Testing Ryzen 9 with SMT On vs. SMT Off

Comparing the latest AMD flagship Ryzen 9 3900X with the Core i9-9900K in 36 games, we found that the 12-core AMD processor was around 6% slower on average in gaming tasks. There were a few selected titles where the Ryzen processor was ahead, but for the most part it was only a little slower.

Overall, it is still performing well for 3rd generation Ryzen, as the gaming deficit is largely negligible while beating the 9900K in all productivity tests.

After we released this feature, readers asked us to retest the 3900X with SMT disabled and essentially convert the 12-core CPU with 24 threads to a 12-core CPU with 12 threads. Many say that this can significantly improve game performance … but why?

We were immediately skeptical because we spent some time testing first-generation Ryzen with and without SMT, and even in 2017, the gains weren't great and often saw a performance regression – in other words, SMT did what it did was supposed to be without great disadvantage.

Since the introduction of Ryzen, Windows and game titles have evolved to better support the multi-core processors. Our recent tests have shown that SMT is effective in gaming performance on both Intel and AMD platforms. We haven't tested the Zen 2 chips with SMT disabled, so we'll give it a try, even if we expect to find the same thing as before.

For this test, the Ryzen 9 3900X was configured with the Wraith Prism RGB box cooler. Otherwise, both the SMT on and off configuration are identical.

Benchmarks

First we have Prey and here we see an impressive 9% improvement in the average frame rate with SMT disabled. Even more impressive was the 13% increase we saw for the 1% low result. The difference to the 9900K has been significantly reduced, although the Core i9 was still the faster chip.

Just Cause 4 also sees a very remarkable 8-10% increase in performance with SMT disabled, and that brings the 3900X close to the 9900K again.

Testing with StarCraft II shows no improvement in performance with SMT disabled when looking at the average frame rate, although we see an increase in performance for the 1% low result.

Testing with For Honor shows a 3-4% performance improvement with SMT disabled and that was actually enough to bring the 3900X ahead of the 9900K in this title. Of course, we have not tested the 9900K with Hyper-Threading disabled, and there is a very good chance that the performance of the Intel processor will also be improved in this title.

With Project Cars 2, SMT performance is slightly improved by 2 to 4%. Even if this is clearly visible, there is no risk for the 9900K of losing the performance crown in this title.

The performance when testing with Total War Three Kingdoms is practically identical. We see an advantage of 1 to 2 fps in favor of the disabled SMT configuration, but that's pretty much within the margin or error.

On PlayerUnknown's battlefields, performance only changes by 1% when SMT is off, so there's nothing to see here.

We see a very small performance regression when testing with The Witcher 3, although again only a difference of 1-2 fps is within the margin or error. So it's fair to say that the performance in this title remains unchanged.

Forza Horizon 4 is a largely GPU-bound game, but we still see a very small performance regression with SMT disabled. Reducing performance by 3% is not a big deal, although it certainly won't help the 3900X approach the 9900K.

A small improvement in performance will be achieved in F1 2019 when SMT is deactivated.

The performance in Far Cry New Dawn remained practically unaffected, at least we see the same 112 fps on average. However, since we have often found that the 1% low performance hit with SMT disabled, we only speak of a ~ 3% decrease in performance, but we don't want to see that.

Strange Brigade tests show an average frame rate change of less than 1%, although we see a 5% reduction for the low 1% number. Therefore, you want to leave SMT enabled for this title.

Interestingly, the average frame rate when SMT was disabled in Battlefield V was practically the same. We see a significant drop in performance for the 1% low result as frame rates dropped 14%. So this seems to be another title where you really don't want to disable SMT.

We went over to Assassin's Creed Odyssey and tested this game more than a few times to make sure there was a regression. We saw a 7% drop in the average frame rate and a 9% drop in the 1% low.

Disabling SMT helps performance in World of Tanks. Here the 3900X managed to increase the average frame rate by 3%, and now the 9900K is only 5% faster, although disabling Hyper-Threading may also improve the 9900K's performance.

