AMD Ryzen 9 9950X

[Jism]

Its the fastest desktop CPU that you can buy at this day.

Apart from that, they have a refresh of nodes every once in a while - this is one.

Clocks cant go higher because the tech is towards it's limits.

 

[sethmatrix7]

Was really hoping these chips could easily do 6600mhz 1:1 at least. I already have this speed achieved on my 7800x3D.

You got a good chip then. Most advice I see says 6400Mhz 1:1 is best case scenario.

 

[Dr. Dro]

True but I will go back to the same ol' tried & tested formula ~ don't need to upgrade every gen!

It certainly dissipated any second thoughts I had about dropping a fat stack on a supremely expensive ROG Apex Encore instead of waiting for this thing, that's for sure.

Its the fastest desktop CPU that you can buy at this day.

Apart from that, they have a refresh of nodes every once in a while - this is one.

Clocks cant go higher because the tech is towards it's limits.

It is the fastest* desktop CPU** you can buy today***

*Your mileage may vary, wildly. Lucky that outliers exist to inflate that weighted percentage faster!
**Does not include "special editions" or their own X3D product lineup
***Buying a 7800X3D today remains a better idea if you're a gamer, but don't tell anyone that

 

[R0H1T]

Anandtech seems to have mentioned there were core parking issues in a handful of scenarios. Wouldn’t be surprising tbh.

It's more than just parking issues, linux shows 17.8% uplift across some ~400 odd tests!

AMD Ryzen 9 9950X & Ryzen 9 9900X Deliver Excellent Linux Performance Review - Phoronix

www.phoronix.com

And to adjust that there's always this ~

https://www.softpedia.com/get/System/System-Info/Quick-CPU.shtml

I wonder why no one seems to be tinkering with the boost options on Windows itself, you know to an extent they can be configured as well?

 

[AnotherReader]

I thought about that too, but then the SMT review TPU did answered some of the questions. In theory, the better and more efficiently you utilize your core in single threading, the less room you have for SMT. If branch predictors have gotten so good that SMT is rendered almost useless. Maybe that explains intels approach to remove SMT altogether. Maybe they predicted we will eventually get there?

Another theory is that windows OS scheduler is failing miserably. This is more likely the case as the Linux reviews paint a completely different picture

SMT still works pretty well for many applications. The chart below is from Chips and Cheese's analysis of the 9950X.

 

[Vayra86]

I think I get AMD's plan here. They're saving the planet for us, by releasing products that make you not upgrade every gen.

Intel was doing the same not too long ago and recently. In fact, it seems to be a pretty frequent occurence. Some generations just hit the nail on the performant head, others are still iterative improvements, but they just don't matter so much, other than perhaps paving the way for more goodies in the future.

I honestly don't get why all these panties are in bunches.

It is the fastest* desktop CPU** you can buy today***

**Does not include "special editions" or their own X3D product lineup
***Buying a 7800X3D today remains a better idea if you're a gamer, but don't tell anyone that

Yeah no shit a gaming optimized chip games better. This is such a complete non argument lol.

I said in another topic about this, I think the Great Plan with Zen 5 here is that the non X3Ds are going to be more expressively not the gaming chips, and the X3Ds are. Product segmentation/optimization.

 

[bitsandboots]

I think this makes more sense than the 9700x. It's clearly the leader in some things, and does so with some increased efficiency. But of course its a bit pricey for what it is.
But, looking forward to what the 9800x3d can do, especially with SMT disabled.

 

[Dr. Dro]

I think I get AMD's plan here. They're saving the planet for us, by releasing products that make you not upgrade every gen.

Yeah no shit a gaming optimized chip games better. This is such a complete non argument lol.

LOL, that's the most "glass half full" take ever

btw I don't necessarily think 3D chips are gaming optimized - it just happens modern games like 3D cache due to numerous factors, primarily available memory bandwidth. it's not exactly a new thing, repurposed high-L3 Xeons ran games surprisingly well for their debilitated clocks even back in the day

 

[Vayra86]

LOL, that's the most "glass half full" take ever

btw I don't necessarily think 3D chips are gaming optimized - it just happens modern games like 3D cache due to numerous factors, primarily available memory bandwidth.

It also happens that AMD is pushing that button for the X3Ds. Gaming chips, and for good reason. These things fix a major drawback most other CPUs always have and still have, even if you buy a swanky degrading Intel. Also, the efficiency is just absolutely bonkers.

 

[nexxusty]

You got a good chip then. Most advice I see says 6400Mhz 1:1 is best case scenario.

