AMD Ryzen 9 3950X Overview: The New Efficiency King

Today we can finally show you how the Ryzen 9 3950X works. At Computex earlier this year, we weren't sure whether there would be a 16-core AM4 CPU, but AMD put it to rest when it later announced the 3950X at E3. Originally on the shelves in September, AMD pushed the release a few months ahead and is expected to hit stores on November 25th. This means that you can see all performance data and data well before the start.

That's good news, considering that the 3950X isn't cheap. At $ 750, it's still good compared to Intel's competition. For example, the 8-core 9900K currently costs $ 500, so pays 50% more for 100% more cores, which may be a pretty good deal.

The Ryzen 9 3950X tries to bridge the gap between mainstream and high-end desktop platforms, at least existing high-end desktop platforms. Third generation Threadripper will change this later this month. The point, however, is that we've never had such an extreme CPU on a mainstream platform before, and this is certainly the most expensive mainstream platform CPU we've seen in a long time.

The Core i7-2600K cost $ 320, the 4790K $ 340, the 8700K $ 360, and now the Core i9-9900K around $ 500. The Ryzen 9 3900X was a big deal for AMD, as it matched Intel for $ 500 and was rightly not like the 1800X.

… the 3950X enables AMD to achieve a premium for desktop computing. But is your new 16-core / 32-thread monster worth the asking price?

But the 3950X goes beyond that, which enables AMD to get a premium on desktop computing. But is your new 16-core / 32-thread monster worth the asking price? Given what we've seen from the 3900X, you'd think it would be if you could get all of those cores up and running, but of course we'll do some testing and get to the bottom of it.

In terms of specifications, the 3950X is similar to the 3900X: you get 16 cores versus 12, and the boost clock speed has been increased by 100 MHz, presumably because the 3950X gets better silicon. However, to stay within the 105 W TDP, the base clock was lowered from 3.8 GHz to 3.5 GHz, although we assume that the all-core clock frequency is similar.

Another difference between these two CPUs is the packaging: the 3900X comes with the Wraith Prism RGB cooler, while the 3950X does not come with a cooler at all. So you have to consider the additional cost of a quality cooler.

For testing we use the new Corsair HydroX cooling device for the Gigabyte X570 Aorus Xtreme. For the block we used the XC7, the XR5 360mm cooler and the XD5 pump / reservoir combination, all based on WetBench. At some point we will upgrade all of this to a real test case, but due to the limited time we have it for now. We also used this setup to retest the 3900X, 3800X and 3700X and update the results if necessary. We also use the DDR4-3200 CL14 memory for all current AMD and Intel processors.

The 8th and 9th generation Intel Core processors were manufactured on the Gigabyte Z390 Aorus Ultra and cooled with the 280 mm all-in-one liquid cooler Corsair Hydro H115i RGB Platinum. Note that the Intel CPUs are not limited to TDP. Therefore, we show the best scenario for an out-of-the-box performance. The graphics card of choice was the MSI Trio GeForce RTX 2080 Ti. Now we come to the benchmark results …

Benchmarks

You can see why Intel tried to downplay the relevance of Cinebench to CPU performance. The 3950X is a whopping 89% faster than the 9900K. For those of you who use Maxons Cinema 4D, the 3950X is worth the premium over the 9900K. That is already the Cinebench R20 score.

The 3950X was also 27% faster than the 3900X, and although this worsens its value, the premium is easy to justify when time is money. While I suspect 2950X owners are interested in 3rd generation thread rippers, those who are thinking of buying the 2nd generation 16-core processor will want to buy it for $ 500 or less because of that 3950X is almost 30% faster. Of course, the 2950X offers more PCI Express lanes, but if you can get by with an X570 card, the 3950X is the better option.

With this maximum single-core boost clock of 4.7 GHz, the 3950X was immediately able to achieve 529 points, which corresponds to an increase of 2% compared to the 3900X. This makes sense, since it should be clocked around 2% higher in this test. Perhaps more impressive is the fact that the 9900KS has actually been displaced, making it the fastest desktop CPU in this single-core test.

Here is a brief overview of the persistent memory bandwidth performance. As expected, the 3950X is very similar to all other 3rd generation Ryzen CPUs we've tested so far.

WinRAR performance is dramatically improved with the 3rd generation Ryzen, but the 12- and 16-core models are still lagging behind the 8-core 9900K from Intel and lagging far behind the 12-core 9920X .

The 3950X didn't seem to be any better than the 3900X either, which is interesting, although we saw something similar with the Threadripper 2920X and 2950X, only to a lesser extent.

The 3950X is only 2% faster than the 3900X in the 7-Zip compression test. That means it was 8% faster than the 2950X, but still 5% slower than the Core i9-9920X.

It is a decompression work that Ryzen has always done excellently. Here the 3950X blows everything out of the water and beats the 9920X by an incredible 73%. It was also 25% faster than the 2950X and just over 30% faster than the 3900X, which is a great result for the new AM4 16-core processor.