Fortnite is the type of game that we expect or suspect that disabling SMT might be helpful, and it appears to be. Honestly, the results are within limits or mistakes …

Hitman 2 is always an interesting title that stands out for one reason or another. Here we see a pretty big 10% improvement in the average frame rate and now the 3900X is just behind the 9900K, although the 1% low performance is only improved by 2%.

The performance in Resident Evil 2 remains practically unchanged.

The performance in Division 2 was almost identical to that without SMT support, at least when you look at the average frame rate. We see a 6% reduction with SMT disabled for the 1% low frame rate. While testing this game without SMT, we also noticed that the 1% low score tends to fluctuate quite a bit, and we haven't seen that with 24 active threads.

Last comes the Shadow of the Tomb Raider and here we see a nice improvement of 7% for the average frame rate, but as we've seen a few times now, the low performance of 1% is a little hit.

36 Compile game results

Deactivating SMT certainly seems a mixed thing, and we've tried switching titles for titles, but we've got results for even more games we've tested, starting with the average frame rate results.

On average, the Ryzen 9 3900X with activated SMT was only 1% slower overall. This is the default configuration when all 24 threads are enabled with simultaneous multithreading. Here we see that the performance margin for the vast majority of the games tested is within 5% and we would argue that we found a negligible performance difference in 31 of the 36 games tested.

Leaving SMT only allowed improved performance in Assassin's Creed Odyssey, while disabling did some benefits in Shadow of the Tomb Raider, Just Cause 4, Prey, and Hitman 2.

However, as we found when looking at individual games, the average frame rate cannot tell the whole story.

Let's see how 1% low numbers stack up …

If you compare the low power of 1%, the results swing back in favor of SMT.

If SMT remained activated, the performance in the 36 tested games could be increased on average by 1%. This means that we see a 5% better performance improvement in 8 games, a 5% or more drop in performance in only 4 games.

closing remarks

Just as we noticed two years ago when testing the Ryzen 7 1800X, disabling SMT can be an advantage, but your mileage varies and is strictly game-dependent (workload) and is also likely to result in a regression of performance.

If you don't disable SMT, the Ryzen 9 3900X can beat or even catch up with the Core i9-9900K in games. On the other hand, we don't really need anything to justify the purchase. The 3900X and all other 3rd generation Ryzen chips do what they're good at, and if you're looking for a solid all-rounder, the Ryzen 9 processor is just the ticket.

As we have already seen, the 3900X is sufficient for high-end games. If he is missing a few frames, this will be offset by significantly improved productivity performance and the ability to do more tasks at the same time. It also runs much cooler, uses a little less electricity and works well with the standard cooler included.

We do not recommend disabling SMT unless you know that this will improve performance in the particular game you will be playing. We're not sure why turning off SMT results in 1% slow performance in so many games. At first, we thought maybe it was because you are moving from a situation where the active threads required for the game can be used in a complex single-core chip with SMT enabled to a situation where those threads with disabled SMT are distributed on both CCDs. However, this doesn't seem to make sense since the 3900X generally offers better gaming performance than the 3700X, and we believe this is because it has two CCDs that have the L3 cache capacity and write traces to the IO Die increase. We contacted AMD to discuss this. So if we hear something interesting, we will report about it here with a footnote.

Purchasing links:
  • AMD Ryzen 9 3900X on Amazon, Google Express
  • Intel Core i9-9900K at Amazon, Google Express
  • AMD Ryzen 7 3700X on Amazon, Google Express
  • AMD Ryzen 5 3600X on Amazon, Google Express
  • AMD Ryzen 5 3600 on Amazon, Google Express
  • Intel Core i5-9600K at Amazon, Google Express
  • Intel Core i5-9400F at Amazon, Google Express
  • AMD Ryzen 5 2600X on Amazon, Google Express
  • GeForce RTX 2070 Super on Amazon, Google Express
  • GeForce RTX 2060 Super on Amazon, Google Express
  • GeForce GTX 1660 Ti on Amazon, Google Express

Leave a Reply

Your email address will not be published. Required fields are marked *