AMD Ryzen 5000 Analysis: Failing to Live Up to the Hype

Unfortunately, due to a lack of supply, we were not able to secure any Ryzen 5000 CPU (though we are working on getting a sample in the next few days). But that’s not going to stop me from making an analysis of AMD’s brand new CPUs, based on the reviews of some respected publications. If you read the title, I’m obviously not all that impressed by these Zen 3 CPUs; below I explain why.

The Ryzen 5000 Series

Firstly, what AMD launched, the Ryzen 5000 series. We have included the older Ryzen 3000 SKUs as a comparison.

As you can see, prices are up $50 across the board. However, because AMD has only launched its higher performance variants, these new Zen 3 CPUs are way, way more expensive than the older Zen 2 CPUs. For example, because AMD did not launch a 5700X or 5700, the cheapest 8 core Ryzen 5000 CPU is actually over $100 more expensive than the cheapest 8 core Ryzen 3000 CPU. Ouch. Hopefully AMD plans to launch these cheaper SKUs soon.

One other thing to note is the decreased base clock speed all across the board. This is something we expected based on preliminary data leaked to us; it seems that the IPC increase of the Zen 3 architecture is not entirely free and will raise power consumption a tad, so base clock speeds had to be brought down by 100 MHz.

Speaking of the Zen 3 architecture, that is the whole point of Ryzen 5000. It still uses the same 7nm process which has powered AMD products since 2019, but the architecture is vastly different compared to Zen 2. There are significant improvements inside and outside the core. The most significant improvement of all is likely the enlarged CCX, which is now 8 cores instead of 4 cores.

Why does this new CCX matter? One core in one CCX trying to communicate to another core or use the cache of another CCX is going to suffer a heavy latency penalty, and this situation was common in Zen and Zen 2 because lots of latency sensitive applications use more than 4 cores. L3 latency should be slightly up within the same CCX, but that is a worthwhile trade for Zen 3. Additionally, each 4 core CCX had 16 MB of L3 cache; this cache has been merged to 32 MB. This wouldn’t seem to make much of a difference at first, but now one core has direct access to all 32 MB of cache, which not only avoids a latency penalty but technically doubles available L3 cache.

My expectations were set the moment AMD accidentally revealed its new 8 core CCX and were repeatedly reinforced by our exclusive leaks, my expectations that AMD would beat Intel in gaming performance. AMD’s recent presentation further convinced me AMD finally had Intel where they wanted them. So, did AMD finally win?

Underwhelming Gaming Performance

I am willing to call Ryzen 5000 CPUs the fastest in gaming, but they did not make Intel obsolete by any means. In fact, based on the numbers I’m seeing, Intel’s upcoming Rocket Lake based 11th Gen has a real chance at taking back the top spot because AMD did not meet my expectations for performance, which surprised me when I started reading the reviews. For my analysis of gaming performance, I have relied on the reviews of TechPowerUp, ComputerBase, Hardwareluxx, Anandtech, Eurogamer/Digital Foundry, and Phoronix.

Let’s start off with the reviews of those first three publications on the list. On average in 1080p high framerate gaming, only one of them found that AMD Ryzen 5000 was faster than Comet Lake in any game, which would be ComputerBase. The others had Intel so slightly ahead that you should just consider it a tie. It should however be noted that ComputerBase was using a 3080 while the other two used a 2080Ti, which could create a GPU bottleneck even in high framerate scenarios.

I’m going to set aside the possibility of a GPU bottleneck for a moment and just take a look at the results. Ryzen 5000’s worst cases seem to be in Battlefield V, Metro Exodus, Assassin’s Creed Odyssey, and Far Cry 5. That last game isn’t too surprising because Zen and Zen 2 also got thrashed in Far Cry games. On the plus side for AMD, these losses aren’t too great, they are just slight losses.

