AMD Ryzen 7 9800X3D

[LittleBro]

Many thanks!

Could you get a screenshot of Ryzen 9800X3D Enhanced as well during benchmark?

 

[Godrilla]

Many thanks!

Could you get a screenshot of Ryzen 9800X3D Enhanced as well during benchmark?

After I posted it I finally got a crash. ( system froze )after 84 hours with set tweak. I am going to clear cmos attempt-35 pbo offset rinse repeat.
Update it could be a conflicting sensor on the Hwiinfo software and Asus's one. The application does pop up with a message that I overlooked the first time if the EC sensor isen't disabled it can cause stability issues.


Update on Ryzen enhanced plus RD2 720p no crashes at -35 PBO.

 

[10tothemin9volts]

Thanks for testing, W1zzard.

So the gaming CPU 9800X3D is only 2.9% (100/97.2) faster in 1440p on average over its predecessor 7800X3D. This is obviously not a reason to upgrade my 7800X3D.
As I currently only play CS2 (64 player community servers only), the 0.3% @ 1440p is..yeah..

Other sources for CS2 @ 1440p: 8.45% (der8auer) and 5.6% (Wccftech). Still not enough and it's going to be even less with my 4070.

Thanks, AMD, for saving me the money. I may invest it in a faster and more-VRAM next-gen GPU (GeForce RTX 50 or RDNA 4 series), so I can load more of Qwen2.5-32B-Instruct-Q5_K_M.gguf / Qwen2.5-Coder-32B-Instruct-Q6_K.gguf (or any other SOTA successor LLMs) into the fast VRAM and get higher tokens per second. CS2 has already become fast enough after the admins disabled player bullets tracking and other decals (because even 4090 player were complaining of the low FPS). Let's hope the prices are going to be reasonable, because I don't strictly need more VRAM because Qwen2.5-32B-Instruct-Q5_K_M.gguf runs at 3.3 t/s to 3.5 t/s on my 12 GB VRAM + 2*32GB DDR5-4800 ECC RAM, which is kinda fast already.

 

[LittleBro]

After I posted it I finally got a crash. ( system froze )after 84 hours with set tweak. I am going to clear cmos attempt-35 pbo offset rinse repeat.
Update it could be a conflicting sensor on the Hwiinfo software and Asus's one. The application does pop up with a message that I overlooked the first time if the EC sensor isen't disabled it can cause stability issues.


Update on Ryzen enhanced plus RD2 720p no crashes at -35 PBO.

Ah, so you're already using CO. I was staring a bit at that 1.18V with 5.325 GHz before. Makes sense now.

You should do a proper stability test. You can't trust games and Cinebench to fully stress your CPU.

 

[Godrilla]

Ah, so you're already using CO. I was staring a bit at that 1.18V with 5.325 GHz before. Makes sense now.

You should do a proper stability test. You can't trust games and Cinebench to fully stress your CPU.

You were right. As soon as I fired up Prime 95 it crashed. after clearing cmoss I found the issue from the lower clocks that are now fixed at high as 5.425 ghz, MC score went up to 1347 with PBO -30. Unfortunately with prime 95 -35 PBO crashed as well. So far -30 ran the test fine although I want to run for longer to be sure. The issue was that there were 2 sections in bios for PBO. One for ai optimizer and one was for manual oc. Both were on -40 lol. I removed the ai setting and the clocks went up to 5.425 ghz instantly.

 

[LittleBro]

Sorry but I doubt it will be stable even with -30. Maybe on two best cores, but at least one core usually can't go lower than -10 or so. Getting maximum out of CO takes time.

What clocks are you getting during test? What's the minimum clock? Could you expand Core clocks section?

 

[Godrilla]

Sorry but I doubt it will be stable even with -30. Maybe on two best cores, but at least one core usually can't go lower than -10 or so. Getting maximum out of CO takes time.

What clocks are you getting during test? What's the minimum clock? Could you expand Core clocks section?

Yes the -30 crashed as well. I ran Prime 95 with -25 CO for past 2 hours and no crashes, Peak temp 80*C and the were around low 4.9 ghz during test. Do you recommended Any other test recommendations? Thank you so much for your feedback.

