Pcie x16 problems

[-rascal-]

14 minutes ago, OutsaneIngenuity said:

So someone else just notified me that the 4090 is gen 4 and x8 gen5 is the same as x16 Gen 4..... makes me sad asf and now I have to re-question my build. If that is the case is there even anything to take advantage of x16 Gen 5? 

 

Yes, the RTX 4000 series is on Gen4 still.

However, available bandwidth from the slot... is not what you think.

 

Let's say PCI-E X16 5.0 is the number of available lanes on a freeway -- X16 Gen5 = 16-lanes, 150 MPH.

You only have 8 lanes of 120 MPH worth of car traffic.

It doesn't cause a traffic jam.

It's when you EXCEED what the lanes can handle, is where you run into performance issues.

 

The RTX 4090 is running as fast as it can. It's not being gimped by the bandwidth on the PCI-E slot.

 

That is not a surprise. It's to give more "future-proofing".

They don't need to come up with a PCI-E Gen6 spec whenever a new RTX 5090 or 6090 comes out.

Why TF would you design something that's "fast enough" for a...RTX 4090 or 7900XTX, and then dump another like $XXX million again a year later?

[RONOTHAN##]

24 minutes ago, -rascal- said:

How is the Apex / Encore CHEAPER than the Hero these days.

No clue. It started back in 9th gen IIRC where the Apex was something like $20 more than the Hero (though availability was sketchy enough to where it wasn't super relevant regardless). 

 

Most of ASUS's ROG lineup can be eliminated since the only board there worth even considering is the Apex, all the other Maximus boards make no sense in comparison, and the Strix boards are usually overpriced compared to the competition from other vendors. 

[AlfaProto]

11 hours ago, OutsaneIngenuity said:

So someone else just notified me that the 4090 is gen 4 and x8 gen5 is the same as x16 Gen 4..... makes me sad asf and now I have to re-question my build. If that is the case is there even anything to take advantage of x16 Gen 5? 

Except, your 4090 is a PCIE gen 4, it will run at gen 4 speed & specs.

 

Both Intel and AMD offer PCIE gen 5 just for future proofing.

 

AFAIK, might more more expensive NIC, or those PCIE 5.0 'dummy' 4xM.2 SSD.

[AlfaProto]

11 hours ago, -rascal- said:

 

I'm just so confused with ASUS' line-up of "ROG" boards these days, and the insanely high price premium for "ASUS" and "ROG".

They were not like this prior to.... 2017?

 

From top -> bottom

• Apex / Encore
• Extreme
• Formula
• Hero
• Impact (ITX board)
• StriX

The -Hero tier was the "budget" ROG line-up, if you didn't want the "High OC" boards.

How is the Apex / Encore CHEAPER than the Hero these days....

ASUS wrecked their ROG product line-up.

 

Strix should be higher than Hero, while Hero is the bottom.

 

I remember back when GeForce 900 series era, ASUS stated that Hero is the most budget line of ROG, close to the 'more budget' TUF variant. Hero is super bare, it's like buying ROG brand only, not the features. They even stated that Strix is the higher version of Hero, more bells and whistles.

 

I forgot what Rampage is, but I'm guessing it's the middle of ROG?

 

I forgot which is which, it's either Formula is OCing or Extreme is for custom WC.

 

There's also Maximus too.

 

Impact and Gene are own its own, they are ITX and mATX.

[-rascal-]

1 hour ago, AlfaProto said:

ASUS wrecked their ROG product line-up.

 

Strix should be higher than Hero, while Hero is the bottom.

 

I remember back when GeForce 900 series era, ASUS stated that Hero is the most budget line of ROG, close to the 'more budget' TUF variant. Hero is super bare, it's like buying ROG brand only, not the features. They even stated that Strix is the higher version of Hero, more bells and whistles.

 

I forgot what Rampage is, but I'm guessing it's the middle of ROG?

 

I forgot which is which, it's either Formula is OCing or Extreme is for custom WC.

 

There's also Maximus too.

 

Impact and Gene are own its own, they are ITX and mATX.

 

Totally forgot about the mATX Gene.

 

Crosshair, Maximus, Rampage were the different platforms.

Crosshair was for AMD's mainstream platform, while Maximus was for Intel's mainstream platform (i.e. Z790) and Rampage was Intel HEDT (i.e. X58, X79, X99).

You can have combinations of Crosshair/Maximus/Rampage with TUF/Strix/Hero/Formula/Extreme/Apex, etc, etc.

 

Rampge VI Extreme was LGA 2066 / X299.

 

Strix owl, Cerberus hound, and then TUF Sabertooth...is now just TUF.

 

[QuantumSingularity]

On 1/20/2024 at 12:15 AM, OutsaneIngenuity said:

So someone else just notified me that the 4090 is gen 4 and x8 gen5 is the same as x16 Gen 4..... makes me sad asf and now I have to re-question my build. If that is the case is there even anything to take advantage of x16 Gen 5? 

