![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Exactly, and with this particular part (5800X3D), I'd venture to say the majority of purchases are people upgrading that are already on AM4. I don't imagine many buying an AM4 board right now to use this thing on. They'll either go Intel DDR5, or wait for AM5 w/ DDR5 if buying new.
I disagree. It makes perfect sense to lower the graphics settings only to the point where the GPU is not creating a bottleneck. On an old title with a 3090, that's quite possibly as high as 1440p ultra although 1080p ultra is likely more reliable.
If lowering the resolution change absolutely anything to the performance, safe to assume the GPU was creating a bottleneck, no ?
They didn't concoct anything, this is a leak of a benchmark in a known game from the same people that leaked non-gaming benchmarks yesterday that showed it being worse than the original 5800x and 12900ks. I would have liked to see more games tested but but at the very least we've confirmed that non-gaming performance will be worse and that there at least some benefits for gaming performance like they claimed.
Their 6 selected titles (tomb raider being one of them) are a little AMD biased, but AMD’s biggest issue in games is coordinating ram access to the cores. A larger cache cuts down on the amount of times the cores need to access Ram and ease off on the schedulers. Regardless it should have a noticeably positive impact on gaming. It won’t always edge out the 12900k in gaming but it will certainly do better than the existing AMD lineup in games, there are a number of media creation or other heavy workloads that the clock speed decrease will be noticeable.
![Zarathustra[H]](https://cdn.hardforum.com/data/avatars/m/9/9298.jpg?1707758471){kind=avatar}
and the AMD chip would get better perfomance with low latency 3800 mhz. if you gonna change one then change the other. nobody running ryzen on cheap 3200 memory!!?
I'm curious where you're seeing 6 games tested, the TH article linked here as well as the one they're quoting only mention the TR title.
If I would try to benchmark Tomb Raider with frame per second right now, without rendering frame at all, how would I do that ?
AMD used 6 games in their slides when they announced the chip. Tomb Raider was one of those 6.
depends on what you're trying to do or the most important thing to you about your pc. if it's gaming performance, well this is the one to get imo. even if it is slightly slower in production/synthetic benches because of a 200mhz clock difference. plus they were using 3200mhz memory. the regular 5800x numbers could be coming from a system that was using the recommended memory speed? everyone knows that's the best way to boost performance on Ryzen.. especially on one with the 3D chip and it's oc lock
Been a couple tests with this. When looking at gaming performance, an average kit of 3200mhz stuff on AM4 is just fine. You can spend a bunch of money/time on something better, and there isn't even a measurable difference when it comes to gaming. People assume benchmark increases with better memory translate to real world performance - It doesn't. The reality is that the 3200mhz kit is already decent increase over the base spec of 2133mhz.
The cache increase compensates for the memory controller. AMD and their many core approach works great for numerous small workloads but struggles when it needs to coordinate all those cores accessing accessing Ram for larger ones. This is an interesting midterm step and I treat it as such.
The CPU isn't even out for another, what, two weeks?
I know some places are already selling them in different countries? I also know that reviewers have them in their hands now.
I don't know what they are thinking. The only thing I can think of is some people spent allot of money upgrading their rig and most wouldn't be too happy if the platform they were on before all of a sudden beat their current one.
guess it depends on who you ask.
The person that ran the tests linked to this post as an explanation of their testing methods though it's worth noting that these test were done at custom low settings too. The short version of the linked explanation is that resolution is the most reliable way to remove the GPU from the equation without also potentially having an impact on cpu usage due to overhead. If that isn't enough to fully remove the bottleneck then start lowering the most gpu intensive settings until the bottleneck is gone.
They could probably swap the GPU's between the test systems, and get results within a half percent of the previous results. But yeah if they want to properly test, the GPU should match.
It might be useful for someone still gaming on say an i7-920, or something 10 years old. But come to think of it, the more useful set of data would have been to do this test across a range of CPU's going back as far as that i7-920. Ensuring each system had fast ram for that system, and just use the same GPU in all runs. Would take awhile, but might show off that CPU Cache a little better. In fact they need to add a 5800X to this test vs the 5800X3D. This data plus data from older CPU's (up thru the newest Intel as well of course),would paint an interesting picture.
Tom's Hardware doesn't know this? That's pretty bad. Or at least, have more benchmarks - comparing one game doesn''t say much.
Hardware Unboxed/Techspot usually does that kind of thing eventually.
Tom's didn't do this test, they're just reporting on someone else's results.