 

[LittleBro]

Depends on what you consider stable. I run CPU stress test for 24 hours, then I call it stable. Some people say 1 hour in OCCT is enough and there's no chance that you will normally stress your CPU more. I would pick at least 2 different stress tools. OCCT can report computational errors, which also points to instability. System will not freeze or crash, but any shown error represents failed test. I'd go for OCCT 60 min. test repeated 3-times (or straight 180 min. if you purchase license) and then Prime95 for 12 hours. You may also consider tool called y-cruncher.

 

[BoggledBeagle]

You really do not need 9800x3d, when you are GPU limited.

One commenter wrote:

The combination of an "RTX 4090 with FHD" and an "RTX 4090 with an i3 14100" are both equally crazy.

It is hard to argue with that....

 

[Godrilla]

You really do not need 9800x3d, when you are GPU limited.

One commenter wrote:



It is hard to argue with that....

Nvidia's own slide shows they recommend 4k dlss set to performance in upcoming Stalker2. What is the resolution is being rendered as the base resolution ? Hint it's a cpu bound resolution territory.
Update the video shows significant better 0.1 and 1% lows for smoother gameplay.

 

[tenguely]

Hi, hello.

I'm on 5900x right now (and 3070) and playing VR games (Valve index) angry me, because in many games i have stuttering (not all, but low fps is bad).
"Internet" told me that is because of my CPU have very low 1% and 0.1% in FPS, and that x3d chips have very good 1% lows....
Do you think it is worth changing my PC to AM5 with 9800x3d in the future? Like 2025 maybe...

 

[Godrilla]

9800x3d vs i9 14900ks head to head. It doesn't look good for the i9 ks.

 

[TSiAhmat]

Hi, hello.

I'm on 5900x right now (and 3070) and playing VR games (Valve index) angry me, because in many games i have stuttering (not all, but low fps is bad).
"Internet" told me that is because of my CPU have very low 1% and 0.1% in FPS, and that x3d chips have very good 1% lows....
Do you think it is worth changing my PC to AM5 with 9800x3d in the future? Like 2025 maybe...

Not sure if that CPU will solve all your Problems, also pretty expensive (CPU + New Motherboard)

Maybe try the 5700X3D? It's still on the same platform -> No new Motherboard & cost most of the time under 200$/€

And you have a 3070... i think that a 9800X3D is overkill even with an 4080 (and 4090 on 4K-Res) in games which are not cpu-bound.


BUT i don't have any experience with VR so take that with a grain of salt.

 

[LittleBro]

AMD's CPU sales are miles better than Intel as 9800X3D launch numbers published

The first sales numbers for the Ryzen 7 9800X3D have been posted online and we can see that AMD continues to dominate the CPU market.

www.pcguide.com

x.com

x.com

https://www.reddit.com/r/Amd/comments/1gqzky2

 

[Godrilla]

AMD's CPU sales are miles better than Intel as 9800X3D launch numbers published

The first sales numbers for the Ryzen 7 9800X3D have been posted online and we can see that AMD continues to dominate the CPU market.

www.pcguide.com

x.com

x.com

https://www.reddit.com/r/Amd/comments/1gqzky2

From the looks of it he brute forced the cpu into the socket. Gives a whole new meaning to brute forced approach. lol.
Cannon fodder for userbenchmark.

 

[Imploser]

Temperature testing on this page uses air-cooling, for consistency and to show comparable results. All performance testing on the other pages is done using an Arctic Cooling Liquid Freezer III liquid cooling solution that keeps temperatures well below all limits, to ensure there is no thermal throttling.

Also add Arctic Liquid 420 temperatures to temperatures page please..

 

[wolf]

Hey @W1zzard , bit late with this question, but have you considered going another one of these (linked below) for the 9800X3D? I remember having the distinct thought at the time that a 4090 would continue to benefit from newer CPU's beyond what was the top at the time, so I think it would be interesting to see the gains from something like a 5800X3D, 14900K, or even 7800X3D to the 9800X3D and how it lets a 4090 stretch it's legs even further.

