>>106932205
If you're mainly using that PC for gaming, 5700X or a 5800XT is better than the 5900XT.
The 5900XT basically acts like a slightly better 5600X in gaming; it doesn't use the other half of the CCD in games because of the cross-CCD latency penalty.
Some games don't even know how to work with these dual CCD chips.
4chan Search
6 results for "2da2e3d20a590c6bac70c80ec2677a1e"
>>106523158
What's the price difference to the 9600X? Can you still return it?
>6 core isn't enough meme
Well, the 9900X is going to be restricted to 6 cores in most games as well.
This chart shows what happens when poorly coded games forcefully try to use all CCDs & ecores.
What's the price difference to the 9600X? Can you still return it?
>6 core isn't enough meme
Well, the 9900X is going to be restricted to 6 cores in most games as well.
This chart shows what happens when poorly coded games forcefully try to use all CCDs & ecores.
>>106141595
>"WELL ACKSUALLY I LOVE CORE PARKING"
Most games will still have to utilize one CCX due to cross-CCD latency.
That cross-CCD latency causes massive performance penalties, where you can see a 9950X performing worse than a 5600X or a 12600K.
>"WELL ACKSUALLY I LOVE CORE PARKING"
Most games will still have to utilize one CCX due to cross-CCD latency.
That cross-CCD latency causes massive performance penalties, where you can see a 9950X performing worse than a 5600X or a 12600K.
>>106105510
>Yeah it's stronger on average but what if it keeps dipping on my setup too?
Then don't buy it. It's so simple.
The e-cores are more likely to fuck over gaming performance on some of the older games/engines.
Gaming is software-bottlenecked more than you think it is. That's where most of HWUB's retested 9070 XT performance gains came from: OS & game patches.
>Yeah it's stronger on average but what if it keeps dipping on my setup too?
Then don't buy it. It's so simple.
The e-cores are more likely to fuck over gaming performance on some of the older games/engines.
Gaming is software-bottlenecked more than you think it is. That's where most of HWUB's retested 9070 XT performance gains came from: OS & game patches.
>>106001548
>They said there was no benefit and produced no proof of that
Someone else already explained the cross-CCD latency penalty, and some shitty game engines proved it's bad to use both CCDs in latency-sensitive tasks.
A 7950X3D performs worse than a 12600K in Total War: Warhammer III because the engine's trying to use all available CPU threads.
AMD initially thought of releasing a 5950X3D with 2x CCDs with a 3D stacked chiplet; their internal tests showed the variant with 1x CCD was optimal for gaming.
https://youtu.be/RTA3Ls-WAcw?t=790
>They said there was no benefit and produced no proof of that
Someone else already explained the cross-CCD latency penalty, and some shitty game engines proved it's bad to use both CCDs in latency-sensitive tasks.
A 7950X3D performs worse than a 12600K in Total War: Warhammer III because the engine's trying to use all available CPU threads.
AMD initially thought of releasing a 5950X3D with 2x CCDs with a 3D stacked chiplet; their internal tests showed the variant with 1x CCD was optimal for gaming.
https://youtu.be/RTA3Ls-WAcw?t=790
If you're wondering why Warhammer 3: Total War seems to favor Intel CPUs, the game engine's old and janky as fuck.
It doesn't know how to properly utilize AMD's dual CCD CPUs.
If an Intel hybrid CPU has too many ecores (like i9s), the 1%/0.1% lows completely tanks.
Creative Assembly employees have been requesting management to invest the resources to unfuck it.
It doesn't know how to properly utilize AMD's dual CCD CPUs.
If an Intel hybrid CPU has too many ecores (like i9s), the 1%/0.1% lows completely tanks.
Creative Assembly employees have been requesting management to invest the resources to unfuck it.