AMD or Intel for GeForce RTX 3080 Benchmarking: Is PCIe 4.0 a Factor?

[Adi6293]

I would probably switch to AMD now and later on just swap the CPU plus like you said in the video it would be good too see how those new cards behave on AMD system vs other revives because right now we do get practically the same benchmarks from everyone

 

[Soaptrail]

Everyone else tests with Intel, so yours would be one of the only benches to use AMD, and thus stand out from the crowd a bit and find any weird behavior that would otherwise be missed.

I concur, set your review apart and use AMD.

 

[Irata]

I have a 6 year old Z97 board with no M2. And I use an adapter on the 3rd slot on the board for an M2 drive. I was convinced this would reduce my lanes to 8 but it doesn’t. GPUz reports 16 lanes still. I didn’t bother checking any further but I don’t think you always lose 8 lanes by using an adapter, it depends on the mobo maybe.

Also, prior to purchasing the adapter I researched and found that 8x PCIe3 doesn’t slow down your GPU, it offers enough bandwidth for modern current cards. Which is why I’m personally skeptical that PCIe4 will yield an advantage for the new upcoming graphics cards, I guess we’ll find out when TechSpot test it and publish the results.

I think PCIe4 is more useful for OEM marketing departments than actual users at this point

As for my example, you can take the Aorus Master B550. If you run the GPU in PCIe 4 x8 mode (same bandwidth as PCIe x16), you can run three nVME drives @ PCIe 4 x4 speed.

From Guru3d‘s review:

Next to the dedicated x4 lanes for a single M.2. SSD, the Gigabyte B550 Aorus Master will split up the x16 PCIe Gen 4.0 graphics slot into a single 8x connection, that frees 2x four lanes.

For those wanting to run several nVME drives, that should be a big advantage.
So I‘d say there will be more and more benefits of having PCIe 4 going forward. If the current top of the line GPU will manage to take advantage of it, or not - we‘ll see soon enough. But looking at the overall system it will be an advantage vs. 16 PCIe 3 lanes that will definitely be needed for the GPU and everything else going via the chipset link.

 

[mAdmAnDingo]

I am good with going with AMD, and I am good if we stay with Intel. Being an AMD owner (and on X570) I am interested in an AMD PCIE-4.0 test system for upcoming GPUs and Zen3 CPUs. But if other readers will be disturbed with switching to AMD and would rather have us stay on Intel for the benchmark system, I am good with that as well.

So I am okay regardless, and I am fine with either choice. Just my two cents.

 

[Shadowboxer]

As for my example, you can take the Aorus Master B550. If you run the GPU in PCIe 4 x8 mode (same bandwidth as PCIe x16), you can run three nVME drives @ PCIe 4 x4 speed.

From Guru3d‘s review:


For those wanting to run several nVME drives, that should be a big advantage.
So I‘d say there will be more and more benefits of having PCIe 4 going forward. If the current top of the line GPU will manage to take advantage of it, or not - we‘ll see soon enough. But looking at the overall system it will be an advantage vs. 16 PCIe 3 lanes that will definitely be needed for the GPU and everything else going via the chipset link.

Absolutely PCIe4 will matter in the future. I’m just talking about now. So far I haven’t seen much in the way I’d benefit from it. But this hasn’t stopped manufacturers from marketing it as a reason to buy now. For a GPU, especially the AMD 5xxx parts, I think it’s a little misleading to advertise it as it doesn’t make that GPU any faster if you have a PCIe4 motherboard.

 

[Decoy]

If you wanted to compare the effect of PCIe 3 vs 4, would it not have been more straight forward to make the comparison between B450 and X570 ?

Other than that, great review and I'm looking forward to the first next gen reviews.

Being on B450 but only interested in mid range GPU , I am curious if PCIe 4 makes a difference with next gen GPU and if so in which tier(s).

You can test PCI-E 3 vs 4 on X570 just fine

 

[Decoy]

I think you should do a couple of quick tests as a sanity check before committing to the whole benchmark suite on one platform. Test on whichever one seems to have the advantage, otherwise people will just slap on a fanboy label.

