[FPS Review] - AMD Ryzen 7 9800X3D Review: Impressive is an Understatement

Posted by davidschroth@reddit | hardware | View on Reddit | 33 comments

[-]

Not_Yet_Italian_1990@reddit

If you're doing a new build, sure.

If you're on AM5 already... just wait for something decent unless you've got great resale value on your 7600/7700/7800X3D...

[-]

TheAgentOfTheNine@reddit

Gaming crown in cpu limited games, no doubt. Power draw increases quite a bit from the 7800x3d, which looks to still be the best performance/Watt cpu.

[-]

masterfultechgeek@reddit

Gaming crown in CPU limited games.

50 way tie for first place if you're GPU limited, tied with a 5 year old CPU you can get off ebay for 1/10th the price.

[-]

3600 it was the budget champ at around $100 on sale for a while.
You have to scroll all the way to the bottom to find it. It's definitely long in tooth now but there's a very real scenario where someone paid $80 for a motherboard, $100 for a CPU and got basically the same experience as someone who paid 2-3x as much and is on track to upgrade at about the same time.

https://tpucdn.com/review/amd-ryzen-7-9700x/images/average-fps-3840-2160.png

[-]

Shrike79@reddit

Yeah no.

Unless you are going out of your way to gpu limit yourself at 4k native ultra settings which nobody in their right mind is actually doing you're going to have a way worse experience on a 3600 vs an x3d cpu or just about anything more modern really.

HUB 4k benchmarks 3600 vs 7800x3d at various quality settings.

That video doesn't take into account upscaling either which would only widen the gap even more. You're giving straight up bad advice.

[-]

Strazdas1@reddit

if you are always GPU limited why would you buy this CPU to begin with? This is for people like me where almost everything i play is CPU limited.

[-]

Atheist-Gods@reddit

Level1Techs mentioned that you can run in 65w eco mode on the 9800x3d without significantly impacting the gaming results if you want that efficiency.

https://www.youtube.com/watch?v=KswGlkrNhP0

[-]

popop143@reddit

Basically the inverse of what they did with the initial 9000-series chips

[-]

mi7chy@reddit

60% more power for 6% more performance on Cyberpunk 2077: Phantom Liberty is terrible trade off. Staying with 7800x3d.

[-]

Recktion@reddit

Regret not buying the 7800x3d when it was on sale. Now the 7800x3d is selling above MSRP. Convinced AMD stopped making them else there would hardly be anyone buying the 9800x3d. But now they stopped making an alternative we have to.

[-]

gusthenewkid@reddit

Yeah, that is really rough actually. Crazy low power draw is the best thing about the 7800X3D.

[-]

zopiac@reddit

I wonder how it would do after matching it to the 7800X3D's clock speed. My guess is that it would be a similar "Zen 5%" scenario but that efficiency would match or exceed the 7800X3D.

But that's probably not why people would buy the 9800X3D anyhow.

[-]

conquer69@reddit

But that's probably not why people would buy the 9800X3D anyhow.

I would cap the power. I don't want it pulling more than the 7950x3d outside of gaming. https://tpucdn.com/review/amd-ryzen-7-9800x3d/images/power-multithread.png

[-]

zopiac@reddit

Oh I absolutely would as well, as I currently do on my 5800X3D. But I'm also not interested in buying a 9800X3D at this point, either.

[-]

ImageLow@reddit

I understand average fps was the historical number to look at, but with everything releasing and getting 144+ fps, can we not ignore the 1% lows?

Frankly, I care about stuttering way more than average fps. If something is a stuttering mess, I simply won't consume it.

Reviews like this are almost completely worthless to me because of the lack of 1% measurements.

[-]

Gippy_@reddit

Reviews like this are almost completely worthless to me because of the lack of 1% measurements.

FPS Review is the spiritual successor of HardOCP which was one of the oldest PC hardware review sites. It shut down when its owner (Kyle Bennett) got hired by Intel. (The current site doesn't reflect what it used to be.)

1% lows didn't really get prioritized until Tech Report, another defunct PC hardware review site, came in the data, though they initially called it "frame time beyond X ms" which sounded way more clunky.

