Review AMD Ryzen 7 9800X3D Review: Devastating Gaming Performance

Page 8 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
TheHerald said:
Completely irrelevant (somewhat I guess) it would be nice if you could share the save files from your tests like PCGH does. Cause testing the same area isn't the same, your save file may behave differently / different time of day / npcs etc. It's just a suggestion for future reviews.
Click to expand...
I published a "how we test GPUs" thing a while back... not sure many people bothered to read it, let alone actually doing any testing. It's very much a situation where the effort required (to share and demonstrate how to test) versus the benefit (people being able to copy what we're doing) usually doesn't make it worth our time to share extended details. It's literally the 1% of the 1% (of the 1%) that might actually try our saves and benchmarks in my experience. LOL

The reality of benchmarking games is that, for anything that doesn't have a built-in benchmark, most outlets just find something that seems to work fine for what they're doing. And since no benchmark is perfect — you can't represent the entirety of a game with a 1~2 minute benchmark run, or even a 30 minute run — they're all just approximations regardless.
 
  • Like
Reactions: -Fran-, helper800, COLGeek and 1 other person
JarredWaltonGPU said:
I published a "how we test GPUs" thing a while back... not sure many people bothered to read it, let alone actually doing any testing. It's very much a situation where the effort required (to share and demonstrate how to test) versus the benefit (people being able to copy what we're doing) usually doesn't make it worth our time to share extended details. It's literally the 1% of the 1% (of the 1%) that might actually try our saves and benchmarks in my experience. LOL

The reality of benchmarking games is that, for anything that doesn't have a built-in benchmark, most outlets just find something that seems to work fine for what they're doing. And since no benchmark is perfect — you can't represent the entirety of a game with a 1~2 minute benchmark run, or even a 30 minute run — they're all just approximations regardless.
Click to expand...
And TBF in my opinion, as long as you keep the runs more or less equal and the results seems to either having significant difference in performance (e.g. how the X3D chips generally smokes everything else in a GPU bounded test) or that it didn't defies logic (say, if the i5 is faster than the same gen i9) I would think the deviation is at acceptable level.

If running the i5 vs i9 or even the X3D vs an i9 differs significantly at the same location/scene in game with solely different no. of NPCs or time of day, I would lean more towards it being a poorly coded game which mess things up at certain situations, so given all else equal in settings and location, or even version of the game are the same, any of those benchmarks will be reliable and worth referencing.

One kind of bad benchmark I would always keep some doubt on are the "tuning" or overclocking tests. As those are guaranteed (at least by vendor warranty) to be working, doing hardware purchase decision on how well one SKU OCs or how one can run ultra fast ram arn't readily repeatable and will be a trap for consumers reading the reviews
 
YSCCC said:
And TBF in my opinion, as long as you keep the runs more or less equal and the results seems to either having significant difference in performance (e.g. how the X3D chips generally smokes everything else in a GPU bounded test) or that it didn't defies logic (say, if the i5 is faster than the same gen i9) I would think the deviation is at acceptable level.

If running the i5 vs i9 or even the X3D vs an i9 differs significantly at the same location/scene in game with solely different no. of NPCs or time of day, I would lean more towards it being a poorly coded game which mess things up at certain situations, so given all else equal in settings and location, or even version of the game are the same, any of those benchmarks will be reliable and worth referencing.

One kind of bad benchmark I would always keep some doubt on are the "tuning" or overclocking tests. As those are guaranteed (at least by vendor warranty) to be working, doing hardware purchase decision on how well one SKU OCs or how one can run ultra fast ram arn't readily repeatable and will be a trap for consumers reading the reviews
Click to expand...
I think you mean OC stuff isn't guaranteed? Yeah, it's an area that is ripe for abuse by people trying to "prove" superiority of one vendor over another. I've been trending away from OC testing, just because it rarely seems to matter. Another ~5%, with potential instability and not guaranteed or warrantied, on a GPU that might be consuming anywhere from 300W to 450W already? Nah. And while undervolting/underclocking is another possibility, it's the same thing: not guaranteed. I prefer the out of box numbers these days.