Unfortunately, we see little or no performance gain in the 3900X when it comes to coding performance in Adobe Premiere Pro CC 2020 Edition. Compared to the 3900X, the 3950X was only 3% faster, which is surprising as it ran at similar clock speeds in this test and the CPU utilization for both parts was over 90%.

However, if we compare each processor side by side with the open Windows Task Manager, you can clearly see that four of the 3950X cores are not fully utilized. For this reason, the 3950X cannot move further away from the 3900X.

The 3950X also performs well in the V-Ray benchmark and offers almost 30% more performance than the Core i9-9920X and over 80% more than the 9900K. This is another really solid result for the red team.

Next up is Corona, where the 3950X was 25% faster than the 2950X, which is a big step forward. It was almost 30% faster than the Intel Core i9-9920X and the 12-Core 3900X. It also blew the Intel Mainstream 9900K out of the water as it was 71% faster.

The last application benchmark we ran was Blender Open Data. Here the 3950X offered a 29% improvement over the 3900X and 37% more power than the 2950X. The 9920X was decimated by 56% and the 9900K by 87%.

It's also crazy to believe that the Core i7-7700K was the mainstream desktop's flagship about three years ago. The 3950X is about 300% faster and offers 130% more money.

power consumption

The Ryzen 9 3950X has something unusual and special about how efficient it is. Although the 3950X in the Blender benchmark offers 29% more power than the 3900X, it actually uses a few watts less, which is absolutely crazy. This also means that you expect identical power consumption for an 87% increase in performance compared to the 9900K.

The Ryzen 9 3950X has something unusual and special about how efficient it is.

We checked these results three times to ensure that the 3950X and 3900X were tested with the other Ryzen processors on the same card with the same BIOS and settings and cooling. We also compared the results with Gamers Nexus employees and they found very similar power consumption values. We are therefore convinced of the accuracy of these numbers.

Why does the 3950X bring total system consumption to the same level as the 3900X? In a word: binning. It is simply better silicon and can therefore be operated at much lower voltages. While the 3900X normally ran between 1.275 and 1.3 V, the 3950X only ran at 1.150 V, which means that the 3900X required an average of 13% more voltage. Combine this with a clock rate increase of 3%. For this reason, the 3950X and 3900X models use a similar amount of power.

Gaming benchmarks

Time for a couple of games. Battlefield V comes first with the DX12 API. We previously tested with DX11, so these are all updated results for this review. As you can see below, the 3950X is a whisker faster than the 3900X, and that only made it a fraction slower than the 8700K and 9700K, while being 12% slower than the 9900K when comparing 1% low power.

It was almost 30% faster than the 9920X when comparing 1% lows. For some reason, Intel's current high-end desktop CPUs often experience performance issues while gaming.

If you increase the resolution with the RTX 2080 Ti to a more realistic scenario at 1440p, the 3950X is now only 7% slower than the 9900K when comparing 1% low data, and again essentially corresponds to the 9700K and 8700K. Again, the performance was much more constant compared to the Core i9-9920X.

The performance of Shadow of the Tomb Raider is a bit poor, here the 9900K was 16% faster at the average frame rate, while the 9920X was 11% faster, the 3950X even a single frame faster than the 3900X.

If we jump to 1440p, we will find the limits of the RTX 2080 Ti if we use the highest quality preset. This was achieved not only by the 3950X, but also by the 3900X and 3800X as well as the Intel Core i7-9700K and 9900KS. Oddly enough, the 9900K, 8700K and 9920X fall slightly behind at 1440p, although they are much faster at 1080p, and we've seen something like this before.

The Ryzen 9 3950X performs very well in The Division 2 with 1080p and essentially corresponds to the average frame rate from Intel. At the same time, it offers a slightly higher result of 1%, although this agrees very well with that of the 3900X.

Oddly enough, increasing the resolution messes up Ryzen and the Intel Core i9-9920X. The 3950X is now up to 10% slower than the 9700K, 9900K and 8700K. This is a very unexpected result considering what was seen at 1080p.

The 3950X was competitive in the Ghost Recon Breakpoint, pushing the 2080 Ti to what it had achieved with the 9900KS within a few frames while maintaining the same 1% low result.

The average frame rate performance of the Ryzen processors dropped slightly at 1440p, although they achieved slightly stronger 1% lower results. In our benchmark, the 3950X and the other Ryzen CPU test delivered very narrow margins between the low and the average frame rate of 1%. This resulted in smooth performance, although the high-end Intel processors were also very smooth.

The F1 2019 results have also been updated with the DX12 API. Here we see some positive results for the 3950X, since it displaced both the Core i7-9700K and the Core i9-9900K and only slightly lost against the 9900KS.