Oh yeah, she's a star. Doing CL28 with 2200mhz FCLK too. I tested the crap out of it as well. It's actually quite stable.

 

[Ayhamb99]

Welp this is also unimpressive, only 3-4% improvement in Applications over the 7950X??? Also AMD making the same mistake they did with the 7950X with the the unnecessarily high TDP, the 7950X3D had a much lower TDP hence lower power consumption and lost like 3% only in applications performance in comparison to the regular 7950X.

If the 9000X3D Chips are gonna have this kind of marginal uplift, it'll just confirm that this gen is essentially AMD's own intel moment of marginal uplift between gens. If people are gonna lambast Intel (and rightfully so) for the period of disappointing uplift in performance between generations, AMD should not get a pass as well.

Again though, due to Intel being a clownshow right now, AMD had the golden chance to capitalise on the opportunity and deliver a final death blow to Intel, but nope, so far Zen 5 has left a poor taste for me.

 

[oxrufiioxo]

Wow! This is bad

 

[Daven]

I decided to look at the 9950X in the same way I see the outcome of Arrow Lake. Both companies are trying to bring down power usage while maintaining the SAME level of performance. AMD succeeded with the 9950X and I think Arrow Lake will use even less power at the same performance as the 13900k/14900k.

With all the focus on NPUs, SoCs, big data CPUs/GPUs, AI, mobile, etc, this is all we are gonna get for the deprioritized desktop market. At least for now.

 

[Spilly44]

Really looking like a move from AM4 to AM5 would be better with a 7950X than a 9950X.

 

[z1n0x]

Like what light/fiber optics?

Nah. Afaik, optical interconnects are still not ready for prime time. Also, i doubt anything outside Data Centers will make use of them.

I was thinking about TSMC's integrated fan-out. There are other options, but those are probably too expensive for Ryzen.

Just doubling the number of GMI links would be enough. They do it for Zen 4c and EPYC {link to PDF). However, this would have to be coupled with a more capable memory controller.

Same old copper routing. High latency, high power consumption.

 

[AnotherReader]

Nah. Afaik, optical interconnects are still not ready for prime time. Also, i doubt anything outside Data Centers will make use of them.

I was thinking about TSMC's integrated fan-out. There are other options, but those are probably too expensive for Ryzen.


Same old copper routing. High latency, high power consumption.

The quoted power consumption of about 2 pJ per bit suggests that the higher idle power draw isn't solely due to sending data over the external link. Of course, integrated fan-out would be a better option.

 

[Wasteland]

I think the Great Plan with Zen 5 here is that the non X3Ds are going to be more expressively not the gaming chips, and the X3Ds are. Product segmentation/optimization.

Sure, but based on what we've seen of Zen 5, there's no reason to expect the 9800x3d to offer a substantial improvement over the 7800x3d. Or even a noticeable improvement.

To what degree is this "gaming vs non-gaming product segmentation," and to what degree is it Zen 5 just being underwhelming? At least for Windows users, Zen 5 looks impossible to recommend over Zen 4. Too bad Intel is too busy stabbing itself to capitalize on this opportunity. Oh well.

 

[TumbleGeorge]

Yes, extremely long latencies everywhere and especially between chiplets, the old cIOd, including the old RAM controller, the old infinite fabric. And more bottlenecks that I'm too lazy to list.

 

[AnotherReader]

Sure, but based on what we've seen of Zen 5, there's no reason to expect the 9800x3d to offer a substantial improvement over the 7800x3d. Or even a noticeable improvement.

To what degree is this "gaming vs non-gaming product segmentation," and to what degree is it Zen 5 just being underwhelming? At least for Windows users, Zen 5 looks impossible to recommend over Zen 4. Too bad Intel is too busy stabbing itself to capitalize on this opportunity. Oh well.

Analysis of CPU performance counters in video transcoding and code compilation indicates that Zen 5 suffers more from memory latency than Zen 4. I expect the uplift from 3D cache to be higher than for Zen 4. Whether it will be enough to be significant is another matter altogether.

 

[rv8000]

Wow! This is bad

Interesting take. It’s okay when Intel offers a smaller performance increase at more power, but bad when AMD offers the same performance increase at less power. Not to mention non consumer oriented workloads see a healthy benefit on top on some more regular single threaded applications.

Spoiler

Not a rosy release considering current prices, but none of the 9000 are really bad CPUs. I hope you’re going to tell me you don’t have any bias… again.

 

[KarymidoN]

It's more than just parking issues, linux shows 17.8% uplift across some ~400 odd tests!

