In China it appears to be selling as a regular retail product that comes in a box with cooler and all. In Australia, you can only purchase the 3500X in a pre-built PC. Which still begs the question, how well does it perfom and what’s changed from the R5 3600?
We’ve heard through channels that in order to acquire the 3500X from AMD to use as a system integrator, there are quite a few hoops to jump through. You have to let them know your plans, what kind of system you intend to build, and the volume you expect to move. We’ve also heard they’re only being sold in large volumes, so 1,000+ chips. Anyway, the Ryzen 5 3500X has been available for purchase at AliExpress for over a month and the second they were listed we snapped one up for what amounted to $157 delivered. It’s been sitting on our desk for a few weeks now, so we thought we better test this thing out or it was just never going to happen. Let’s quickly go over the specs of the 3500X…
Like the 3600, it packs half a dozen Zen 2 cores that come clocked at 3.6 GHz for the base. Depending on the workload they can clock as high as 4.1 GHz out of the box, just 100 MHz lower than that of the 3600. The key difference is the lack of SMT (simultaneous multithreading) support for the 3500X, where the R5 3600 packs 6-cores / 12-threads, the 3500X features just 6 threads, much like the Core i5-9400F. Is dropping SMT support for a ~30% discount from the R5 3600 worth it? (in Australia, again, this will vary on your location). For US-based shoppers it’s just a 20% discount, so we doubt that will be worth the savings, in any case we will know shortly. For testing we have the usual battery of gaming and productivity benchmarks.
All CPUs have been tested with 16GB of DDR4-3200 CL14 memory and a RTX 2080 Ti to reduce any potential GPU bottleneck. All the latest Windows 10 patches, games, drivers and BIOS updates available at the time of testing have been installed. Let’s get into the results.
Benchmarks
First up we have the Cinebench R20 multi-core results and here the 3500X doesn’t look great. Yes, it’s still faster than the Core i5-9400F, but it’s slower than even the first-gen Ryzen 5 1600X. It is 6% slower than the R5 2600 and a whopping 27% slower than the R5 3600. Naturally, for those running wanting to run core-heavy applications, the 3500X might not be the best fit.
The advantage the Ryzen 5 3500X has over the similarly priced first and second-gen Ryzen parts (1600 and 2600 are 6-core/12-thread processors) is the improved single core performance offered by those Zen 2 cores. Thus in single-core Cinebench it’s basically on par with the Core i7-8700K, making it 25% faster than the R5 2600 and even 9% faster than the Core i5-9400F.
AMD’s SMT implementation for first and second-gen Ryzen wasn’t particularly useful for compression and as a result the 3500X is able to match the 1600 and 2600 series processors. That said, it is 22% slower than the R5 3600, so a reasonably large drop in performance there.
AMD’s SMT has always worked very well for decompression work, so the 3500X gets blown away as it can only compete with the Intel Core i5 processors in this test. This time it was 36% slower than the 3600 and 25% slower than the 2600.
As we saw when testing with Cinebench, for those looking for a budget productivity processor, the 3500X isn’t it. In V-ray benchmark we see that the 3500X was 6% slower than the old Ryzen 5 1600 and almost 20% slower than the R5 2600.
Things look even worse when running the Corona benchmark, much worse in fact. Here the 3500X is only able to match the Core i5-9400F, which means it’s 27% slower than the R5 1600 and 43% slower than the R5 2600. When compared to the R5 3600 we’re talking about a 60% increase in render time, so creators on a budget are much better off with a Ryzen 2600 or similar.
We find a similar story when testing with Blender, the 3500X took 14% longer to complete the workload than a first generation Ryzen 5 1600, meaning it was much slower than the R5 2600 and R5 3600.
The removal of SMT support means the 3500X loses quite a bit of its power efficiency. For example, where it was 14% slower than the first-gen R5 1600, total system draw is just 10% lower, That figure includes all system components, but still you’d expect the newer 7nm part to fair much better. Yet again, power consumption is hardly a concern here as 141-watt total system load for a core-heavy workload is nothing.
Gaming Benchmarks
Time for some game benchmarks and we’ll start things off with Assassin’s Creed: Odyssey at 1080p using the very high quality settings. As a reminder, we’re using a high-end RTX 2080 Ti for testing. Here the 3500X is comparable to the Core i5-9400F along with the first and second-gen Ryzen 5 parts. That’s to say performance is good, but we get the feeling it won’t be long before 6-thread processors such as the 3500X and 9400F go the way of the Core i5-7600K.
ACO does become a little more GPU bound at 1440p and this helps the 3500X out as it’s not able to match the Ryzen 5 2600X, and really deliver similar performance to the R5 3600. Performance from the mid-range to top end is pretty similar under these more GPU limited test conditions.
