) / Gen4x4 (PCIE5))*
AMD Ryzen series CPUs (Raven Ridge) (Picasso)
- 1 x PCI Express 3.0 x16 Slot (single at x8 (PCIE1))*
- 1 x PCI Express 4.0 x16 Slot (single at x4 (PCIE5))*



CPU does not have 40 lanes. You're looking at a sum of both cpu's and chipset's lanes.
Check again what puts out what, and the reason why even the high end motherboards come up in "x16 ,8 ,4" setting will be clear

I am sorry for misquoting I did mean to say platform.

 

[Midland Dog]

Titan V is not a consumer GPU.

and? it just proves the point more

 

[lexluthermiester]

and? it just proves the point more

Oh? Please explain your logic..

 

[Midland Dog]

Oh? Please explain your logic..

if the best compute sillicon in the world cant max 3.0 16x nothing can (at this point)