RTX 4090 & 53 Games: Ryzen 7 5800X vs Ryzen 7 5800X3D Review

​

 

[W1zzard]

Hey @W1zzard , bit late with this question, but have you considered

Definitely, but no time, too many things happening right now

 

[wolf]

Definitely, but no time, too many things happening right now

I can appreciate that! Before you know it, it'll be a 5090 scaling review.

 

[Godrilla]

How not to break your 9800x3d by Buildzoid 101.

Shame.

How to fight scalpers? Amazon has the 9800X3D available shipping in January in mid to late January at MSRP.

https://a.co/d/10tr4dT

 

[SL2]

MC stores still have some, for those who live nearby.

• CA - Tustin (in stock)
• CO - Denver (in stock)
• FL - Miami (in stock)
• GA - Duluth (in stock)
• GA - Marietta (in stock)
• IL - Chicago (out of stock)
• IL - Westmont (in stock)
• IN - Indianapolis (in stock)
• KS - Overland Park (in stock)
• MA - Cambridge (in stock)
• MD - Rockville (out of stock)
• MD - Parkville (out of stock)
• MI - Madison Heights (in stock)
• MN - St. Louis Park (in stock)
• MO - Brentwood (in stock)
• NC - Charlotte (in stock)
• NJ - North Jersey (in stock)
• NY - Westbury (out of stock)
• NY - Brooklyn (out of stock)
• NY - Flushing (out of stock)
• NY - Yonkers (out of stock)
• OH - Columbus (in stock)
• OH - Mayfield Heights (in stock)
• OH - Sharonville (in stock)
• PA - St. Davids (out of stock)
• TX - Houston (in stock)
• TX - Dallas (in stock)
• VA - Fairfax (out of stock)

 

[LittleBro]

From the looks of it he brute forced the cpu into the socket. Gives a whole new meaning to brute forced approach. lol.
Cannon fodder for userbenchmark.

How can you brute force CPU into the socket? It can't just partially fit. It can only fit or not from my experience.

Man, some users are artists, not users.

 

[Godrilla]

How can you brute force CPU into the socket? It can't just partially fit. It can only fit or not from my experience.

Man, some users are artists, not users.

From Buildzoid perspective it was installed vertically. Steve from gn bought the cpu and is probably going to do a deep dive. Let's give hime a benefit of doubt. So if it's not user error what's your take on it?

 

[chrcoluk]

Depends on what you consider stable. I run CPU stress test for 24 hours, then I call it stable. Some people say 1 hour in OCCT is enough and there's no chance that you will normally stress your CPU more. I would pick at least 2 different stress tools. OCCT can report computational errors, which also points to instability. System will not freeze or crash, but any shown error represents failed test. I'd go for OCCT 60 min. test repeated 3-times (or straight 180 min. if you purchase license) and then Prime95 for 12 hours. You may also consider tool called y-cruncher.

I tend to not got overboard on these utilities, as they focus on only testing full load, not transients, and push everything too hard for my liking.

For me, any instability is not stable, but generally speaking, if it can run what you throw at it and it runs fine daily with absolutely no signs of instability, then I consider it good.

 

[LittleBro]

From Buildzoid perspective it was installed vertically. Steve from gn bought the cpu and is probably going to do a deep dive. Let's give hime a benefit of doubt. So if it's not user error what's your take on it?

I do consider this users error, I'm just amazed.

I only touch PC components (CPU, mobo, GPU when my case is laying on the large side (motherboard tray is positioned paralelly to table's surface). It is the only way to do modifications and does not stress the components.

Once I saw how one dude was installing twin tower CPU cooler while the motherboard was in upright position to the ground. There were 2 screws to fix that tower to the base. He was tightening that small screws in order to bring tower to CPU's surface and fix it. Using that 2 small screws to sit bring closer and sit that 1000g tower cooler ...

And it only takes putting the case on the side to do that absolutely comfortably and not stressing anything.

That user is one lazy dumbass.