Not to say that I'll never do any sort of OC testing, but I just heavily caveat such things with clear statements that a sample size of one means the results are statistically meaningless.

I really dislike trying to test games that have varying time of day, crowds, etc. Some of the games we use do have those (Cyberpunk manual testing, for one example), but since we load from a save spot at least the TOD is always the same. And having run most of these tests hundreds of times, I can say that the run to run variance from crowds and such appears to be extremely small (less than 0.5%). Anything that shows more than about a 2% variance between runs ends up on my chopping block unless I really want to deal with the variability for whatever reason (usually by running more tests).
 
JarredWaltonGPU said:
I think you mean OC stuff isn't guaranteed? Yeah, it's an area that is ripe for abuse by people trying to "prove" superiority of one vendor over another. I've been trending away from OC testing, just because it rarely seems to matter. Another ~5%, with potential instability and not guaranteed or warrantied, on a GPU that might be consuming anywhere from 300W to 450W already? Nah. And while undervolting/underclocking is another possibility, it's the same thing: not guaranteed. I prefer the out of box numbers these days.

Not to say that I'll never do any sort of OC testing, but I just heavily caveat such things with clear statements that a sample size of one means the results are statistically meaningless.

I really dislike trying to test games that have varying time of day, crowds, etc. Some of the games we use do have those (Cyberpunk manual testing, for one example), but since we load from a save spot at least the TOD is always the same. And having run most of these tests hundreds of times, I can say that the run to run variance from crowds and such appears to be extremely small (less than 0.5%). Anything that shows more than about a 2% variance between runs ends up on my chopping block unless I really want to deal with the variability for whatever reason (usually by running more tests).
Click to expand...
Yes, I mean OC stuff isn't guaranteed, be it CPU or ram outside of the XMP spec at most, better not too far from what the IMC of the said CPU is rated for (I personally can accept testing up to DDR5 6400 in dual channel, 1R for RPL or DDR5 6000 running EXPO in Zen 4 just to say) as they are proved to be highly probably to be working, tuning down timings or clocking the 6400 kit to 7200+ or with ultra tight timings.

Same goes for undervolting, since alder lake the intel chips are running so hot that a lot of channels started teaching others to undervolt their chip, while it is beneficial to the thermal throttling (and in RPL case, could be degradation stopping), some bad bins just can't undervolt much or at all, so IMO those tests are bad representation to how likely a general consumer will achieve and not a good idea for reviews
 
  • Like
Reactions: JarredWaltonGPU
JarredWaltonGPU said:
I published a "how we test GPUs" thing a while back... not sure many people bothered to read it, let alone actually doing any testing. It's very much a situation where the effort required (to share and demonstrate how to test) versus the benefit (people being able to copy what we're doing) usually doesn't make it worth our time to share extended details. It's literally the 1% of the 1% (of the 1%) that might actually try our saves and benchmarks in my experience. LOL

The reality of benchmarking games is that, for anything that doesn't have a built-in benchmark, most outlets just find something that seems to work fine for what they're doing. And since no benchmark is perfect — you can't represent the entirety of a game with a 1~2 minute benchmark run, or even a 30 minute run — they're all just approximations regardless.
Click to expand...
You are right, not many people would, im just in a small circle of "enthusiast idiots" that do this kind of crap and share our results with each other...
 
TheHerald said:
You are right, not many people would, im just in a small circle of "enthusiast idiots" that do this kind of crap and share our results with each other...
Click to expand...
I'm working on my revised test suite for GPUs (for 2025, maybe even December of this year), so I'll see if I can capture some stuff and share it for the "interested parties" like yourself. It might be a few weeks...
 
  • Like
Reactions: helper800 and TheHerald
TheHerald said:
Is there any chance of putting this into an ITX case with a tiny pure lock LP type of cooler or the increased power draw makes this a nogo? Kinda thinking of pulling the trigger.
Click to expand...