The nature of the data flow is such, that a GPU is rarely held down by the interface connecting it to the motherboard. Even with 1000W VBIOS 4090 and overclocked to the moon and back it very very rarely hits the limit of PCIe Gen 3, leading to about 1%-2% performance difference compared to PCIe Gen 4, which again - x16 Gen 3 = x8 Gen 4 = x4 Gen 5. If you compare components based on their data throughput, a lot of stuff will be faster than a 4090. Its job simply isn't transferring data as fast as possible, its computing and working with that data as fast as possible. Marketing strategies aside, we won't need PCIe Gen 5 on GPUs for at least 2 more generation, but to make it sound better and "futureproof" Nvidia most likely will launch their RTX 5000 series GPU with PCIe Gen 5 support, although it will go totally wasted. At least you would be able to stuck the GPU in the bottom x16 slot of your motherboard where it won't have any issues running in x8 config, with the main slot populated a NVMe expansion card and 4 SSDs.

[AlfaProto]

3 hours ago, -rascal- said:

 

Totally forgot about the mATX Gene.

 

Crosshair, Maximus, Rampage were the different platforms.

Crosshair was for AMD's mainstream platform, while Maximus was for Intel's mainstream platform (i.e. Z790) and Rampage was Intel HEDT (i.e. X58, X79, X99).

You can have combinations of Crosshair/Maximus/Rampage with TUF/Strix/Hero/Formula/Extreme/Apex, etc, etc.

 

Rampge VI Extreme was LGA 2066 / X299.

 

Strix owl, Cerberus hound, and then TUF Sabertooth...is now just TUF.

 

Can all board vendors return back to LGA1155 to 1151 naming scheme and feature allocation?

All I know is that creators board are: Gigabyte Aero, ASUS ProArt, MSI Creation, Asrock Creator. Those can stay.

 

Also entry/decent line-up: Gigabyte Ultra Durable (UD), MSI Pro, ASUS Prime and Asrock Extreme

I guess the gaming line-up are completely fk, and strayed away from their original intent.

[Queen Chrysalis]

On 1/19/2024 at 3:08 PM, OutsaneIngenuity said:

Im confident there will be occasions I push it towards its limit. And hey, it's cheaper than building a car. Gotta be efficient with your dopamine per dollar ratio and not all of us won the lottery on our allotted categories. That aside it comes with the sweet benefit of when I get blamed for a "hardware" issue I can list my specs and tell them to pound sand! (This happens more than you'd think and every time the issue has not been on my end)

I promise you that won't happen, even if you spend less than half of this.  And it can still happen even if you spend all of this.  You don't buy reliability with PC parts, that's a car thing.  You buy features.  Reliability is 1) way higher, 2) a bit of a crapshoot anyway.  You are almost completely paying for features you'll never use.  Efficient dopamine per dollar is like a $1500 soft cap, and IMO a $2000 hard cap for gaming.

[OutsaneIngenuity]

On 1/19/2024 at 3:18 PM, -rascal- said:

 

I'm just so confused with ASUS' line-up of "ROG" boards these days, and the insanely high price premium for "ASUS" and "ROG".

They were not like this prior to.... 2017?

 

From top -> bottom

• Apex / Encore
• Extreme
• Formula
• Hero
• Impact (ITX board)
• StriX

The -Hero tier was the "budget" ROG line-up, if you didn't want the "High OC" boards.

How is the Apex / Encore CHEAPER than the Hero these days....

Supply/demand?

[OutsaneIngenuity]

On 1/20/2024 at 8:27 AM, AlfaProto said:

Can all board vendors return back to LGA1155 to 1151 naming scheme and feature allocation?

All I know is that creators board are: Gigabyte Aero, ASUS ProArt, MSI Creation, Asrock Creator. Those can stay.

 

Also entry/decent line-up: Gigabyte Ultra Durable (UD), MSI Pro, ASUS Prime and Asrock Extreme

I guess the gaming line-up are completely fk, and strayed away from their original intent.

Yea I've noticed that "gaming motherboards" are usually just marking bs and best an ignored term

[OutsaneIngenuity]

On 1/20/2024 at 8:00 AM, QuantumSingularity said:

The nature of the data flow is such, that a GPU is rarely held down by the interface connecting it to the motherboard. Even with 1000W VBIOS 4090 and overclocked to the moon and back it very very rarely hits the limit of PCIe Gen 3, leading to about 1%-2% performance difference compared to PCIe Gen 4, which again - x16 Gen 3 = x8 Gen 4 = x4 Gen 5. If you compare components based on their data throughput, a lot of stuff will be faster than a 4090. Its job simply isn't transferring data as fast as possible, its computing and working with that data as fast as possible. Marketing strategies aside, we won't need PCIe Gen 5 on GPUs for at least 2 more generation, but to make it sound better and "futureproof" Nvidia most likely will launch their RTX 5000 series GPU with PCIe Gen 5 support, although it will go totally wasted. At least you would be able to stuck the GPU in the bottom x16 slot of your motherboard where it won't have any issues running in x8 config, with the main slot populated a NVMe expansion card and 4 SSDs.

