AMD Ryzen 7 9800X3D Linux Performance: Zen 5 With 3D V-Cache

[ZeroPointEnergy]

Secondly, the world doesn't revolve around [x86] games.

This is a thread about a CPU that is specifically tailored for gaming workloads, so yes, it does revolve around games.

Thirdly, as a Linux user you could be a little more modest as there are no (AAA) games for Linux at all. The last one was released probably 5 years ago.

Linux has an implementation of the windows API that lets us game basically all (except titles with anti-cheat) windows titles at native performance. This CPU can be fully utilized on a Linux gaming setup.

The Apple processors have a completely different instruction set. Their graphics chip is tailored to their own proprietary graphics API that virtually no one uses. It's a complete joke to even mention those products in the context of this thread. They are not even remotely competitive.

[DumbFsck]

What exactly would be so better about it? ARM isn't some dark magic and Apple is just licensing ARM. The only thing this chips are better at is power efficiency, because this platform was mainly used in power restricted use cases and hence a big junk of the engineering went into that aspect.

This just isn't true. Firstly because Apple extended the ISA, so it is different than ARM from the get go (and yes this is something that holders of the specific license can and do do, not only apple). Second what they did with UMA and Cache Hierarchy was different than what basically anyone else did (or do!) and gave them (at least at the time) best in class latencies, add to that that their memory controller is 100% custom.

Then with their different GPU architecture than Mali or afreno or whatever, they get closer linear scaling with GPU tiles they add to the package. And IMO they had the best way to glue processors together until Intel made foveros work (it is foveros right? I can't remember if foveros is 3D stacking only and there's another name to gluing together) (I recognize how good infinity fabric is and how innovative it was). IMO I'm benefits for the consumer I'd say foveros > "ultra fusion" > infinity fabric >>>>>> not doing anything of the sort.


Lastly I'd say Apple chip designers were very good in deciding what accelerators to add and created some nice designs, like the media accelerators and all the work with the ISA and circuits and accelerators to make Rosetta emulate x86 able to (on Apple Silicon) perform faster than any non x86 architecture ever could.


Thermals, power, etc is also good I guess.


So what would be better, you ask? If premiere or whatever can export videos as fast as in some ridiculously more expensive CPUs (judging price from the fact that a Mac mini can cost less than a top of the line CPU, and this is true from the M1) and custom GPU benchmarks and TFLOPS or whatever run very well as well, if games or other software not made for M CPU/GPUs or ported with their compatibility layer can run "well" - if they were made or optimized for it I *guess* they'd also get very high frames/USD compared to other architectures.

All the x86_64 CPUs, no matter if Intel or AMD completely outclass the Apple stuff in almost every regard, including performance, newer hardware support, more and faster connectivity. It's not even close.

You just fell for the Apple marketing, that is basically all.

Laughable.

[_ReD_]

Secondly, the world doesn't revolve around [x86] games.

The computer gaming world does, and this CPU (not the M4) is the freaking king of gaming.
Sorry, if it's not much to you.

Thirdly, as a Linux user you could be a little more modest as there are no (AAA) games for Linux at all. The last one was released probably 5 years ago.

And who cares?
Since Valve blessed Linux with Proton, I don't really see what you're blabbing about.

Fourthly, a user of the OS with no games tells me something about Apple which actually has a ton of modern AAA titles.

WOW! That's precious. A ton??? Let me tell you, I do have an Apple and I beg to differ.

[yump]

People complaining about power efficiency are barking up the wrong tree.

120 W / 65 W is a ratio of 1.85. To just match, not even beat, the perf/watt of the lower-powered predecessor when you have that kind of disparity, you need something like a 1.85^(2/3) ~= 50 % architectural performance advantage. That just doesn't happen anymore. The thermal improvements that let the chip pull over 120 W without slagging itself are the technical achievement.

In single-threaded and lightly-threaded tasks, the ECO power limit makes almost very little difference to either performance *or* power, because those tasks don't hit the power limit. If you want the 9800X3D to be as efficient as the 7800X3D, limit the boost clock, not the power.

[avis]

A processor who can't run 99.99% of all games natively. Ok Apple fanboy

I'm not an Apple fanboy, I've owned zero Apple devices in my entire life.

Secondly, the world doesn't revolve around [x86] games.

Thirdly, as a Linux user you could be a little more modest as there are no (AAA) games for Linux at all. The last one was released probably 5 years ago.

Fourthly, a user of the OS with no games tells me something about Apple which actually has a ton of modern AAA titles.

Lastly, get lost with your name calling.

[ZeroPointEnergy]

How good the *hardware IS* would be more noticeable if *some* software support was different, that gaming part.

What exactly would be so better about it? ARM isn't some dark magic and Apple is just licensing ARM. The only thing this chips are better at is power efficiency, because this platform was mainly used in power restricted use cases and hence a big junk of the engineering went into that aspect.

All the x86_64 CPUs, no matter if Intel or AMD completely outclass the Apple stuff in almost every regard, including performance, newer hardware support, more and faster connectivity. It's not even close.

You just fell for the Apple marketing, that is basically all.

[tenchrio]

Zen5 3D:

1. A blatant rip-off price-wise, "just" $480 for the CPU alone, not counting cooling (and lots of reviewers ran it with liquid cooling/AiO)

2. A pathetic gaming performance increase:



3. A horrible decrease in power efficiency:



AMD fanbase? A wonderful CPU!!! MY NEXT CPU FOR SURE. INTEL DESTROYED AND RECKED!