Oddly enough, when the resolution is increased, both the Ryzen processors and the Core i9-9920X fall off, and this is very similar to what we saw while testing The Division 2. It's not a massive drop, but given the 1080p results you'd expect, the 3950X, at 1440p, is almost 5% behind almost all Intel CPUs.

The last game we're going to watch is Borderlands 3, and again you can see that the 3950X performs well at 1080p. Here it fits the 9900KS, although it wasn't faster than the 3900X.

Even at 1440p it only suited the 3900X and although it was 2-3 fps less on the Intel CPUs, it was 1-2 fps faster for the 1% low result. While this may not be the most CPU-demanding game, the 3950X does well.

Overclocking

Here is a brief overview of the overclocking performance: With 1.325 V, the 3950X was 100% stable with an all-core value of 4.3 GHz. We might have been able to get it to 4.4 GHz, but we stopped at 4.3 GHz with limited time to check overclocking. We still exceeded the 10 km mark with a Cinebench value of 10046 points, which corresponds to a performance increase of 9%. We also found that simply activating PBO increased performance by 5%.

Blender's growth was milder, with PBO boosting performance by 3%, while manual overclocking improved things by 7%.

If you're looking for maximum performance at the expense of everything else, overclocking the 3950X makes sense, but if you're interested in efficiency at all, it really isn't.

For the small 3 to 5% increase in performance that PBO offers, expect an increase in overall system consumption of over 25% and an increase in manual overclocking by 4.3 GHz by almost 40%.

Because of the limited time, we don't have a whole bunch of thermal data for the 3950X, at least not with a number of different cooling solutions. At the moment we only have the data with the custom Corsair HydroX loop and for comparison we have thermal data with the 3900X.

The 3950X reached a maximum of only 64 ° C after a one-hour Blender stress test in a 21c room. The CPU ran at 1150 V and maintained a clock rate of 4025 MHz until the end of the stress test.

For comparison: the 3900X reached a maximum of 68 ° C under the same conditions, which means that the 3950X actually ran a bit cooler. The 3900X maintained a clock speed of 4100 MHz, although, as we saw when measuring power consumption, the voltage was slightly higher, which resulted in the extra warmth.

In other words, any cooler that can handle the 3900X has no problem with the 3950X.

The new performance king, but who is he for?

In this test, we were excited to finally test the Ryzen 9 3950X, but we weren't expecting any major surprises. It's basically a 3900X with four additional cores. If all things are the same, it will be up to 33% faster, and we have certainly seen that in applications like 7-Zip and Blender.

However, when measuring power consumption, we found how impressive the 3950X is. We're still excited about the fact that it offers almost 30% more power and still uses a few watts less than the 3900X. The 3950X really shows what is possible with the 7 nm process from TSMC.

This improvement in silicon quality makes the 12% price premium you pay per core for the 3950X over the 3900X worth. By the way, if you adjust the clock speeds and voltage of the 3900X and 3950X, the 16-core model consumes about 30-40% more power than expected because it contains 33% more cores.

In the past month we have been asked many times whether you should wait for the 3950X or just buy a discounted Threadripper 2950X. Our advice was to wait for 3950X reviews and we are happy to have suggested this as the 3950X is good. The TR 2950X currently costs $ 680, but is 30% slower and increases overall system consumption by 30%. In terms of performance and efficiency, the 2950X is the poorer choice except for these additional PCIe lanes.

The Ryzen 9 3950X also offers the advantage of cheaper motherboards. You can throw this thing on a decent B450 board for just over $ 100, and that's exactly what we'll be doing in a new test soon. A decent X399 board now costs around $ 300. Ultimately, we believe the 3950X and 3900X have discontinued the 1st and 2nd generation Threadripper series for most people.

At Intel, the R9 3950X is clearly a much better CPU than the $ 1,050 Core i9-9920X. However, this part will be updated very soon as the Core i9-10920X for only $ 700. While the Intel CPU with 12 cores consumes similar performance, it is often far superior in terms of raw performance. There are cases where the 10920X wins, but with a slight lead, while there are many more cases where the 3950X completely burrows it. For similar workstation performance, you'll need to spend $ 800 on the 10940X or $ 1,000 on the 10980XE to beat AMD's best AM4 CPU.

AMD has delivered a "mainstream" AM4 part that can compete with and outperform the best high-end desktop CPUs from Intel. This explains why 3rd generation Threadripper is so damn expensive as AMD tries to separate its mainstream and high-end desktop platforms. But with the $ 1,400 24-core 3960X, which offers more PCIe lanes than most of you know, and the double L3 cache of the 3950X, 3rd generation Threadripper becomes a real high-end desktop Product.

We'll have an opportunity later this month to evaluate 3rd generation Threadripper along with the Intel Cascade Lake-X series. But at the moment we're seriously impressed with what the 3950X offers for $ 750. It is an exceptionally good option for anyone looking for a powerful 16-core processor.

Purchasing links:

Leave a Reply

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