As for games Ryzen 5000 does well in, only ComputerBase found any games where Ryzen 5000 had a clear lead: Gears Tactics, Total War Troy, and Valorant. Otherwise, all three publications found Ryzen 5000 tying comparable Intel CPUs. This is not the kind of performance that makes a CPU a clear leader.

Because ComputerBase showed better performance with a faster GPU, I will consider the possibility that there was a GPU bottleneck in the other two. But, at these sorts of framerates (around 100-200 FPS) it’s likely that there is a CPU bottleneck, no matter what GPU was used. So, these three reviews were not decisive enough for me, so I moved on to reviews from more scientific oriented publications: Anandtech, Eurogamer/Digital Foundry, and Phoronix (which tested in Linux).

I’m going to start with Eurogamer/Digital Foundry. Again, I’m using 1080p gaming results. As expected, Ryzen 5000 does poorly in Assassin’s Creed Odyssey and Battlefield V. However, AMD manages some convincing victories in Far Cry 5, Kingdom Come Deliverance, and the Witcher 3. This review is finding victories for AMD where others found ties, and it’s not because this review is using a faster GPU, as it was conducted with a 2080Ti. My best guess is that Digital Foundry’s testing methodology made a clear difference; this could depend on the usage of built in benchmarks or the testing route of a manual benchmark. But overall, it looks like a slight victory for AMD.

Moving on to Anandtech’s review. At 1440p low, Anandtech found mostly ties, a few wins, and a couple of losses for Ryzen 5000, which would make Ryzen 5000 about 10-20% faster than Ryzen 3000. However, Anandtech also tested at many sub 1080p resolutions (like a shockingly low 360p) and in every single title AMD pulled off a win against Intel, usually by a significant margin. So, AMD’s theoretical best is far better than Intel’s theoretical best, but the problem is that we’re talking about framerates of several hundred at extremely low resolutions. Nobody is buying a 5950X and wanting to game a 360p or even 720p. Furthermore, these framerates are far in excess of the refresh rates of most high end gaming monitors, so the only benefit of these higher framerates are reduced latency. At normal settings, AMD and Intel are about tied in Anandtech’s review.

Finally, Phoronix, which tests pretty much everything in Linux. Here AMD is overall somewhat faster than Intel, making this review one of the most positive for Ryzen 5000. Of course, this is in Linux, so it’s not totally comparable to the other reviews I’ve considered in this analysis, but people do game on Linux, and I was desperate to find a review that didn’t just show a tie overall.

To say the very least, I was disappointed with these gaming results. I expected AMD to pull off convincing leads across the board but I found few. Again, this could be down to a GPU bottleneck, because ComputerBase was the only one here which tested with something faster than a 2080Ti. However, it should also be noted that both Digital Foundry and Anandtech still found victories for AMD with a 2080Ti. All of these 1080p/1440p benchmarks resulted in framerates within the 100-200 FPS region, which is ideal for CPU gaming benchmarks; if the framerate on the faster CPU really needs to be increased to 500 or even 1000 FPS in order to produce a gap, are we even benchmarking real gaming performance anymore?

Good Productivity, Synthetic, and Other Performance

As for performance in productivity applications, synthetic benchmarks, and other things, AMD is extending their lead. There’s not a whole lot more to say other than AMD is even faster than they were in 2019. TechPowerUp, ComputerBase, Hardwareluxx, and Anandtech recorded AMD victories on almost every single one of these benchmarks. In some cases AMD was taking the top spot from Intel; in others, AMD was just widening the gap. So not much has changed here between Intel and AMD other than the amount of work Intel has to do in the future.

One important thing to note is that AMD is seriously contesting or perhaps even winning in single core performance. The increased L3 cache latency is pulling Zen 3 down a bit but the doubling of cache plus architectural improvements within the core itself have propelled AMD to the top of many synthetic benchmarks, especially Cinebench R20. In this arena, AMD not only leads in core count but per core performance.

