AMD Ryzen 9 9950X

[Daven]

That's extremely subjective. Because this CPU trades blows with the former 7950X, and the performance difference is in the range of the statistical error, and the natural system-to-system performance deviation.

So, extremely pathetic. Maybe it needs faster than DDR5-6000?

View attachment 358916
View attachment 358917

View attachment 358918

View attachment 358919

View attachment 358920

View attachment 358921

View attachment 358923

It's not subjective. It's called cherry picking. You selected to show all the benchmarks with the least performance relative to the 7950X. How cunning of you but I'm not sure why you took so much time to copy and paste the worst data into your comment. Here is the chart for the rest of us not looking for the worst performance but performance across all apps. Take what you like from it but this is at least ALL the data versus the 7950X.

 

[oZ65]

Server first. That's it

 

[Sarajiel]

It is not slow, but I'll play along .. future reviews will use DDR5-6000 CL30 memory anyway

What about testing CPUs at their actual stock settings, aka DDR5-5600 at JEDEC timings in the case of Ryzen 9000?

Personally, I don't mind a 2nd, 3rd or even 4th result with static OC, PBO maxed, or extremely tightened DRR5 timings, but it always feels a bit weird to look at energy efficiency and generational uplifts, if the CPU's IMC is already overclocked.

 

[W1zzard]

What about testing CPUs at their actual stock settings, aka DDR5-5600 at JEDEC timings in the case of Ryzen 9000?

Absolutely no plans for that, there was SO MUCH drama in the past because people kept complaining about slow memory, like 100 posts asking the same thing

 

[Carillon]

Where does he state this? DDR5 6000 c32 is cheap and shouldn’t have any issues running on z790 or b650/x670.

In the comments of the 9700x review https://www.techpowerup.com/forums/threads/amd-ryzen-7-9700x.325081/post-5305257

 

[rv8000]

To play it a bit safe with compatibility, because if I can run CL30 only on Intel and not on all AMD I'm fucked because you will crucify me for running mismatched settings.

It is not slow, but I'll play along .. future reviews will use DDR5-6000 CL30 memory anyway

Okay, makes sense.

It is definitely slower though, sometimes upwards of 3-4% depending on the game/application.

From what I’ve seen most people pick up c32 6000 kits, just seems like a more appropriate baseline. Overall it’s not going to make any drastic changes.

 

[Metroid]

9950x 3% faster than 7950x and yet it costs almost 300 usd more money. The only good thing about this product x last gen is the thermals, anyway, this might be the greatest flop in recent times. Basically nothing really to see here. If the 9800x3d ends up worse than the 7800x3d then I will give up on AMD.

 

[Sarajiel]

Absolutely no plans for that, there was SO MUCH drama in the past because people kept complaining about slow memory, like 100 posts asking the same thing

Too bad...
That kinda defeats the purpose of the "stock" run. If you are going for the popular option, why not also run a curve optimizer of -10 by default? According to the AMD super-fanboys, every CPU should do co -30 out of the box...

 

[Crackong]

As a gamer I will keep waiting for the 9800X3D

 

[BoggledBeagle]

As a gamer I will keep waiting for the 9800X3D

What if it is just 3% faster than 7800X3D?

 

[Sarajiel]

9950x 3% faster than 7950x and yet it costs almost 300 usd more money.

Over here in my part of Europe, the MSRP for the R9 9950X is 49% higher than the street price of the R9 7950X.

 

[sethmatrix7]

Performance is great in some apps, but slower than the 7600 in virtualization? What the heck is going on there

 

[john_]

9000 series is by far the best advertisement campaign I have seen until now, for the 7000 series models.

 

[starfals]

What a boring ""new"" generation. At least i got a motherboard supporting 3 generations of CPUs I can't wait for the next 1, we might even get 10% better performance than the 7000 series XD This whole... buy a new motherboard, we support many generations of CPUs seems to be a giant lie for AM5. Yes, im happy with my CPU from the 7000 series, but if the updates are soooo small, why was this new platform so expensive if longevity is a total lie? I was sold on that idea. That i can eventually get a future CPU thats much better than mine and never having to replace the MOBO. What will the 10900X will do? 10-20 more than what i got now? So? I was hoping to get at least 50-60% more after 3 generations... not 10-35 lol. If that's the case (why wouldn't it be? do you honestly expect 30% more performance from the next gen stuff in 2026? Especially after this 9000 fail)... i kind of regret buying a motherboard for 500 bucks. These were the only prices back when i bought my CPU. Everything was super expensive, and for what? For nothing. I had less USB and SATA ports, no error code... 15 bios updates to make my CPU finally work... My PC still needs 95 seconds to boot, sure down from 5 mins... but yeah. My 300 bucks laptop boots in 2 seconds, no joke, and it has the old Sata SSDs, not the new kind (samsung 990pro)... that boot in 95 secs lol. Go to the AMD forums, many still complain from this.

