Are PCIe 4.0 speeds worth it?

[ABunchaHerbs]

Are PCIe 4.0 speeds really noticable? I have always wondered whether it is worth buying a X570 Mobo to pair with your, say, Ryzen 5 3600?

And in what areas would I see improvement?

[Skiiwee29]

You wont see any areas of improvement. Its only good and improved for things that need the IO throughput, usually reserved for machine learning and AI stuff, not for gaming. Hell, in gaming and normal everyday tasks, you wont notice a different going from a good sata 3 SSD to NVMe pcie 3.0x4. 

[For Science!]

2 minutes ago, ABunchaHerbs said:

-

Usually if you have to ask this question, it is not worth it. The PCIe 4.0 Bandwidth is pretty much only relavent for very fast SSDs or arrays of SSDs doing very heavy i/o workloads. So if you aren't sure why it would beneficial for you, you probably are not doing anything that can utilize it.

[ABunchaHerbs]

1 minute ago, Skiiwee29 said:

You wont see any areas of improvement. Its only good and improved for things that need the IO throughput, usually reserved for machine learning and AI stuff, not for gaming. Hell, in gaming and normal everyday tasks, you wont notice a different going from a good sata 3 SSD to NVMe pcie 3.0x4. 

Ok thanks! I can save a bit of money now and not worry as I'm only planning gaming, streaming and video editing.

[porina]

See it as a future technology. If you don't know why you need it, you probably don't.

[mr moose]

16 minutes ago, Skiiwee29 said:

you wont notice a different going from a good sata 3 SSD to NVMe pcie 3.0x4. 

call me caught in a placebo then, because I could swear my new NVME was faster than my older sata SSD.

[Arika]

34 minutes ago, porina said:

See it as a future technology. If you don't know why you need it, you probably don't.

Yup. The technology needs to exist before it can be of any use.

 

What comes first? The hardware, or the support for the hardware?

[SnowyMus]

I guess the short answer to this question is: for most people, it may become beneficial later but it really isn't right now.

 

You might find some PCIe 4.0 SSDs or a configuration of SSDs that saturate a PCIe 3.0 interface and require PCIe 4.0 to get the full speed. I'm sure if you threw enough SSDs together in RAID 0, you could probably even saturate a PCIe 4.0 interface.

 

As for PCIe 4.0 graphics cards, you can transmit data to and from the GPU faster. For stuff like high resolution textures, this could be useful? AMD APUs also happen to be limited to 8 PCIe lanes, so you'd basically effectively make them have the same bandwidth as PCIe 3.0 x16 (assuming they use a PCIe 4.0 graphics card), but I'm very doubtful someone will buy an AMD APU and an X570 board together unless the APU was temporary. 8 PCIe 3.0 lanes is still fine for gaming, anyway.

 

We might see more technologies use the high bandwidth as time goes by.

 

Perhaps you may be able to equate this to asking someone in 2009-2011 if 16 GB of RAM was worth it. Yes, you could get 16 GB of RAM if you wanted to, but the high price and limited number of use cases made it not worth it for most people at that time.

[Kilrah]

27 minutes ago, Kavawuvi said:

For stuff like high resolution textures, this could be useful?

In theory yes, but people have run tests and those haven't shown any significant difference on current hardware and software.

But yeah as you said it's already considered that running a graphics card at PCIe3 x8 instead of x16 results in only about 5% difference in performance, so we're still far from saturating even PCIe3 x16 at this point.

[bowrilla]

Ignore it for now. There are no PCIe 4.0 capable GPUs on the market and won't be for a while up to the point that the next mainboard generation will be available. Current gen consumer (and most prosumer) GPUs usually can't even saturate PCIe 3.0 16x. NVMes are a different story but this is more for bragging rights then actually perceivable improvements for the average user. 

 

Edit: see this test by techpowerup. The 2080Ti manages to saturate PCIe3 8x but not by much considering the bandwith of PCIe3 16x is twice the one of an PCIe3 8x interface: the gain is on average 2-3%. There's a lot of bandwith left unused.

[caldrin]

3 hours ago, bowrilla said:

Ignore it for now. There are no PCIe 4.0 capable GPUs on the market and won't be for a while up to the point that the next mainboard generation will be available.  co

The new AMD cards are PCIe 4.0

 

 

[bowrilla]

1 hour ago, caldrin said:

The new AMD cards are PCIe 4.0

 

 

I forgot about those. I really wonder why AMD went that route. It's basically useless when looking at the benchmarks: TechPowerUp PCIe4 performance scaling

[Mathieu9836]

If you are looking for a higher tier B450/X470 board then i would grab the X570 with a similar price. If you don't need a high power delivery, 2-3 Sata ports, 2 GPU slot, lot of fan connetors, IMO the X570 are not worth the money.

[Tan3l6]

It's as useful as highway for golf carts.

at this time

[RejZoR]

Main problem why PCIe 4 doesn't help much with games is the way they are coded. If games streamed more data on the fly instead of stuffing it all in one go into VRAM, PCIe 4 would make a HUGE difference in performance.