Conclusion

Like I said before, we do not have any Ryzen 5000 samples yet, so our review is pending and so is our final verdict. But just based on third party reviews, the results with Ryzen 5000 are disappointing, primarily in respect to gaming performance. This was the generation where I expected AMD to finally become the clear leader in gaming benchmarks and it didn’t happen.

I know this analysis is probably the only one on the planet that is this negative for AMD and will likely make alot of people mad (perhaps even AMD themselves) but I’m just basing this on the benchmarks of respected publications. It is a fact that Ryzen 5000 usually ties or barely beats Intel 10th Gen CPUs in benchmarks that use typical resolutions and achieve framerates which are realistic for gamers wanting a high framerate. AMD did not make Comet Lake or Zen 2 obsolete in gaming; a 10% or even 20% victory does not warrant that kind of rhetoric.

You can make arguments all you want about reviews not using 3080s or 3090s, or reviews not testing below 1080p, or reviews not testing specific titles. These publications are not biased against AMD and they’re not testing in a way that specifically harms AMD. The fact of the matter is that 5 publications failed to produce results that put AMD in the lead for gaming by a convincing margin.

To be clear, Ryzen 5000 is arguably the fastest CPU for gaming, but it’s only by a hair. If I had a Comet Lake CPU, I wouldn’t upgrade to Ryzen 5000 unless I was gaining more cores or needed a more efficient CPU (which Ryzen 5000 is by a massive margin). Personally, I have a 3700X, and the reviews have convinced me to just upgrade to a 3900X once the price drops. If you’re not gaming at over 200 FPS, you will probably not notice the difference between Zen 2, Comet Lake, and Zen 3.

As for productivity and other sorts of cases, Ryzen 5000 is doing much better. Still, the performance increase is not earth shattering in every case compared to Ryzen 3000. I could see someone upgrading from Ryzen 3000 to 5000, but only if one is also increasing the core count (like getting a 5950X, up from a 3900X). Oh, and of course Comet Lake is the clear loser in this arena, but that was mostly the case in 2019 anyways.

The biggest credit I can give to AMD is that you can finally have the great multi core performance of Zen 2 and the great gaming and single core performance of Comet Lake all in just a single CPU. That certainly has value for the people who need both, not just one or the other. Smart Access Memory support with RX 6000 series GPUs is also interesting, but it’s only supported on 500 series boards (and this is probably why 400 series boards were not intended to support Ryzen 5000 initially).

The thing that Ryzen 5000 does that really disappoints me is the pricing. $50 across the board is too much, at least for gaming. How in the world do you recommend these CPUs for gaming when they often provide very little or no performance uplift at all? The only person I can recommend these CPUs to in good faith are e sports players, who ideally play at framerates far above refresh rates in order to cut down on latency; at these extremely high framerates, Ryzen 5000 is overwhelmingly the victor. But for the rest of the gaming crowd, they can keep cruising along with their Zen 2 or Comet Lake CPUs, or even their Zen and Coffee Lake CPUs.

There’s also some bad news for AMD; Intel’s Rocket Lake CPU is coming. It will most certainly not be winning any points in productivity benchmarks and other applications, but it could pose a serious threat in gaming. I don’t expect Rocket Lake to make Intel the clear leader again but it could seriously dampen AMD’s claims of having strong, leading gaming performance. That of course wouldn’t have been such a big problem if AMD just didn’t overhype Ryzen 5000. But Intel always has room to fail, so perhaps AMD’s position will truly be unassailable.

So, that’s my take on Ryzen 5000. I think I might be the only person on the entire planet who isn’t ecstatic at the idea that AMD is (kind of) winning. But I don’t care if AMD or Intel is winning, I only care if the products are good enough. Ryzen 5000 isn’t bad but what are people missing out on if they don’t buy one? The magnitude of AMD’s achievement today is only great because people say it is.

Liked it? Take a second to support Matthew Connatser on Patreon!