AMD Ryzen 9 9950X & Ryzen 9 9900X Deliver Excellent Linux Performance Review - Phoronix

www.phoronix.com

And to adjust that there's always this ~

https://www.softpedia.com/get/System/System-Info/Quick-CPU.shtml

I wonder why no one seems to be tinkering with the boost options on Windows itself, you know to an extent they can be configured as well?

idk if anyone mentioned this, but Wendell from Level1techs discovered that running a game as Administrator in windows was boosting the performance, that there is something strange happening in windows, he even said some games where running better on Linux than on Windows native. Something is not right with zen5 on Windows.

 

[Vincero]

My biggest take away from the 9900X/9950X reviews is that unless you have something that is heavily threaded and those threads don't run in to a bottleneck somewhere (e.g. 7-zip / encryption / some server workloads), there appears to be very little reason to get any of the Ryzen CPUs that have more than 1 CCD on.

There is literally hardly any games where that 2nd CCD die is making more than a token percentage or two difference (and is slower in some cases) compared to the 9700X parts. Even the normal productivity / office apps and Adobe media applications, the benefits are minor for the cost.

For sure if you need as much CPU power as you can get then on AM5 the 9950X will be the choice but, in real world application usage for the average user, it seems having that 2nd CCD is a waste of silicon - at this point the 'platform' (be it IO die / infinity fabric / IMC, etc.) can't exploit it. Maybe it's an MS Windows OS limitation... but something is not delivering.

@W1zzard I know it's likely outside of the scope of TPUs normal thing, but it would be interesting to see how well the EPYC products scale performance vs core count - if there is a much better / linear scale then it would suggest that the extra headroom could be exploited if whatever is holding back the desktop can be overcome - in which case a Zen1 Threadripper style product/platform may salvage this for HEDT / bleeding edge enthusiast crowd.

 

[oxrufiioxo]

Interesting take. It’s okay when Intel offers a smaller performance increase at more power, but bad when AMD offers the same performance increase at less power. Not to mention non consumer oriented workloads see a healthy benefit on top on some more regular single threaded applications.

Spoiler

View attachment 358989View attachment 358990

Not a rosy release considering current prices, but none of the 9000 are really bad CPUs. I hope you’re going to tell me you don’t have any bias… again.

Interesting take consideration I haven't been a fan of 13th or 14th generation mostly due to how they are configured out of the box and the E cores. I liked 12th generation though and 13th offered more uplift over 12th than this and is better at gaming even though it's 2 years old. Intel offering meh performance increases isn't a excuse for AMD offering almost 0 uplift over a 2 year old architecture.

The but Intel did this or that really doesn't matter I only expect these to be mildly impressive over 2 year old cpus. Assuming this generation last another 2 years going nearly a half decade at a similar perfomance level isn't good for anyone it wasn't good when Intel did it from 2012-2017 it isn't good now.

I've owned at least 2 ryzen cpus from each of the last 3 generations and 1 from 2000 so Intel really hasn't been on my radar whatsoever I do have a decent amount of hands on time with i7/i9s from the previous 3 generations though.

I agree that although they are terrible when it comes to generational improvements pricing is the biggest issue currently these make the 4060ti 8G blush when it comes to P/P uplifts. AMD also better hope Arrow Lake is a dud it doesn't have to do much over raptorlake to be significantly more impressive.

 

[gurusmi]

I'm still running a 3800x on AM4. I planned to get a X950x on AM5 to substuitude my actual workstation. But i'm still glueless wich generation. As i was able to read the new 950x is better in singlecore performance. But on multicore the 7950x and the 9950x are not that far away from each other. But the prices differ. With that singlecore performance the new 9950x is faster in calculating spreadsheets. But in slicing both have a similar performance. Now i have to define if it is worth to pay additionally for the higher single corre performance. And that question is not that easy to answer.

 

[Hecate91]

Yeah no shit a gaming optimized chip games better. This is such a complete non argument lol.

I said in another topic about this, I think the Great Plan with Zen 5 here is that the non X3Ds are going to be more expressively not the gaming chips, and the X3Ds are. Product segmentation/optimization.

I guess it makes sense from a marketing standpoint, 9950X if you want a productivity/editing system, though I expected at least a 10% increase in gaming.

But, but it wasn't tested in Lunux...
We truly gone full circle with the AMD fanboys for years crying about Intel barely doing any performance uplifts each new gen only for AMD to come and do the same. All the mental gymnastics on full display.

To be fair, at least AMD doesn't require a new socket with every new generation.