CPU Scaling with 9070 XT / 5070

[RandallFlagg]

Hard to find, useful data from Hardware Canucks.

 

[rusty caterpillar]

Hard to find, useful data from Hardware Canucks.

View attachment 407554

What is strange about it that are tested only K and X variants

 

[Krit]

What is strange about it that are tested only K and X variants

Difference are small for such a big work. At least for Ryzen usually X versions are waste of money.

 

[Sithaer]

Thanks that 5070 part was actually something I was curious about since I do happen to have that 12600K/5070 combo with a 100-120FPS global cap depending on the game and for the most part I can hit that at my resolution so I really wouldn't get anything out of a better CPU w/o wasting even more of my money.

 

[RandallFlagg]

What is strange about it that are tested only K and X variants

There's a 9800X3D in there, which serves as a good benchmark here for the least CPU limited situation. It illustrates that with a 5070, a 9800X3D won't give you a whole lot if you have a 12600K / 5600X or higher - on average.

Usually CPU scaling tests are only done with 4090 / 5090, which doesn't say much if you are looking at more mid to upper midrange like the 5070.

OFC, the posted chart is an average. There are cases which don't fall into the generic implications of that chart.

This type of game for example really likes a beefy CPU, and its taking a hefty penalty with anything less than a 14600K.

 

[Tyler-98-W68]

Interesting how the 1% lows are higher on the nvidia 5070 intel side of things vs amd/amd

 

[Dr. Dro]

Interesting how the 1% lows are higher on the nvidia 5070 intel side of things vs amd/amd

Topology and L3$ allocation makes or breaks Ryzen most of the time. The 9800X3D is a fantastic gaming processor, but that's usually where it stops.

 

[nguyen]

5070 only lose 1.5% avg FPS when going from 9800X3D --> 265K, meanwhile 9070XT lose 5.3%

Radeon getting better FPS when paired with Ryzen CPU is not something new.

Every reviewer nowadays use the 9800X3D in their test system, that make RTX appear weaker than they actually are

 

[RootinTootinPootin]

but that's just your typical YTuber stuff (I don't base my decisions over those, there's plenty of forums out there with a mix of enthusiasts and normal user base)

 

[droopyRO]

Those graphs and charts are kind of useless in general.
Because they don't show you how often your games stutter and how often they hit that .01% lows. I upgraded my Ryzen 2700X to a Ryzen5600. The stutter, lag and spikes are all gone in the games i play. Even if one is a 8 core CPU it's outdated for modern games. While those 6 core R5 5600 has no issues running those game smoothly.
Your best bet, as a customer looking for an upgrade, is to find a clip of *whatever game you play or want to play* with *CPU* + *GPU* of your choice.

Those charts like the ones here on TPU or Hardware Canucks are just to narrow your choices, not to make a choice.

 

[CheteriagalenTriagalnik]

Those graphs and charts are kind of useless in general.
Because they don't show you how often your games stutter and how often they hit that .01% lows. I upgraded my Ryzen 2700X to a Ryzen5600. The stutter, lag and spikes are all gone in the games i play. Even if one is a 8 core CPU it's outdated for modern games. While those 6 core R5 5600 has no issues running those game smoothly.
Your best bet, as a customer looking for an upgrade, is to find a clip of *whatever game you play or want to play* with *CPU* + *GPU* of your choice.

Those charts like the ones here on TPU or Hardware Canucks are just to narrow your choices, not to make a choice.

Well, the interior bar shows the 1% lows, so that should give you some indication for stutters.

 

[droopyRO]

Yes and no, they don't show you how frequent they are. If it's once a few minutes it's nothing, if it happens once every few seconds, the gaming experience is bad.

 

[RandallFlagg]

5070 only lose 1.5% avg FPS when going from 9800X3D --> 265K, meanwhile 9070XT lose 5.3%

Radeon getting better FPS when paired with Ryzen CPU is not something new.

Every reviewer nowadays use the 9800X3D in their test system, that make RTX appear weaker than they actually are

It's only with the X3D really. When you look at a more CPU bound game like BG 3 you can really see it.

Basically it shows what we've known for a long time - Radeon GPUs suck more CPU cycles than Nvidia. So much so that in this game, the 5070 handily outperforms the 9070 XT (a $200 more expensive GPU) on any CPU less than a 9800X3D.

 

[rusty caterpillar]

It's only with the X3D really. When you look at a more CPU bound game like BG 3 you can really see it.

Basically it shows what we've known for a long time - Radeon GPUs suck more CPU cycles than Nvidia. So much so that in this game, the 5070 handily outperforms the 9070 XT (a $200 more expensive GPU) on any CPU less than a 9800X3D.

Yes, but that higher usage of CPU paired with 9070 XT is maybe to do with the OS > Wintel, on the other hand can you bet on 5070 for a smooth experience in 1440 P for the next 3 years?