 

[Decoy]

Then you need to hire more staff or increase your budget... because this is probably the single most important thing your site does.

In order to accurately assess these new video cards - as well as AMD’s which we must assume will be released fairly soon - the user needs to know which platform to use with them.

Not having 2 systems is simply unacceptable.

You can't test at the same time with two systems. Then you're just as likely to see variance between the two GPUs. The same GPU must be used for both systems or your testing is invalid.

 

[Decoy]

How is this even a thing? and why ignore the fact that so many AMD users don't have a 3950x much less a 550 or 570 board to use it on? A person could be running a 3950x on an x470 board where pci-e 4 is irrelevant. Are you going to label every graph as pci-e 4 so they know?

If you run a 3950x and it turns out its bottle-necking (same for an intel CPU), you're doing your viewers a disservice. The reality also might be that faster GPUs result in a wider performance delta between the 10900k and 3950x.

A poll I see as relevant is asking what systems they are using in case you would rather tailor your tests to the majority. This noise is premature. The article should be about what the findings are when the cards actually launch.

The point is to remove any potential CPU bottleneck to show the maximum potential of the new GPUs. Everything else will be tested after. It's like you don't understand the basic concept of a review.

 

[Lew Zealand]

Then you need to hire more staff or increase your budget... because this is probably the single most important thing your site does.

In order to accurately assess these new video cards - as well as AMD’s which we must assume will be released fairly soon - the user needs to know which platform to use with them.

Not having 2 systems is simply unacceptable.

You're gonna have to head to another site which does that then.

In the YT comments to the video accompanying this, Steve said that of course they will do a head to head test of Intel and AMD with PCIe 4.0 video cards to set the baseline, but moving forward they will have only one system to test with.

 

[Irata]

Absolutely PCIe4 will matter in the future. I’m just talking about now. So far I haven’t seen much in the way I’d benefit from it. But this hasn’t stopped manufacturers from marketing it as a reason to buy now. For a GPU, especially the AMD 5xxx parts, I think it’s a little misleading to advertise it as it doesn’t make that GPU any faster if you have a PCIe4 motherboard.

I think it will be an advantage very shortly, of course depending on use case.

I guess it‘s nice to have a platform that is forward looking both wrt CPU upgradeability as well as better IO. That will help a system stay relevant for longer, even if it‘s for the next owner.

Edit: Computerbase did a review comparing RDNA1 cards on PCIe 3 vs 4. While there were no noticeable differences on the 8GB 5700XT, there were sometimes with the 5600 XT with 6GB. They said they suspect that this happened in cases the card did not have sufficient memory so had to fetch textures over the PCIe bus.
The test was done on 1080p, so we could expect next gen cards with 8 or 10 GB to run into a similar situation on 4k resolutions with high details. Again, we‘ll see.

In German, so via Google translate:

PCIe 3.0 vs. 4.0 in Spielen im Test

Navi-Grafikkarten von AMD unterstützen PCIe 4.0. Wie der Test zeigt, kann dies durchaus von Vorteil sein. Das gilt aber nur in einem Fall.

translate.google.com

 

[Maxiking]

There is techpowerup review of 17games with every amd cpu except 3950x, amd is slower in 4K by a few %.

AMD is still a huge bottleneck for 2080ti and will be even a bigger bottleneck for Ampere. Not to mention that 1080p and 1440p are the most used resolutions.

Maybe wait for Ampere before spreading nonsense.

 

[brucek]

Both? If that's too much for one reviewer / one site, maybe pick a partner and agree to share the results? Crowd sourcing the data could be an answer too?

 

[Shadowboxer]

I think it will be an advantage very shortly, of course depending on use case.

I guess it‘s nice to have a platform that is forward looking both wrt CPU upgradeability as well as better IO. That will help a system stay relevant for longer, even if it‘s for the next owner.

Of course. But how much is it worth paying for. And is it worth sacrificing something single core speed for? Currently that’s the cho

I think it will be an advantage very shortly, of course depending on use case.

