CPU and GPU Scaling with the Ryzen 7 5800X3D: 4 GPUs, 4 games, 3 resolutions

[Lounds]

I still cant see these benchmarks as being right. The rainbow 6 at 1080p for example. The CPUs are all the same fps on the 9070 but when you look at the 5080 they step up, but the 5600 gets more fps with the 9070 by over 100 fps. It does not seem right that the 5600 has problems maxing out with a 5080 Seems strange. Someone explain why this happens. This does not add up.

It's well known that Geforce drivers have a significant CPU overhead compared to Radeon.

 

[Nintenboy01]

It's well known that Geforce drivers have a significant CPU overhead compared to Radeon.

I think it's mainly due to the scheduler being part of the driver for GeForce cards and a hardware component in Radeon cards. I think Nvidia is still relatively more performant in DX11 games though, and of course OpenGL

 

[Maday]

This would be helpful except the 5700x3d is sold virtually everywhere (im not talking about scalpers on ebay)

 

[Lounds]

This would be helpful except the 5700x3d is sold virtually everywhere (im not talking about scalpers on ebay)

Literally take away 10% off the 5800X3D scores. Chances are it sits higher than that but on average due ot the clock speed reducation it's around 10% max fps but similar 1% scores which is what matters the most for smooth gameplay.

 

[Inthenstus]

MMORPGs are a special use case, they're worse then even your run of the mill PVP FPS. Why? Well a couple of reasons.

There is virtually no graphical exchange between a player's computer and the game server. The server is a massive database containing 3 dimensional values for every location in the game world. Then there's other server/s with location data for every moving element like NPC's and players, where they are, where they're heading, and their speed. As well servers also keep track of the interactions between well... virtually everything.

So there's a lot of data flowing between a player's computer and a the game server/s. And this puts a massive load on the CPU to begin with. But it gets worse the more players are in any one area. Because not only are the server/s keeping track of everything, they also have to supply each player with information about every other player. How well things work has a lot to do with the Net code. You might have <30ms latency, but other players might have 50, 75, or 100+ ms latency and that will affect how well the game server can refresh what you see.

So how does a better CPU help? Well there's more resources available for all the non-graphical workload and this can free up more resources for graphics. It's the main reason why the article was looking specifically at single player experiences. Even your run of the mill shooter has the same problems as a MMORPG. It's why most have a hard limit on player count. Something the MMO designation doesn't allow or it wouldn't be a MMO.

There's also one other factor that has a minor effect. There's something called host processing, which simply means that most on-board "hardware" like your sound, or network are only interfaces. All the actually processing is done by the "host" processor, your CPU. In most normal use cases your CPU doesn't even notice the extra load they add. But remember how I mentioned that a MMORPG has a really high network load? So your NIC is adding a abnormal load to the CPU and more CPU resources will help mitigate it.

A perfect example of this is Ironforge in vanilla WoW. It was the only auction house in Azeroth, and due to that there was a massive amount of players there at any one time. So much so that it created all sorts of havoc. I can remember running just outside the bank one second and finding myself in the moat between the bank and the AH the next. Blizzard did a lot of work on the net code, but what really helped was simply making more AH locations to reduce the player count. Later they upgraded to phasing, so that the load was split between servers.

That was my point, I know how MMORPGs work, I even have an addon for WoW called Hyperframe that changes a bunch of hidden settings to improve performance.

I just wish in these articles they’d add more about how bottlenecks aren’t all the same, and can vary game from game, place by place. I have a 9800X3D or 9700 drawing a blank on the name and I get better FPS in the city now in raids. Flip flopped from my non-X3D CPU. I appreciate your detailed post, I hope others see it and maybe authors will start mentioning things like this in their reviews. GGS

 

[Gezzer]

That was my point, I know how MMORPGs work, I even have an addon for WoW called Hyperframe that changes a bunch of hidden settings to improve performance.

I just wish in these articles they’d add more about how bottlenecks aren’t all the same, and can vary game from game, place by place. I have a 9800X3D or 9700 drawing a blank on the name and I get better FPS in the city now in raids. Flip flopped from my non-X3D CPU. I appreciate your detailed post, I hope others see it and maybe authors will start mentioning things like this in their reviews. GGS

But they said very clearly that they were only looking at single player games for the very reasons I stated. Too many variables that will muddy the results when you include anything beyond single player I guess...