PCI-e 3.0 x8 may not have enough bandwidth for RX 5500 XT

[Waldorf]

Nope. you dial down res/settings, or upgrade your card.

And if the card is "new", then ppl didnt do their research and bought the wrong card,
or badly planned upgrade.
its like ppl hitting the key (ignition) to keep a dead car going with the alternator.

 

[Flaky]

+1

RAM is VRAM's insurance policy just as pagefile is RAM's insurance policy.
You're supposed to avoid scenarios which cause any of them to be used.

 

[Xuper]

gigabyte says : PCI-E 4.0 x 16 , Asrock says : PCI-E 4.0 x 8

Radeon™ RX 5500 XT GAMING OC 8G Specification | Graphics Card - GIGABYTE Global

Discover AORUS premium graphics cards, ft. WINDFORCE cooling, RGB lighting, PCB protection, and VR friendly features for the best gaming and VR experience!

www.gigabyte.com

asrock.com/Graphics-Card/AMD/Radeon%20RX%205500%20XT%20Phantom%20Gaming%20D%208G%20OC/

so PCIe 3.0 x8 mode only applies to 4GB? or 8GB?

 

[Flaky]

@Xuper
5500 is an x8 chip, and nothing can change that.
Gigabyte seems to be unable to decide whether they want to specify physical PCIe (x16) or electrical (4.0).
You even have a review at tom's of this exact card with gpu-z shot included.

 

[SomeOne99h]

(the 4gb only benefits from pcie 4.0. Thanks for clarifying this steve. You can rest now. And merry christmas)
Steve's response: "and probably only if you have at least 16GB's of system memory "

 

[FordGT90Concept]

Yeah, PCIe performance only seriously comes in to play when the VRAM is saturated (using more PCIe bandwidth to reach the system RAM). Don't want problems with modern games? Get a card with 8 GiB of VRAM. 4 GiB is budget now.

I'm running a 5500 XT 8 GiB on PCIe 3.0 and I'm not having any performance issues.

 

[eidairaman1]

Yeah, PCIe performance only seriously comes in to play when the VRAM is saturated (using more PCIe bandwidth to reach the system RAM). Don't want problems with modern games? Get a card with 8 GiB of VRAM. 4 GiB is budget now.

I'm running a 5500 XT 8 GiB on PCIe 3.0 and I'm not having any performance issues.

The card is only wired for 8x right or 16x?

 

[SomeOne99h]

The card is only wired for 8x right or 16x?

8x if I am not mistake.

 

[EarthDog]

8x if I am not mistake.

Correct. The whole reason why it is slower once the 4gb VRAM fills up.

 

[eidairaman1]

8x if I am not mistake.

Ok so its a 560/560D

 

[HTC]

8x if I am not mistake.

Yes. It benefits from PCIe 4.0 because 8x bandwidth with 4.0 is equal to 16x bandwidth with 3.0.

What surprises me is how this is possible since none of @W1zzard 's tests with PCIe bandwidth thus far (and there have been several) show the kind of differences in performance shown in the OP's link.

 

[EarthDog]

Yes. It benefits from PCIe 4.0 because 8x bandwidth with 4.0 is equal to 16x bandwidth with 3.0.

What surprises me is how this is possible since none of @W1zzard 's tests with PCIe bandwidth thus far (and there have been several) show the kind of differences in performance shown in the OP's link.

because his testing is with a card that doesnt saturate the available vram nor is it wired for x8.

 

[HTC]

because his testing is with a card that doesnt saturate the available vram nor is it wired for x8.

And that suggest he needs to change the way he tests GPU's PCIe bandwidth saturation, no?

Unless i'm grossly misunderstanding this "phenomena" ...

 

[Waldorf]

No.
That was only to show this isnt an issue with pcie bandwidth, as it only applies to cards that are vram limited (and swap to ram).

 

[arbiter]

And that suggest he needs to change the way he tests GPU's PCIe bandwidth saturation, no?

Unless i'm grossly misunderstanding this "phenomena" ...