Reality:

• A pathetic showing vs Apple M4 Pro
• Extremely overpriced; costs as much as Sony PlayStation 5
• Barely any progress in 2 years for the vast majority of people
• A regression in terms of power efficiency despite using a more advanced node
• Continues the tradition of desktop AMD Zen CPUs idling at 25W+

Trump becoming a president again is now not the worst news of the day.

All of the points you make also apply to the i9-14900K and even UC9 285K to an even worse degree.
1. The launch price of the i9-14900K was $589 same for the UC9 285K, both more expensive than the AMD chips (also no cooler, almost as if it is normal these days on higher end CPUs).
2. Not only are they more expensive they aren't even topdog in the charts you provided, also the generational uplift from the 13900K to the 14900K is 0.8%, almost 1/10th the difference between the 7800X3D and the 9800X3D. The UC9 285K does even worse and loses 5%?! It is as expensive as the 14900K on launch. And they still both lose to the 7800X3D, at least the 9800X3D could beat that.
3.Same story the i9-14900K is also less power efficient compared to the i9-13900K, the UC9 282K at least gets a bit more power efficient but that is of course ignoring that you are losing performance. And even here they lose out to both the 7800X3D and the 9800X3D.
And before you go "Uhh bUt ThE i9 =/= Ryzen X800 CPUs", the same shtick applies to the i7 CPUs (it's even worse, the i7 13700K is only 0.1% behind the i7 14700K in the chart you provided and the launch price even went up, it is at least cheaper than the 9800X3D but that is kinda what you expect with literally worse performance).

"ZOMG SO BAD, clear rip-off! Don't buy!! The imaginary AMD fanboys I am making up are wrong because AMD can't beat AMD!!!"
As if in the cases you provided AMD isn't just competing against itself for the top spot (not to mention it still beats Intel in terms of bang for buck) and as if almost every comment preceding yours isn't people pointing out that yes the performance is better but power draw is larger, really making one wonder where your vitriol and AMD hate come from.

Not to mention how nonsensical your comparisons are, like extremely overpriced compared to the PS5? Ah yes let's compare the top performing gaming chip that is meant to pull every FPS possible with high end GPU hardware so Cyberpunk 2077 on 8K is playable to the console whose $700 pro version had to remove the 8K label from its packaging. Man it is almost as if the CPU offers more performance compared to the one in the PS5 (Pro even) which in turn explains the price difference.

Even bringing up the Apple M4 Pro, a chip we barely have benchmarks for and most are synthetic or from Apple themselves and wasn't your entire argument that it was too expensive? The M4 Pro Mac Minis start at $1400 and that is with 24GB of RAM and 512GB SSD storage both of which can't be upgraded once you order it as they are soldered on (people with a 7000 series CPU can upgrade to the 9800X3D if they want to which is a lot less expensive than the upgrade cost from an M2 Ultra Mac Studio as that is buying an entirely new Mac mini, god forbid you bought one that doesn't have enough RAM). Not to mention in the synthetic benchmark it barely beats the M2 Ultra who in turn could not actually beat the Ryzen 9 7950X in more real life CPU benchmarks:
image.png
Which is ignoring the fact that most of your previous arguments were on gaming (hilarious as you forgot to mention that the Techpowerup article you got those benchmarks from call it "The Best Gaming Processor") in which case.... really you argue for a Mac?

Tomsguide review of the Mac Mini 4 Pro has it scoring 83FPS @ 1080P in the Shadows of the Tomb Raider benchmarks, I had to look up RX7600 reviews using 12th gen Intel CPUs for the test bench (specifically the one from PC Mag) and even there the RTX 2060 (again paired with an Intel 12th gen I9) managed to score 101 FPS.
Now I don't know what to call the M4 Pro losing to the combination of a 3 year old CPU that is beaten by last gen entry level CPUs like the 7600X (again in the charts you provided) and a 5 year old entry level GPU by about 20% but "pathetic showing" could work.

[sophisticles]

Didn't you know that the president has a little nob in their office that controls inflation? And there's another one that controls the price of gasoline? I'm sure it's not more complicated than that, like impacts from extended periods of crazy low rates to try and ward of a recession during the global pandemic, OPEC is fake news, etc.

Whether rightly or wrongly the leader of an organization always gets the credit or the blame for most of what goes right or wrong.

The CEO is a company is a genius when the company does well, a moron when it does lousy.

The coach is a genius when the team is winning and doesn't know what he;s doing when they are losing.

But presidents do have tremendous influence on inflation. oil prices, the economy in general via the policies they implement.

[skeevy420]

It obviously depends on what you value, but I think most 9800X3D buyers will view that as a pro, not a con.

As in, they don't really care about power use as long as it's not too crazy, and the way the prior gen X3D chips were thermally restricted was a major design flaw that's now been overcome.

If all you care about is efficiency, though, the 7800X3D is almost certainly going to continue to be the chip to beat. AMD had to bin those really hard to get the best possible silicon in them.

I consider it both a pro and a con. I like that it's an option, but the gains are minimal for the amount of extra power used so it's not like trying to crank the dial to 11 is a great option regarding stock settings. That the user has more control over it is why I think it'll be neat seeing what the undervolters can achieve, like how it's possible to lower the power of the 7800X3D to get more performance out of it due to the increased thermal headroom which can allow it to run at boosted speeds for longer. I wonder if the 7800X3D will remain the low power performance king or if the 9800X3D just needs the right knobs turned to overtake it. I assume the latter.

[DumbFsck]

So if it was completely different then it would be awesome? Ok.

Is that how you read it?

I though I said the hardware works really well.


How good the *hardware IS* would be more noticeable if *some* software support was different, that gaming part.

So no, not "if it was completely different"....

But I guess you'll misread my post to say whatever you'd like, so why am I wasting my time.. Xkcd386