[Mathieu9836]

1 hour ago, RejZoR said:

Main problem why PCIe 4 doesn't help much with games is the way they are coded. If games streamed more data on the fly instead of stuffing it all in one go into VRAM, PCIe 4 would make a HUGE difference in performance.

That would probably be at the cost of something, like having faster CPUs, or lot of RAM.

[SnowyMus]

9 hours ago, Kilrah said:

In theory yes, but people have run tests and those haven't shown any significant difference on current hardware and software.

But yeah as you said it's already considered that running a graphics card at PCIe3 x8 instead of x16 results in only about 5% difference in performance, so we're still far from saturating even PCIe3 x16 at this point.

I suppose, today, if your SSD and GPU have more bandwidth, though, then loading times could be better. That's about all I can think of if we mean the present.

 

But for performance, no "optimized" game will take advantage of PCIe 4.0 today. Most GPUs are still on PCIe 3.0, and most games still perform very well with a PCIe 3.0 x8 interface. Trying to do anything that would constantly require more than a x16 PCIe 3.0 interface would be completely insane today, because only certain mid-range GPUs would not get dumped on in performance.

 

Here's an example: Let's say you're on a GTX 1660 Ti (6 GB GDDR6, PCIe 3.0 x16, 288 GB/s memory bandwidth). Let's also consider twenty 8192 x 8192 32-bit uncompressed textures with mipmaps. First we have the main image which is 8192 x 8192 pixels, or over 67.1 million pixels total. Then, for mipmaps, you divide each dimension by 2 until you get to 0.5. So, you get this: (8196^2 + 4096^2 + 2048^2 + 1024^2 + 512^2 + 256^2 + 128^2 + 64^2 + 32^2 + 16^2 + 8^2 + 4^2 + 2^2 + 1^2) pixels * 32 bits per pixel = 358 176 148 bytes per texture.

 

For simplicity's sake, let's assume there's no overhead. If there was no overhead, then PCIe 3.0 can send that at 1 GB per second with one lane. Note that PCIe bandwidth is measured using IEC GB, so 1 GB = 1 billion bytes in this case. Thus it'll take a little over 0.358 seconds to send one texture to the graphics card. Obviously, this would make the game unplayable if it had to load many assets, but if you have 16 lanes, you can divide that by 16 and you get about 22.4 milliseconds. With 20 8K textures, you can do that in 0.448 seconds. But there's a problem: you're on a GTX 1660. With a 6 GB (6 GiB) GDDR6 frame buffer, you can only hold 17 uncompressed, 32-bit, 8K textures. This means the game has to switch textures in and out during gameplay every once in a while.

For your game to run at a stable 60 FPS, it has to draw a frame every ~16.667 milliseconds. This means that every time it loads a single texture during gameplay, there is going to be a hitch of at least around 22.4 milliseconds (or a drop to under 44.7 FPS). This is very noticeable. But if you used PCIe 4.0, it'd only take half that time: ~11.2 milliseconds. You'll still get a hitch, and it'll definitely show up in your 1% lows, but it won't be nearly as significant, plus it's entirely possible it'll still maintain 60 FPS. Also, if you were on 8 GB of VRAM, more things can be stored in memory, so you wouldn't even need to worry about this nearly as much.

 

Today, we would probably call such a game an "unoptimized" game. Not because it's computationally demanding but because it's constantly trying to shove large, uncompressed textures through a small interface. More likely, you'd see games using some form of (probably lossy) compression supported by the GPU to get such a high resolution texture as low as possible. But ten years from now? Maybe! We might even see more games using 16K textures, too, as more GPUs support it.

[TheVillageIdiot]

Pci-E 4 means more for storage and certain workloads in relation to storage. Gpu's not so much.

[caldrin]

On 10/8/2019 at 5:21 PM, bowrilla said:

I forgot about those. I really wonder why AMD went that route. It's basically useless when looking at the benchmarks: TechPowerUp PCIe4 performance scaling

Yeah its a bit strange but i guess as they were promoting PCIe 4.0 on their new motherboard range then it made sense.. plus i guess it will now be part of all Graphics cards in the future so gaining more experience working with the tech is a good thing even if here is basically now performance gain in the current gen cards...

[ABunchaHerbs]

On 10/8/2019 at 8:59 PM, Kavawuvi said:

I guess the short answer to this question is: for most people, it may become beneficial later but it really isn't right now.

 

You might find some PCIe 4.0 SSDs or a configuration of SSDs that saturate a PCIe 3.0 interface and require PCIe 4.0 to get the full speed. I'm sure if you threw enough SSDs together in RAID 0, you could probably even saturate a PCIe 4.0 interface.

 

As for PCIe 4.0 graphics cards, you can transmit data to and from the GPU faster. For stuff like high resolution textures, this could be useful? AMD APUs also happen to be limited to 8 PCIe lanes, so you'd basically effectively make them have the same bandwidth as PCIe 3.0 x16 (assuming they use a PCIe 4.0 graphics card), but I'm very doubtful someone will buy an AMD APU and an X570 board together unless the APU was temporary. 8 PCIe 3.0 lanes is still fine for gaming, anyway.