Nothing wrong with testing used but way the headline was worded is the problem. When you word stories like that its called click bait. I can't read german so maybe in story they note that its vram is limited but if that was the case would said something so kinda tells me they didn't.

 

[FordGT90Concept]

Yes. It benefits from PCIe 4.0 because 8x bandwidth with 4.0 is equal to 16x bandwidth with 3.0.

What surprises me is how this is possible since none of @W1zzard 's tests with PCIe bandwidth thus far (and there have been several) show the kind of differences in performance shown in the OP's link.

It's only going to rear its head in games that need more than 4 GiB VRAM. It does show up in W1zzard's testing.

RX 5500 XT 4 GiB (1737/1845/1750) is 9-26% slower than RX 590 8 GiB:

Sapphire Radeon RX 5500 XT Pulse 4 GB Review

Sapphire's factory-overclocked Radeon RX 5500 XT Pulse comes with an amazing cooler that delivers excellent temperatures and unbelievable noise levels at the same time. Idle fan stop and a dual BIOS are included, too, for a very reasonable price increase of $10.

www.techpowerup.com

RX 5500 XT 8 GiB (1737/1845/1750) is 3-4% slower than RX 590 8 GiB.

MSI Radeon RX 5500 XT Gaming X 8 GB Review

The MSI Radeon RX 5500 XT Gaming X features twice the VRAM of the 4 GB OEM model, which we found to make a surprising difference in games, even at 1080p. As expected, the card includes an extremely quiet cooler, idle-fan-stop, and a backplate.

www.techpowerup.com

That's because some of the games in his test do exceed 4 GiB. It soars to 26% at 4K because they're having to fall back on the system RAM. He tested using a motherboard that's PCIe 3.0 x16.

The 4 GiB card numbers wouldn't be quite as terrible had he tested with a PCIe 4.0 motherboard but they'll never reach the 8 GiB card unless the games that exceed 4 GiB were purged from the benchmark. Even so, TechSpot testing shows 8 GiB 3.0 -> 8 GiB 4.0 produces frames about 2 ms quicker in Wolfenstein Youngblood so it seems that Vulkan benefits from faster lanes.

 

[Waldorf]

@arbiter
I know pcghw since they started, not a mag/site that does that kind of crap.
They wanted to show that even with a card limited to x8, running it on pcie 4 (vs 3.0) does help with performance,
not tha 4.0 wasnt an improvement, and their title was reflecting that (in german).

 

[(*^^*)]

Present a PC that can play a light game to the siblings. Because it is troublesome to buy a new motherboard, I decided to use AMD. The set of CPU and X570 was cheap and bought.

By the way, if the 5500XT4GB works like 8GB for Gen4, that's thankful.
I was worried about buying 5500XT8GB or 5500XT4GB. I will buy 5500XT4GB. With the savings. Cancel DDR4 3600 8GB × 2. Buy DDR4 3600 16GB × 2!

Good to hear good things! ,Thank you!
Thanks AMD for leaving unexpected options!

 

[EarthDog]

Present a PC that can play a light game to the siblings. Because it is troublesome to buy a new motherboard, I decided to use AMD. The set of CPU and X570 was cheap and bought.

By the way, if the 5500XT4GB works like 8GB for Gen4, that's thankful.
I was worried about buying 5500XT8GB or 5500XT4GB. I will buy 5500XT4GB. With the savings. Cancel DDR4 3600 8GB × 2. Buy DDR4 3600 16GB × 2!

Good to hear good things! ,Thank you!
Thanks AMD for leaving unexpected options!

it doesnt work like 8gb... but works better in vram full situations. honestly... I'd get the 8gb card and leave the memory alone... 16gb is enough... and if you need later, get more ram...

 

[HTC]

It's only going to rear its head in games that need more than 4 GiB VRAM. It does show up in W1zzard's testing.

RX 5500 XT 4 GiB (1737/1845/1750) is 9-26% slower than RX 590 8 GiB:

Sapphire Radeon RX 5500 XT Pulse 4 GB Review

Sapphire's factory-overclocked Radeon RX 5500 XT Pulse comes with an amazing cooler that delivers excellent temperatures and unbelievable noise levels at the same time. Idle fan stop and a dual BIOS are included, too, for a very reasonable price increase of $10.

www.techpowerup.com

RX 5500 XT 8 GiB (1737/1845/1750) is 3-4% slower than RX 590 8 GiB.