I guess it‘s nice to have a platform that is forward looking both wrt CPU upgradeability as well as better IO. That will help a system stay relevant for longer, even if it‘s for the next owner.

Well of course. I’m just saying that today there’s little advantage for it. But this won’t stop GPU manufactures marketing PCIe4 as a reason to buy their GPU when it offers no actual benefit over PCIe3.

I can’t see the advantage anytime soon for gamers. The only thing I can think of is the PS5 is leveraging PCIe4 to delivering very fast loading times for games. If that can make its way to the PC then maybe we’ll get that benefit. And maybe the new GPUs will use the extra bandwidth. But I doubt it. An 8x PCIe3 GPU doesn’t really run any slower than when at 16x, so I don’t think the bandwidth of the PCIe connect is the limiting factor.

 

[Alfatawi Mendel]

Intel don't have PCIE V4.0 implementation as yet.
That is why it could only be an AMD platform for testing the new upcoming GPU's, to see if they are capable of being hampered by a PCIE V3.0 interface.
I know it's tough on the Intel fanbois, but Intel are behind the curve at the moment.

 

[Lionvibez]

I am good with going with AMD, and I am good if we stay with Intel. Being an AMD owner (and on X570) I am interested in an AMD PCIE-4.0 test system for upcoming GPUs and Zen3 CPUs. But if other readers will be disturbed with switching to AMD and would rather have us stay on Intel for the benchmark system, I am good with that as well.

So I am okay regardless, and I am fine with either choice. Just my two cents.

I'm in the same boat already on a AM4 x570 platform and using a Pcie 4.0 M2 drive and will go RDNA2 in the fall which will also be PCIe 4.0 so looking forward to more reviews covering it.

 

[Decoy]

There is techpowerup review of 17games with every amd cpu except 3950x, amd is slower in 4K by a few %.

AMD is still a huge bottleneck for 2080ti and will be even a bigger bottleneck for Ampere. Not to mention that 1080p and 1440p are the most used resolutions.

Maybe wait for Ampere before spreading nonsense.

TPU also uses a version of Windows and chipset driver that's a year old. You're looking at <10% differences at 1080p and calling it a massive bottleneck. Maybe reevaluate who is spreading the nonsense.

 

[Squid Surprise]

You're gonna have to head to another site which does that then.

In the YT comments to the video accompanying this, Steve said that of course they will do a head to head test of Intel and AMD with PCIe 4.0 video cards to set the baseline, but moving forward they will have only one system to test with.

I already have.... but I have the ability to surf to multiple sites.... as techspot should have multiple benchmarking PCs

Intel’s new chipset (Whenever it actually arrives) will almost certainly have PCIe 4 as well....

Honestly, there should be top of the line AMD and Intel PCs used for testing on any serious tech site.

 

[moobg]

Then you need to hire more staff or increase your budget... because this is probably the single most important thing your site does.

In order to accurately assess these new video cards - as well as AMD’s which we must assume will be released fairly soon - the user needs to know which platform to use with them.

Not having 2 systems is simply unacceptable.

Find me a reputable/notable site that does thorough testing of GPUs across two test systems, or has stated that they'd do the immense battery of tests required once these new GPUs are released on two different test systems from now on. I'll wait.

As has already been stated, they'll do a baseline test for both systems, which is great, but moving forward they're probably going to move to a single system, which is what the vast majority of reviewers do (I can't think of any reviewers off the top of my head that does GPU reviews with two systems).

 

[candle_86]

Absolutely PCIe4 will matter in the future. I’m just talking about now. So far I haven’t seen much in the way I’d benefit from it. But this hasn’t stopped manufacturers from marketing it as a reason to buy now. For a GPU, especially the AMD 5xxx parts, I think it’s a little misleading to advertise it as it doesn’t make that GPU any faster if you have a PCIe4 motherboard.