If you realize that Brent Justice, the author of this review, is part of the old HardOCP guard, then you'll understand why this review is written the way it is.

[-]

Raiden_Of_The_Sky@reddit

1% has almost nothing to do with stuttering. One of the notorious examples is CS2 which has low 1% fps relatively to avg fps on all CPUs in the world simply because of tick updates every 15 ms. And it usually feels smooth regardless.

Also, stutter is a game code defect which feels very bad regardless. I don't care if it's 0.3 or 0.2 seconds - it's still a stutter, and it's annoying, but these numbers will affect avg 1% fps in figures. That's why this metric is useless.

[-]

keenOnReturns@reddit

? sure not every stutter is from 1% lows, but a good number of them are. And CS2 is a terrible example when you’re getting 200+ fps 1% lows on any modern chip: of course it’s gonna feel smooth. You’re telling me if your 1% lows in CS were 30 fps you wouldn’t feel that??

[-]

Raiden_Of_The_Sky@reddit

? sure not every stutter is from 1% lows, but a good number of them are.

Neither stutter is from 1% lows because it's the other way around - reduced 1% lows are from stutters (and not always, as we can see). That's why this metric is useless, it doesn't show the actual picture of what's happening with a game.

You’re telling me if your 1% lows in CS were 30 fps you wouldn’t feel that??

Well in CS there're occasions when I get as low as 150 1% fps (warmup map aim_botz), but 500+ avg. And it doesn't feel nearly as bad as 150 fps avg. That's what I'm saying.

[-]

keenOnReturns@reddit

thats cause 150 fps never equals 150 1% fps... what are you saying. like does it need to be said that 150 1% fps obviously feels better than 150 fps average?? huh???

I literally dare you to find me a single modern game in existence in which the average fps is exactly equal to 1% fps. wut.

[-]

Raiden_Of_The_Sky@reddit

I literally dare you to find me a single modern game in existence in which the 1% fps isn't lower than avg fps but equal.

Any single player without asset streaming or with flawless one. Doom Eternal for example.

[-]

keenOnReturns@reddit

? are you ok? quick google search literally shows Doom Eternal benchmarks with 1% low being lower: 2080 ti on 1080p Ultra is 171 1% fps and 239 avg fps. Are you ok.

[-]

Raiden_Of_The_Sky@reddit

This is unless you limit fps with RTSS or NVCP. Yes, I am OK.

[-]

Strazdas1@reddit

in CS2 you are lucky to get 60 FPS in a large city even on this CPU. And i mean Cities Skylines 2, not the later released Counter Strike 2.

[-]

gusthenewkid@reddit

It’s not that simple. You can get frametime spikes where the fps doesn’t drop much, but you 100% feel it. That’s why UE5 games feel so bad to play.

[-]

Raiden_Of_The_Sky@reddit

You can get frametime spikes where the fps doesn’t drop much, but you 100% feel it.

Normal people call them traversal stutters, shader compilation stutters and etc, not "1% reduction", "frametimes spikes" or other bs.

[-]

keenOnReturns@reddit

? are you daft? frametime and fps measure the same thing. FPS is averaged which is why it doesn’t drop as much, but that’s why pple measure 1% to capture those spikes. Or to put it another way, if you similarly measure average frametime, those “spikes” wouldn’t be as much either; I suppose you could display a plot of frametime vs time instead of just displaying a raw number? But most reviewers don’t do that + it would be harder to read.

[-]

boobeepbobeepbop@reddit

In the review I watched on Hardware unboxed I think the 1% lows on this CPU were very strong. It wins across the board on 1% and .1% measures.

Of course, that's not to say that the 5800x3d or 7800x3d don't do well enough on those measures that this isn't that noticeable.

[-]

Vegetable-Source8614@reddit

One thing I'd like to see tested is what is the performance of the X3D chip once the V-cache is full? In Factorio for instance for the 7800X3D, it's amazingly fast, but once the game instance exceeds the V-cache capacity it becomes slower than the 14900K.

[-]