 

We might see more technologies use the high bandwidth as time goes by.

 

Perhaps you may be able to equate this to asking someone in 2009-2011 if 16 GB of RAM was worth it. Yes, you could get 16 GB of RAM if you wanted to, but the high price and limited number of use cases made it not worth it for most people at that time.

What about for like, NVMe SSD's or PCIe SSD's? Will there be an increase in transfer rates with PCIe 4.0?

[Turtle Rig]

On 10/8/2019 at 1:17 AM, ABunchaHerbs said:

Are PCIe 4.0 speeds really noticable? I have always wondered whether it is worth buying a X570 Mobo to pair with your, say, Ryzen 5 3600?

And in what areas would I see improvement?

Intel says no!  They say the PCIe 4.0 is really not faster then a 3.0.  But then again this is Intel who doesn't have a PCIe 4.0 board out.  They say its useless yet tout the fact PCIe 6.0 specs are coming 2021.  But honestly its not really going to do anything.  Even PCIe 4.0 SSD cards that have high benchmarks will work just like PCIe 3.0 or M.2 or SSD.   Their all the same speed in the real world unless you do copying from one to the other.  Like 2 PCIe 4.0  and copy between them it will be lightning fast as for OS drive its same sH*T even tho benchmarks are 4000mbps and m.2 2500mbps and SSD is 550mbps..... Benchmarks show the difference but real world everything will work the same.

[ABunchaHerbs]

3 minutes ago, Turtle Rig said:

Intel says no!  They say the PCIe 4.0 is really not faster then a 3.0.  But then again this is Intel who doesn't have a PCIe 4.0 board out.  They say its useless yet tout the fact PCIe 6.0 specs are coming 2021.  But honestly its not really going to do anything.  Even PCIe 4.0 SSD cards that have high benchmarks will work just like PCIe 3.0 or M.2 or SSD.   Their all the same speed in the real world unless you do copying from one to the other.  Like 2 PCIe 4.0  and copy between them it will be lightning fast as for OS drive its same sH*T even tho benchmarks are 4000mbps and m.2 2500mbps and SSD is 550mbps..... Benchmarks show the difference but real world everything will work the same.

But what about Ryzen?

[Turtle Rig]

Just now, ABunchaHerbs said:

But what about Ryzen?

Ryzen just has the motherboards x570 that offer the PCIe 4.0 and all this nonsense does is put a fan in your chipset that makes a crummy noise.  You bring up Ryzen... As far as RAM yes Ryzen will like faster RAM better then Intel.  But PCIe 4.0 is a standard has nothing to do with Ryzen.  AMD just adopted it while Intel is hell bound its a waste and not really faster then PCIe 3.0 which is why their not bothering with it for their new version x299 boards for their upcoming 1080XE CPU.  But Intel contradicts in a way unless they really know what their talking about.  Their already boasting PCIe 6.0 specs in 2021.  Apparently according to Intel PCIe above 3.0 doesnt matter until its PCIe 5.0 or 6.0 lol.  Nice way to weisel out of the fact AMD beat them to it.  But they do have valid point as to why it is really not needed right now unless you wanna buy 2 PCIe 4.0 SSD drives and transfer between them.  This will yield at least 3000mbps potentially.  I mean even m.2 drives are 2500mbps.  When I went from SSD 550mbps to M.2 the only difference I felt and saw was the M.2 was 2500mbps.  But the OS loaded same time apps launched same time and files copied same time.  The M.2 is a form factor.. No need for cables and tiny and what not.

[ABunchaHerbs]

3 minutes ago, Turtle Rig said:

 But the OS loaded same time apps launched same time and files copied same time.  

That's weird. I've heard from many other people that an NVMe M.2 SSD will transfer/read files significantly faster and the OS boot would be much faster...

[Turtle Rig]

Just now, ABunchaHerbs said:

That's weird. I've heard from many other people that an NVMe M.2 SSD will transfer/read files significantly faster and the OS boot would be much faster...

Well heres the thing my M.2 is on a Adapter cuz my motherboard is very old.  I run crystal benchmark and I get 2500mbps ....... But honest to God I feel no difference in boot up times or game map loading times or in my DAW.  Unless a M.2 on a adapter is botched what I say stands.  Most people will say anything just so their happy with their cute little m.2 stick ya know.  As I said just a form factor that's it.  One advantage it has that possibly SSDs don't is you can do many things at once.  Like if the drive is in usage copying stuff you can still use your computer and apps and other things will be just as fast even tho your doing some copying say from a m.2 to another drive like a SSD.  If you copy from m.2 to m.2 then it will be very fast tho and you will near the 2500mbps.  But how many people do this ?  People put their OS drive on the M.2 or PCIe 4.0 SSD.  But ya dont expect a twin turbo Porsche once you go M.2 or PCIe 4.0 compared to your SSD drive.  Real world Im talking just note,, on benchmarks its a lot like I said.