It's advantage like anything badwidth related is about stress. It's likely the extra bandwidth would benefit ray tracing,dlss and 4k as the amount of data the chip needs increases as resolution and features add up. It's the same reason why in 2008 ddr2 and 3 looked the same but by 2010 you could tell phenom ii was hobbled by ddr2 and you needed am3 to get those most out of it.

To quote one of the best films of all time, "if you build it, they will come"

 

[Evernessince]

I’d say pick the CPU that benches the fastest. If PCIe4 makes cards faster on Ryzen the faster go with Ryzen, if it doesn’t and Intel is still faster than use Intel.

Neither is a typical scenario, I don’t think that many gamers are buying either CPU when it comes to the i9 or R9, instead a 3600 or a maybe a 10600K is a best representation of what users may buy with a new GPU when building a new rig.

But looking at the GPUs and the fact that we aren’t likely to see any bottlenecks I don’t think it matters that much. AMD fanboys will be filled with glee to see Ryzen on a gaming GPU test bench so you may aswell throw them a bone when the CPU performance doesn’t really matter.

But I do think it’s important that if PCIe4 doesn’t affect performance that you let users know. Motherboard manufacturers and AMD seem to be very keen to promote PCIe4 but in reality most users wouldn’t be able to tell the difference over PCIe3. And I have no doubt that PCIe4 will be written all over the GPU boxes even if it has no benefit.

It also may worth bearing in mind that Ryzen 3 is apparently due to drop within the next few months and if it performs faster it will do Ryzen a dis-service as over the next year when people are reading the reviews in a few months (which is very possible, not everyone buys GPUs on release). I can definitely see confusion between Ryzen 3xxx and Zen 3 occurring. We wouldn’t want to lead users to buying Intel because it’s faster than Ryzen 3xxx when Ryzen 4xxx stuff is faster (if it’s faster).

There were 3 games tested above where the same GPU and CPU combination performed 1-3% faster on PCIe 4.0 over PCIe 3.0.

Whether or not that will increase with new video cards remains to be seen but it does appear that it certainly could play an impact, which is why the conclusion to this article says something to a similar effect.

 

[MaxSmarties]

You should use the better platform, that overall is the AMD at this time.

 

[Maxiking]

TPU also uses a version of Windows and chipset driver that's a year old. You're looking at <10% differences at 1080p and calling it a massive bottleneck. Maybe reevaluate who is spreading the nonsense.

Sure, <10%

https://imgur.com/oleSHWh

HZD, the amd "optimized", I mean gimped title

https://imgur.com/RkK80pd

AMD is a massive bottleneck in plenty of games maybe watch someone doing proper reviews like GamersNexus.


Performance of amd hasn't changed almost at all over the time, if anything, it decreased due to less aggressive bioses and lower voltages after reviews. So please, spare of this fine wine conspiracy, it has been debunked 100 times that the performance of AMD cards do not increase in the older titles and it is the same compared to Nvidia. It is just Nvidia releasing a new arch unlike AMD, which was rebranding and introducing refreshes after refreshes in the past, 480rx got rebranded 3 times - 480rx, 580rx, 590x. Look how Vega performs terribly in new games, slower than 1060gtx, so fine wine got busted.

Techpower up has 3900xt, 3600xt and 3800xt reviews up, they used the bios AMD had told them to use.

Any other excuses of yours I shall address?

 

[ViPeRGT2]

I'd like to see benchmarks run a bit differently.....

Most people have a Chrome(or other Web browser), Discord, 2 or 3 game launchers, and/or Spotify running at the same time as the game. Would love to see some benchmarks with other programs also running in the background, like most users, to see if that makes any difference to the results.

Even if it's just a once off, to see if AMD vs Intel really makes a difference in everyday use with multiple cores being used, while gaming.

If AMD is the CPU most users are currently purchasing then using them as the benchmark makes sense.

 

[Bamda]

Finally a story I can relate too. I have not gamed at 1080p in 8 years plus! As I already knew gaming at 1440p and 4k is less about the CPU and more to do with GPU. Good to see TechSpot finally caught up with modern hardware.