MSI Radeon RX 5500 XT Gaming X 8 GB Review

The MSI Radeon RX 5500 XT Gaming X features twice the VRAM of the 4 GB OEM model, which we found to make a surprising difference in games, even at 1080p. As expected, the card includes an extremely quiet cooler, idle-fan-stop, and a backplate.

www.techpowerup.com

That's because some of the games in his test do exceed 4 GiB. It soars to 26% at 4K because they're having to fall back on the system RAM. He tested using a motherboard that's PCIe 3.0 x16.

The 4 GiB card numbers wouldn't be quite as terrible had he tested with a PCIe 4.0 motherboard but they'll never reach the 8 GiB card unless the games that exceed 4 GiB were purged from the benchmark. Even so, TechSpot testing shows 8 GiB 3.0 -> 8 GiB 4.0 produces frames about 2 ms quicker in Wolfenstein Youngblood so it seems that Vulkan benefits from faster lanes.

I didn't explain myself properly: i meant this kind of PCIe Bandwidth tests:

Also, i was referring to the original PCGamesHardware article where it's demonstrated the difference is quite a bit higher than the one TechSpot found (different methodology, i guess).

If we do the following to a 5700XT card (as per W1zzard's 2080Ti PCI-scaling tests):

For all our PCI-Express bandwidth testing, we limit the bus-width by physically blocking the slot wiring for lanes using insulating tape. The modular design of PCI-Express allows for this. The motherboard BIOS lets us limit the PCI-Express feature set to that of older generations, too. We put the card in its various PCI-Express configurations through our entire battery of graphics card benchmarks, all of which are real-world game tests.

Does it also behave the same as a 5500XT card? And what about non-XT versions of both?

 

[eidairaman1]

Present a PC that can play a light game to the siblings. Because it is troublesome to buy a new motherboard, I decided to use AMD. The set of CPU and X570 was cheap and bought.

By the way, if the 5500XT4GB works like 8GB for Gen4, that's thankful.
I was worried about buying 5500XT8GB or 5500XT4GB. I will buy 5500XT4GB. With the savings. Cancel DDR4 3600 8GB × 2. Buy DDR4 3600 16GB × 2!

Good to hear good things! ,Thank you!
Thanks AMD for leaving unexpected options!

I would suggest a RX570 8G.

 

[HTC]

Present a PC that can play a light game to the siblings. Because it is troublesome to buy a new motherboard, I decided to use AMD. The set of CPU and X570 was cheap and bought.

By the way, if the 5500XT4GB works like 8GB for Gen4, that's thankful.
I was worried about buying 5500XT8GB or 5500XT4GB. I will buy 5500XT4GB. With the savings. Cancel DDR4 3600 8GB × 2. Buy DDR4 3600 16GB × 2!

Good to hear good things! ,Thank you!
Thanks AMD for leaving unexpected options!

It doesn't: suggest you read the original article while paying more attention to it because that's not what it suggests @ all.

 

[Waldorf]

@(*^^*)
your wasting your money on the wrong thing.
having 6 or 8gb vram is always better than having another 16gb ram, at least for gaming rigs.

 

[FordGT90Concept]

Does it also behave the same as a 5500XT card? And what about non-XT versions of both?

Not sure what you're asking. 5500 XT and 5500 both have fingers for x16 but only up to x8 are physically connected. AMD likely does that because fewer traces through the PCB means lower cost for AIBs.

 

[HTC]

Not sure what you're asking. 5500 XT and 5500 both have fingers for x16 but only up to x8 are physically connected. AMD likely does that because fewer traces through the PCB means lower cost for AIBs.

Test PCIe scaling the exact same way the 2080Ti PCIe scaling was tested (check 1st and 3rd links of my previous post and compare the methodology).

Are there other cards with x16 but LESS physically connected, other than the 5500(XT) cards, regardless of manufacturer? If there are, those too should be tested.