So if that's the case there's no reason to avoid putting a m.2 in the 1st slot and knocking the first pcie down to x8 then?

[QuantumSingularity]

Yup. With Gen 5 speeds you can drop down to x4 and still the GPU would barely feel the difference. But really, if PCIe lanes compatibility and availability is you main concern, you should turn to AMD and X670E.

[brob]

36 minutes ago, QuantumSingularity said:

Yup. With Gen 5 speeds you can drop down to x4 and still the GPU would barely feel the difference. But really, if PCIe lanes compatibility and availability is you main concern, you should turn to AMD and X670E.

 

As long as the GPU was PCIe 5.0. I don't believe there are any such models.

[-rascal-]

It's just because the current-generation Intel CPUs only provide:

• 16x PCI-E 5.0 lanes (i.e. graphics card)
• 4x PCI-E 4.0 lanes (i.e. NVMe M.2 SSD).
• 8x PCI-E 4.0 for communication between CPU and Motherboard chipset

If you want to make use of a PCI-E 5.0 M.2 slot, the only option it to split the PCI-E 5.0 lanes....

 

AMD went with:

• 16x PCI-E 5.0 lanes (i.e. graphics card)
• 8x PCI-E 5.0 lanes (i.e. NVMe M.2 SSD, up to 2x SSDs at full 4x speed each) 
• 4x PCI-E 5.0 for communication between CPU and Motherboard chipset

[AlfaProto]

On 1/23/2024 at 7:29 AM, -rascal- said:

AMD went with:

• 16x PCI-E 5.0 lanes (i.e. graphics card)
• 8x PCI-E 5.0 lanes (i.e. NVMe M.2 SSD, up to 2x SSDs at full 4x speed each) 
• 4x PCI-E 5.0 for communication between CPU and Motherboard chipset

That's straight up false. I'm using Gigabyte block diagram, since they are the only ones that provide useful mobo block diagram:

X670E (X670E Aorus Pro X):

X670 (X670 Aorus Elite AX):

B650E (the most useless chipset from AMD, B650E Aorus Elite X AX Ice) :

Unfortunately, A620 doesn't provide block diagram, since it's pretty simple, straightforward, and it's the more useful chipset over the B-series.

 

I wonder where you got your:

 

8x PCI-E 5.0 lanes.

 

It has no dedicated PCIE 5.0 lanes for SSD.

[-rascal-]

7 hours ago, AlfaProto said:

That's straight up false. I'm using Gigabyte block diagram, since they are the only ones that provide useful mobo block diagram:

X670E (X670E Aorus Pro X):

 

 

Gigabyte's block diagram supports what I am saying, so does their details specification.

 

Look at AMD's specification sheet for a AMD Ryzen 7000-series CPU, and the block diagram AMD released.

For an example, the Ryzen 7800X has 28 PCI-E 5.0 lanes supplied by the CPU, 24 are usable, while the remaining 4 is allocated for the communication bus between the CPU and Chipset.

 

From the X670E AX motherboard block diagram.

 

However, how to split, distribute, manipulate, and wire the PCI-E lanes, it is up to the board manufacturer.

On the CPU side, and how AMD designed the I/O, it provides 28 PCI-E 5.0 lanes.

[AlfaProto]

17 hours ago, -rascal- said:

 

Gigabyte's block diagram supports what I am saying, so does their details specification.

 

Look at AMD's specification sheet for a AMD Ryzen 7000-series CPU, and the block diagram AMD released.

For an example, the Ryzen 7800X has 28 PCI-E 5.0 lanes supplied by the CPU, 24 are usable, while the remaining 4 is allocated for the communication bus between the CPU and Chipset.

 

From the X670E AX motherboard block diagram.

 

However, how to split, distribute, manipulate, and wire the PCI-E lanes, it is up to the board manufacturer.

On the CPU side, and how AMD designed the I/O, it provides 28 PCI-E 5.0 lanes.

 

Quote

 

AMD went with:

• 16x PCI-E 5.0 lanes (i.e. graphics card)
• 8x PCI-E 5.0 lanes (i.e. NVMe M.2 SSD, up to 2x SSDs at full 4x speed ea

 

Except, unlike Intel, which is a true chipset maker, AMD isn't very straight forward with their datasheet.

Intel published their datasheet slightly more detail:

 

 

Intel at least shows that they have a dedicated PCIE lanes for the SSD. Meaning if you installed a PCIE 3.0 GPU and PCIE 4.0 SSD, each of them runs at their own specified max speed & specs.

 

And using my own mobo as a reference:

 

With AMD, anyone could straight up assume that they are all tied together, from the 'datasheet' and block diagram.

[OutsaneIngenuity]

On 1/22/2024 at 3:37 PM, QuantumSingularity said:

Yup. With Gen 5 speeds you can drop down to x4 and still the GPU would barely feel the difference. But really, if PCIe lanes compatibility and availability is you main concern, you should turn to AMD and X670E.

Already bought i913900k so that's what Imma have to go with