Unless you're dead set on a particular ITX case, any one that'll fit something like the Noctua NH-D12L would be a good choice. I have built several ITX PCs for people in the CoolerMaster NR200 using either the Noctua or Thermalright 120mm air cooler. 5900x, 12600K, 9700x (OC) all kept cool with these.
 
  • Like
Reactions: TheHerald
blkspade said:
Unless you're dead set on a particular ITX case, any one that'll fit something like the Noctua NH-D12L would be a good choice. I have built several ITX PCs for people in the CoolerMaster NR200 using either the Noctua or Thermalright 120mm air cooler. 5900x, 12600K, 9700x (OC) all kept cool with these.
Click to expand...
I figured that there is no ITX that can fit the hardware I want (big 1200w PSU + 4090) so I went for a jonsbo d41, which ended up being a little bit bigger than I originally expected but it's still decently small. It's like 1/3rd the size compared to my main PC (fractal torrent).
 
  • Like
Reactions: helper800
TheHerald said:
I figured that there is no ITX that can fit the hardware I want (big 1200w PSU + 4090) so I went for a jonsbo d41, which ended up being a little bit bigger than I originally expected but it's still decently small. It's like 1/3rd the size compared to my main PC (fractal torrent).
Click to expand...
The NR200 can fit up to a triple slot GPU, and SFX PSU do come in that wattage. Not that you would end up needing such a beefy one. I tend to stick to 850w-1000w. The max I pull from the wall gaming with a 7950x3D and 7900XTX is ~550w, in a full tower. Can't even plug enough things into an ITX system to pull much more power.
 
Cooe said:
Yes, IMO I would 110% return it!!!

Microsoft Fight Simulator 2020 (& likely 2024 as well) aren't particularly heavily multi-threaded (anything >=6c/12t is good), but ARE ABSOLUTELY as cache, IPC, & clock-speed hungry as games can basically get!!!

Aka, the Ryzen 7 9800X3D is going to be notable better than your current Ryzen 9 7950X3D in literally every way that actually matters! 🤷 (Framerates, frame-times, performance consistency, future-proofing for future GPU upgrades, etc...)

If you absolutely NEED >8x-cores for something outside running flight simulators though, then you might not have any choice as the 16-core/32-thread Ryzen 9 9950X3D isn't coming out until January, but if you don't, THEN MOST DEFINITELY RETURN/EXCHANGE!!!
Click to expand...
I wouldn't make that bet about MSFS2024. It seems they intend to rely more heavily on online Map streaming, which is going require doing decompression in real time. I believe that is why higher core count CPUs are in the Ideal Spec. Even other games that do dynamic shader compilation lean really hard on 8 core CPUs.
 
blkspade said:
The NR200 can fit up to a triple slot GPU, and SFX PSU do come in that wattage. Not that you would end up needing such a beefy one. I tend to stick to 850w-1000w. The max I pull from the wall gaming with a 7950x3D and 7900XTX is ~550w, in a full tower. Can't even plug enough things into an ITX system to pull much more power.
Click to expand...
For sure I could make it work but I already have a spare PSU so I didn't feel like buying a new one, sadly the one I have is kinda massive even for ATX standards so I had to go for a bigger case. I want to build a tiny SFX but with the way gpus are going...oh well 😱
 
TheHerald said:
For sure I could make it work but I already have a spare PSU so I didn't feel like buying a new one, sadly the one I have is kinda massive even for ATX standards so I had to go for a bigger case. I want to build a tiny SFX but with the way gpus are going...oh well 😱
Click to expand...
Did you look at the Fractal Design Define 7 Nano for an ITX case? It can take a PSU up to 200mm in length if you don't have the HDD-tray and modular air duct installed. It also allows for a GPU up to 2.8 slots wide.
 
  • Like
Reactions: helper800
  • Like
Reactions: helper800 and thestryker
-Fran- said:
More than likely: socket defect and bad installation, looking at the pictures.

Still, rather be cauteous and wait for more information.

Regards.
Click to expand...
If it's pebkac then tough luck, if it's socket defect from the factory oh well, let's hope it's just a small batch.

Receiving mine tommorow, I'm actively praying to Lisa to send me a good one.
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom