Review AMD Ryzen 5 9600X and Ryzen 7 9700X Review: Zen 5 brings stellar gaming performance

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
>In the first paragraph of the review it says "The processors also deliver class-leading single-threaded performance but still trail Intel in heavily threaded applications." That right there says all you need to know.
Seriously, what do you want here? That's as about impartial a statement there is, that states the obvious.

Part 1: "The processors also deliver class-leading single-threaded performance - This is a true statement.
Part 2: "but still trail Intel in heavily threaded applications." - This is a true statement.

There's no need to go back to a deep dive on why the results are that way. It's really clear. That you see differently, well, that's for you. No need to criticise the author too much.
 
Last edited:
For anyone wondering why tom's testing is so far away from every other reviewers, it's because they use weird ram configurations
image.png
 
I would have really liked to have seen a temperature graph with PBO enabled because based on what TPU showed the lower temps seemed to be mostly related to the better power consumption.

While I certainly wouldn't say these chips are bad by any stretch I was expecting a bit more all around improvement. It seems like in certain things they've got a really good improvement and others none. The multithreaded improvement not being as high as the single is also interesting given the efficiency improvements.

It's going to be very interesting to see how the dual CCD parts end up given the different power profiles.
 
The gaming benchmarks appear to diverge wildly from the testing conducted by others (e.g. Eurogamer, Anandtech, Techpowerup) despite having very similar test rigs.
Yeah, I was noticing that as well. From other tests I watched on YT, these CPUs weren't all that impressive in gaming vs 7000 X3D. Which I expected anyway. But the productivity uplifts are impressive if anything. According to Tom's at least. I'm looking to get at least an 8c/16t CPU this year. But currently it looks like the 7000 series is the better price to performance AM5 choice. (I'm upgrading from a mini PC w/6900hx mobile APU. So any desktop 8c will already be higher performant in my workload, lol)
 
PBO turns it into a different beast, similar power consumption to the previous gen but much better all core frequencies, 4.4/4.5 GHz to approximately 5.3GHz. (The friendly German overclocker demonstrated this).

PBO off it looks a little meh compared to the 7000 series. It will be a good upgrade from my 3900x.
I'll be upgrading from a mini PC based on a 6900hx... So I'll be seeing a huge performance gain as well! Especially going from a mobile to desktop CPU. Lol
In skimming over the THW review piece, I'm struck by how much subjectivity (of the reviewer's) is layered on top of what should be a straightforward showing of "objective" benchmarks, which would allow the reader to form his own opinion.

As an in-your-face example, instead of the gushing "Zen 5 brings stellar gaming performance," the reviewer could've gone the other way and say that "Zen 5 still substantially trails Intel multicore performance that are two generations old." You see how much slant (and the writer's bias) goes into these pieces.

For me, I'm not a gamer in the "MOAR FPS" sense that these sites primarily cater to (I play strategy games), so perhaps I'm not the targeted in-crowd. But these Ryzen 9K don't make much sense for gamers regardless. As other reviews point out, their bang/buck value sucks. If you want gaming perf, it's X3D. If you want value, either the Ryzen 7K or Intel RPL. Proclaiming "stellar gaming perf" is just wrong on so many levels.

Anyway, this only reinforces my practice of looking at multiple reviews to get a more balanced perspective. From the few I've seen thus far, Ryzen 9K comes off as meh at best.
I agree with some of what you said for sure. I will be upgrading (6900hx based mini PC). But it won't be 9000. I will be getting a 7000 for sure, hoping for a bundle from Microcenter. But even without a bundle, the 7000 is the better choice it seems for my workload and price/performance.
 
Those are the stock settings so it isn't wonky at all.
Tom's testing the memory settings from the CPUs product specification and not just enabling the unsupported XMP/EXPO for all tests seems valid for a CPU test.

It would be different if the baseline AMD profile wasn't also using the officially supported memory speeds.

I would like to see at least one Intel CPU with XMP/Extreme profile enabled to go with the EXPO/PBO test.
 
>I will be upgrading (6900hx based mini PC). But it won't be 9000. I will be getting a 7000 for sure, hoping for a bundle from Microcenter.

Ryzen 9K's forte is power efficiency, so would be suited to SFF builds, if that's your preference. 9700X vs 7700X's price delta ($359 - $290 = $69) can be viewed as a premium for going small. The difference is nominal when considering the cost of the overall build.

I had tended toward SFF for personal use, but now I favor mini-PC more--cheaper, more compact and portable, has better iGPU options, and much simpler build. The trade-off of course is lower performance vs SFF. I'd just do a regular tower build for more perf. The SFF cases I've seen are only incrementally smaller than mATX cases anyway, having to account for the humongous GPU.

Intel is getting the pile-on treatment from the online crowd right now for the 13/14th-gen instability issue. But it still wins hands down in productivity, and 13500 (SFF) or 13600K is the value leader. For the perf-bar-none folks, Arrow Lake is expected to improve upon 13/14, and given the lackluster AMD perf, is worth the wait-and-see.

It's not hard to see AMD's rationale with Ryzen 9K, which is the same as with its Radeon 7K release. They're slow-walking the improvements, because all of their attention is on AI and competing against Nvidia in that market. As an investor, I understand it's the obvious business choice, and would not want them to do any different.
 
...

As an in-your-face example, instead of the gushing "Zen 5 brings stellar gaming performance," the reviewer could've gone the other way and say that "Zen 5 still substantially trails Intel multicore performance that are two generations old." You see how much slant (and the writer's bias) goes into these pieces.
...
Comparing to Intel's last two generations is basically academic at this point as really, any 65W TDP SKU's and above have complete stability issues. We don't know what 13th and 14th gen performance really is until the voltage-fixing BIOS is released as performance could change.

That said, I don't entirely disagree with you on the subjectiveness. I, like many others, was surprised to hear the meh's from GN and other tech tubers and then come here to see "stellar" used. I think a lot of it really stems from the overall performance-per-watt characterstic, not necessarily the raw performance. Then again, also interesting to see PBO make such a big difference this time around when I don't think it did for Zen 4.
 
  • Like
Reactions: KyaraM
Already happened.

IMO, it'll be very interesting to see the socketed version of Strix Point. It's a 4 + 8 configuration, yielding a max of 24 threads. When given a decent power & thermal budget, it could deliver some impressive multithreaded performance at a price point below that of the 9900X.
I'm hoping for a mini PC with Strix Halo myself. Or even MoDT also! (Or is that pretty much the same? Lol). I'm running a 6900hx based mini and I'm hoping for Strix (Point or Halo) versions. That's if I don't just build a new AM5 PC first, lol
 
  • Like
Reactions: bit_user
>I will be upgrading (6900hx based mini PC). But it won't be 9000. I will be getting a 7000 for sure, hoping for a bundle from Microcenter.

Ryzen 9K's forte is power efficiency, so would be suited to SFF builds, if that's your preference. 9700X vs 7700X's price delta ($359 - $290 = $69) can be viewed as a premium for going small. The difference is nominal when considering the cost of the overall build.

I had tended toward SFF for personal use, but now I favor mini-PC more--cheaper, more compact and portable, has better iGPU options, and much simpler build. The trade-off of course is lower performance vs SFF. I'd just do a regular tower build for more perf. The SFF cases I've seen are only incrementally smaller than mATX cases anyway, having to account for the humongous GPU.

Intel is getting the pile-on treatment from the online crowd right now for the 13/14th-gen instability issue. But it still wins hands down in productivity, and 13500 (SFF) or 13600K is the value leader. For the perf-bar-none folks, Arrow Lake is expected to improve upon 13/14, and given the lackluster AMD perf, is worth the wait-and-see.

It's not hard to see AMD's rationale with Ryzen 9K, which is the same as with its Radeon 7K release. They're slow-walking the improvements, because all of their attention is on AI and competing against Nvidia in that market. As an investor, I understand it's the obvious business choice, and would not want them to do any different.
I'm looking into the same thing as you with the Mini PC. I have a Minisforum UM690s (6900hx based system). I have two plans waiting for my upgrade. Either wait for next year hoping to see a Strix Halo mini or build a desktop AM5 system. The advantages of having a desktop CPU/dGPU are huge. But I also really like the Mini PC thing too. So I'm torn... Maybe do both? Lol

Anyway, as of now my choice would be 7000 series for sure. I'm keeping an eye on my local Microcenter for 7000 bundle deals or just good price cuts on the CPU side. I can get a mobo b650 for less than $150. If I went a620 series even under $100 for cheap quick PC build.

The price/performance difference between 7000 and 9000 seems pretty much negligible for my workflow/workload. And both X gens do actually perform just fine on gaming for me. I just need a good stable 60fps on 1440p. (I don't game a lot anyway, lol)

Us Mini PC enthusiasts need to unite!!! Hahaha
 
This is an extremely generous review, given the performance of the chips.

It's clear that AMD all but abandoned the gaming segment (with the exception of their X3D series, but the notable inclusion of Radeon products) to focus on "AI".

The Zen 5 architecture overhaul is intended to benefit server workloads, and causes regressions in gaming.

I can't help but question the motives of people praising its gaming performance in particular. My guess is that in the case of Tom's Hardware, the intention is to get as many Amazon referral clicks as possible.
 
PBO turns it into a different beast, similar power consumption to the previous gen but much better all core frequencies, 4.4/4.5 GHz to approximately 5.3GHz. (The friendly German overclocker demonstrated this).

PBO off it looks a little meh compared to the 7000 series. It will be a good upgrade from my 3900x.

PBO is an overclock preset, and Intel chips have always had more headroom in this regard. The only thing AMD has going for them is the argument that PBO is a "one-click" overclock.

Though a "press this button to get between 0% and 2% more performance" feature is hardly an asset...

And more importantly, if these chips could run those frequencies consistently, then AMD would ship them stock like that.

PBO is nothing more than a marketing device, and AMD's way of sitting on two chairs at the same time.
 
The Zen 5 architecture overhaul is intended to benefit server workloads, and causes regressions in gaming.
Source?

In order to be fair, microarchitecture analysis should compare vs. other non-X3D models. So, I hope that's not what you're talking about.

FWIW, we know Ryzen 9000X3D models are coming. So, if regressions vs. prior X3D is what's got you concerned, this is not really different than what we saw with Ryzen 7000 struggling to best the 5800X3D and will similarly be remedied.
 
This is an extremely generous review, given the performance of the chips.

It's clear that AMD all but abandoned the gaming segment (with the exception of their X3D series, but the notable inclusion of Radeon products) to focus on "AI".

The Zen 5 architecture overhaul is intended to benefit server workloads, and causes regressions in gaming.

I can't help but question the motives of people praising its gaming performance in particular. My guess is that in the case of Tom's Hardware, the intention is to get as many Amazon referral clicks as possible.
As far as the X SKUs go, they're not gaming focused CPUs, as you know. That's what the X3D SKUs have been from the start. So saying AMD has "all but abandoned" the gaming side of things is factually false. As you also know, the 7800x3d is considered the best gaming CPU currently in circulation. With that in mind, why do you think AMD abandoned the gaming market? They still have Radeon as well (even though the next cards probably won't be competitive, they still make them and update software on a regular basis). Not to mention, when the 9000 X3D SKUs come out (probably next year) I think they'll shake up the gaming CPU hierarchy. Best t they could be complete duds, we just have to wait. But AMD's current lineup of gaming focused options show that they have not abandoned gamers.
 
My next CPU will be probably an AMD.

My all-time cpu history :
1989 ATARIST 68000
1991-1995 AMIGA 500,600,1200 68000 - 68EC02
1995 Pentium 120Mhz
1999 Pentium Celeron Mendocino 400Mhz
2003 Athlon XP 2000Mhz
2011 Core i5-2500K
2023 Raptor Lake i5-13500
203? Ryzen ...
....
 
  • Like
Reactions: strobolt
PBO is an overclock preset, and Intel chips have always had more headroom in this regard. The only thing AMD has going for them is the argument that PBO is a "one-click" overclock.

Though a "press this button to get between 0% and 2% more performance" feature is hardly an asset...
According to the review, it's a 14% uplift!

uFQnLhSPmV9nYHeRvKboGd.png


And more importantly, if these chips could run those frequencies consistently, then AMD would ship them stock like that.
Maybe not. It hurts efficiency and means using a higher-end cooler than a user or system builder might want.

PBO is nothing more than a marketing device, and AMD's way of sitting on two chairs at the same time.
It's like a car that gives you a "sport mode" button. It hurts fuel efficiency, which is why it's disabled by default.
 
>That said, I don't entirely disagree with you on the subjectiveness. I, like many others, was surprised to hear the meh's from GN and other tech tubers and then come here to see "stellar" used.

I want to be clear here and state that I don't take issue with this THW piece because it's an outlier to the consensus. I take issue because of its effusive billboarding, on what is an irrelevant angle.

Sure, play up the "power efficiency" bit if that tickles your fancy. But proclaiming "stellar for gaming" when gaming is the lowest cut on the totem pole is going beyond the pale. Again, Ryzen 9K is a poor choice for gaming, going by either performance (X3D is better) or value (Ryzen 7K is better). It's outright misleading.

Every review has some bias. That's par. Misleading reviews are a different thing.

>Then again, also interesting to see PBO make such a big difference this time around when I don't think it did for Zen 4.

That's one thing I wasn't surprised about. With TDP lowered to 65W, it makes sense that removing the TDP would net larger perf increase.


>I'm hoping for a mini PC with Strix Halo myself...I'm running a 6900hx based mini and I'm hoping for Strix (Point or Halo) versions. That's if I don't just build a new AM5 PC first, lol

Strix Point is a premium part. It will show up on premium laptops first, then trickling down to mainstream laptops next year. Large OEMs will get priority allocation, which means small players, ie miniPC vendors who are mostly small Shenzhen companies, will be the last to get their hands on them. In short, you'll be waiting for a while, a year or more.

It'd be interesting to see how Halo will be marketed. It's too power hungry for handhelds, or the largest laptop category which is thin&light. MiniPCs are still way too niche to merit a dedicated part. Best guess would be in the gaming segment, where it can compete with low-tier dGPUs. Nvidia has dominated the mobile dGPU market, so perhaps this is a way AMD can get a foothold in gaming laptops.

>I have a Minisforum UM690s (6900hx based system)

What's your reason for upgrading?
I get that the APUs are laptop first. I'm just hoping (probably in vain) for Halo to be used in minis in a timely manner. But I know it will likely be a while for mini PC companies (Shenzen based like you said) get allocations for Halo. I'm prepared to wait if I need to. Unless I do go DT first.

And the reason for upgrading ... Eh, *shrugs* just for having a better performance if I want to do more than Fallout 4 in the gaming side of things. The 680m iGPU is getting older now. It's going to be 2 gems behind soon. Currently the 780m shows good performance in 1080p (for an iGPU). I do have one other idea though to be able to get to 1440p gaming. And that's to get an eGPU setup. I do know on USB4 (mine has no thunderbolt or Oculink) the full power of a dGPU is seriously handicapped. But it would make for some better experiences in gaming. I've been looking at a 4070 super. That way, if and when I build a DT, I'll have a decent GPU already and cut cost dramatically. I can put that money into a higher tier CPU if I need it, or just save around $800+ USD.
 
Expecting a non-X3D part to surpass an X3D part is simply wishful thinking. Hitting RAM is slow, in terms of bandwidth, and has high latency cycles versus on-chip caches. Versus Intel, you're not going to beat Intel's MT with less cores; they may be small linked quartets (E-cores), but they do push benchmark scores higher. Where Intel (and even AMD's Strix Point with dual-CCX again) will falter is when ops do boundary crossing (P-to-E or vice-versa) to access dependencies, so schedulers will try to avoid that at all costs.

What 3D V-cache shows us is that games want low-latency, very high bandwidth memory for data accesses. This is simply not possible with current DDR5 RAM and any type of HBM is simply too expensive to implement. OC RAM sort of throws power considerations out of the window, so V-cache is the more efficient way to go. AMD simply needs to get V-cache in the mobile chips at some point by moving the entire lineup to chiplets using denser fanout connections, as this will allow iGPU to become its own chiplet.

I think being on N4P muted some of the gains, and I think EPYC's Zen 5c part will show its full potential on N3B. I personally don't think Apple should be allowed to reserve all of TSMC's N3 allocations in the first year, but that's between those two companies and their supply contract terms. I tend to think that's anti-competitive behavior, but Apple is willing to pay and TSMC isn't a charity. It just slows the rest of the industry down, IMO, as others are forced to wait for new lithography nodes.