Anyways, lets see how good the X3D chips are... imagine if they are only 5-10% better. Meh.

 

[Warigator]

The gains are so small it's very boring and very disappointing. Two years and they offer almost the same thing...

It's the repeat of Intel's 2-4 cores every year with +20-25% performance or efficiency gains every two years. Except now it's with 6-16 cores.

 

[Daven]

9000 series is by far the best advertisement campaign I have seen until now, for the 7000 series models.

Lol. What's even more funny is that the top CPUs from both Intel and AMD are all within 10% performance of each other and cover two generations and up to four years of time (7950X, 7950X3D, 13900K, 14900K, 9950X). It's like they've hit some kind of wall and performance gains on the client desktop are over.

 

[AphexDreamer]

Multithread efficiency WORSE THAN PREVIOUS GEN?! I wonder how they managed that...

You have to go down to go up again?

 

[napata]

To be honest, Intel's improvements were even worse than these

Eh, it's very similar on average. However Intel released something every year instead so you'd need to look at 2 gens to properly compare.

 

[persondb]

So, extremely pathetic. Maybe it needs faster than DDR5-6000?

For AVX512 workloads it`s certainly memory bandwidth starved but it already can't use the entire DDR-6000 bandwidth(which is about 96 GB/s).

This is almost certainly why it barely had improvements over the 7950X in Y-Cruncher, it just doesn`t have the bandwidth to do more.

 

[Warigator]

Lol. What's even more funny is that the top CPUs from both Intel and AMD are all within 10% performance of each other and cover two generations and up to four years of time (7950X, 7950X3D, 13900K, 14900K, 9950X). It's like they've hit some kind of wall and performance gains on the client desktop are over.

It's likely they need a totally new way of designing and producing processors for any meaningful gains in performance. Perhaps they at least can lower TDP, so we get 14900KS@6.3GHz performance at Ryzen 5600@3.8GHz power draw, this would still be helpful, especially for battery life and maintaining lower temperatures during summer months.

 

[Metroid]

Over here in my part of Europe, the MSRP for the R9 9950X is 49% higher than the street price of the R9 7950X.

Fail.

 

[Warigator]

Eh, it's very similar on average. However Intel released something every year instead so you'd need to look at 2 gens to properly compare.

Q6600 -> Nehalem -> Sandy -> Haswell -> Skylake was about +20-25% gains roughly every two years and Coffee Lake was similar, because of apps not being designed to use more than 4 cores most of the time, 8600K was also usually +20-25% faster than the 6600K after roughly 2 years

 

[Dr. Dro]

Strong AI/machine learning and AVX-512 performance, but it remains underwhelming and a total loser for people wanting to build a gaming system. Removing the outliers, it doesn't shine in productivity, and letting the processor run wild with uncapped power consumption does not improve its performance either.

If Intel delivers on the promises of Lion Cove and Skymont, Arrow Lake will murder this thing. It fails to be better than a 2 year old Raptor Lake processor in many more ways than it ever ought to - I am not impressed. Still, it's priced right (going by MSRP alone, street prices seem another thing entirely right now) and if someone is building with socket AM5 with an emphasis on multitasking performance, might as well just purchase this one over the 7950X.

Well, not that I was expecting anything else after the 9700X review. Fiasco of an architecture for the client segment, this is a server processor through and through. And a server processor for the AI cloud at that.

 

[rv8000]

For AVX512 workloads it`s certainly memory bandwidth starved but it already can't use the entire DDR-6000 bandwidth(which is about 96 GB/s).

This is almost certainly why it barely had improvements over the 7950X in Y-Cruncher, it just doesn`t have the bandwidth to do more.

View attachment 358933

Yea, latency is especially bad. Memory overclocking doesn’t belong in initial reviews, but 6400 c32 2133 FCLK gets the bandwidth for read/copy over 100gb/s and latency down to 56-58ns on dual ccd 7000 skus.

I want one of these 9000 series just to play with, I just wouldn’t keep it.

 

[tvshacker]

@W1zzard Are the prices correct on the table in the first page? The 7900X3D costs 70$ more than the 7950X3D??? (the same values are on the table on the 9900X review)