Battlefield V is an example where 6 threads or even 8 threads are becoming insufficient. Here the 3500X delivered a similar level of performance to a Core i7-7700K and Core i5-9400F, and although the average frame rate looks quite good it’s the 1% low performance that suffers. Granted the game is still very playable and significantly better than the quad-core 7600K, but you will see noticeably larger frame dips when compared to the 12-threaded and better CPUs. The 1% low performance was improved by 11% with the R5 1600 and 14% for the R5 2600. However, it’s the R5 3600 that really steps things up, boosting 1% low performance by almost 30%.
The situation doesn’t change at 1440p, rather things just get worse for the 6-core processor. Here the 3500X still suffers a big dip in 1% low performance and as a result the R5 1600 was 25% faster, so not a good result for the affordable 3rd-gen Ryzen processor.
Shadow of the Tomb Raider isn’t quite as CPU demanding as the first two titles, and as a result the 3500X performs quite well. It’s only able to match the 2600X and 9400F, but overall quite a good result.
The margins are similar at 1440p, the 3500X does slip slightly behind the 2600X and 9400F, but again overall performance was decent.
The Ryzen 5 3500X also struggles a bit in The Division 2, delivering poor 1% low performance when compared to even the first-gen Ryzen 5 1600. Again we see that the average frame rate performance is quite good, but it’s the 1% low performance that you should focus on here.
Even at 1440p the 3500X falls behind the R5 1600, albeit by only a few frames, but you wouldn’t expect a 6-core Zen 2 processors to fall short of a first-gen part.
Far Cry New Dawn is the least CPU demanding game we’ve tested so far, so unsurprisingly the 3500X doesn’t suffer poor 1% low performance here, in fact it’s a little better than the 3600 in that regard. Overall performance was the same and not a great deal better than the 2nd-gen Ryzen.
At 1440p we see the same behavior, so for older and less demanding titles the six-threaded 3500X will be perfectly fine, no surprises there.
Hitman 2 played well on the 3500X, certainly well enough despite the 1% lows taking a bigger hit than what you’ll experience on a modern 8 and 12-thread processor. Performance overall was comparable to the Ryzen 5 2600, at least when testing at 1080p.
Even at 1440p the weaker than expected 1% low performance persists and although average frame rates are similar to the 2nd-gen 2600 and 2600X, the minimum frame rates are much lower.
Finally we have Total War: Three Kingdoms and here the 3500X is comparable to the first-gen Ryzen 5 1600X, so the 3600 was much better when looking at 1% low figures. Here the 3600 was almost 40% faster which is a massive performance uplift thanks to the use of SMT.
Once we move to 1440p where the GPU is the most important factor, all margins are nullified and here the 3500X performs much like any other CPU in our chart, it’s just the R5 1600 that falls a little off the pace here.
Is It Worth It?
That’s all the blue bar graphs we have for you (almost!), now it’s time to decide if the Ryzen 5 3500X is worth buying, should you come across it. For those buying today, the best price we’ve found online is $155 and as we said earlier, that makes it about ~20% cheaper than the R5 3600 in the US. For fellow Aussies, it’s about 30% cheaper. Depending on pricing in your region for first, and in particular second-gen Ryzen, the 3500X may or may not be worth it. In the States, the Ryzen 5 2600 can be had for just $120 and frankly at that price it’s a much better deal, and that’s regardless of what you plan on doing with your system.
For productivity tasks the R5 2600 is clearly the better choice as it was faster than the 3500X in every single core-heavy workload we ran. Couple the better performance with the lower price and you have an obvious winner. Even when it comes to gaming, the superior IPC of the 3500X’s Zen 2 cores can’t make up for the lack of SMT support and to better illustrate that, here’s a look at some updated cost per frame graphs.
Here we can see when looking at frame rate performance at a 7-game average, the 3500X was only able to match the second-gen Ryzen 5 2600. You’re looking at spending 6% more per frame which makes it a worse value choice for US and Australian gamers right now.
Moreover, if you only had the choice of a 6-core/6-thread processor for about $150, the 3500X or the 9400F, we’d go with Intel every time for gaming. But since the Ryzen 5 2600 exists, that’s the obvious choice here. For those concerned about gaming performance not just today, but in a year or two, the 3500X makes even less sense. If we look at 1% low performance across the 7 games tested you get a good idea of how limited this CPU will eventually be.
In the absence of first or second-gen Ryzen parts with more threads, would the 3500X worth buying? If you’re mostly gaming we’d recommend the Core i5-9400F instead, while for productivity you could go either way. The actual scenario is that AMD is offloading Ryzen 3600 processors that didn’t pass quality assurance to OEMs, and for budget pre-built PCs, this might make sense. For buying off the shelf though, given the warranty issues you might run into with the 3500X, the Core i5-9400F seems like the better choice at $150. As a 6-core/6-thread processor, the 9400F probably isn’t the wisest investment either. The best budget CPU right now is the Ryzen 5 2600 or Ryzen 5 2400G (if you need integrated graphics) or coughing up the extra $50 for the Ryzen 5 3600, which is an astonishing CPU for the money that will no doubt hold up well for years to come.
