Phenom vs Ryzen: Hexa-cores a decade apart (Part 5)
Gaming Benchmarks
The arrival of Ryzen 5000 series finally dethroned Intel from the 'best gaming processor in the world' title. The IPC gains coupled with other improvements like reworked cache, improved boosting algorithm finally let AMD take the absolute performance crown. As can be comprehended from professional reviews, Ryzen 5 5600X demolishes every single processor Intel could offer in gaming at launch.
Games do love high clock speeds owing to it being lightly threaded, in most of the games even now(DirectX 12 titles are better at utilizing threads though). As a result, games really reap benefits from overclocking. AMD had multi-threaded performance down cold but where it lagged behind was gaming. With each launch, AMD chipped away at Intel's lead. Ryzen 3000 series was almost neck-on-neck with its Intel counterparts, until the Intel processors were overclocked that is. Contrary to that, 5600X outpaces the flagship i9 10900K overclocked to 5.1 GHz even at 1080p mind you, where the results are more CPU limited. That is AMD punching above its weight with 5600X. Compared to chips in its class, the delta averages at ~+25%.
This really is a testament to the constant innovations by AMD, gen after gen to establish its products as premium rather than being relegated to just the budget alternative, as it used to be. Intel's comeback attempt with 11th generation let it contest again for the gaming performance crown, by throwing power efficiency out the window of course.
With some history done and dusted, lets get into the gaming benchmarks. But before we begin though, we need to understand some metrics. Average Frames per Second(FPS) is pretty self-explanatory. It gives you a measure of the central tendency of how the system fared in the entire test loop, in terms of fps. What it doesn't tell is the distribution. Outliers can heavily affect gaming experience, which is hard to glean just from average FPS alone. Another metric, called the 1% low fps is taken for that. 1% low fps tells you about the distribution of values in the sample, meaning 99% of the frame rates is higher than this value. This is an indicative of the smoothness you experience in gaming - closer it is to the average fps, the better.
Contrary to popular beliefs, the graphics card is not always the only factor determining gaming performance. CPU plays a big role too, as we will see. If you have too fast a CPU with the GPU lacking, you will be GPU bottlenecked. Conversely, too good a GPU paired with a lackluster CPU will limit the GPU's performance and will be CPU bottlenecked. Identifying the bottleneck is a crucial step for upgrading your PC. In this set of benchmarks, we will see both sides of the coin - how the system went from being CPU bottlenecked to now the latter.
"Lets get into the gaming benchmarks". Good Joke. No, for real this time!
We start by taking a look at some synthetic tests.
Performance Test 2D graphics mark shows even with the same graphics card, we see a 2X (or more precisely, 111%) increase in performance when paired with Ryzen 5 5600X versus the 1055T. Other factors such as faster memory (DDR4 3600 MHz vs DDR3 1333 MHz) may also influence the results a bit but even so, we see a sizable lead when upgraded to 5600X.
The 3D graphics test shows relatively low performance increase as compared to 2D graphics test. We see 5600X system is only 7% faster versus the 1055T system with same GPU.
3DMark 11 Fire Strike benchmark shows 5600X pulling ahead by a significant 37%.
Only the Graphics score was taken here and we already looked at Physics score in our CPU benchmarks.
Hitman 3 sees a substantial increase in both average frame rate as well as 1% low frame rate with upgrade to 5600X. The test was performed in the Paris level by taking a fixed route from outside to the fashion show area with lots of NPC.
The upgraded 5600X system was an insane 60% faster in average fps while also being an impressive 68% faster in 1% low fps, which resulted in much smoother gameplay.
A 15-20% performance increase is often found in GPU a tier higher than the one being considered. Here, the CPU upgrade itself improves the performance of the GPU to a level which you would have got with a higher tier card, say Radeon R9 270X. Basically the game goes from unplayable below 30 fps, to much more playable state averaging around the 40s. 1% low fps is over 30 too, giving a much better gameplay experience.
Injustice 2 really isn't a good game to test graphics performance on since the game is capped to 60 fps. Still, the game being already installed on the PC, it was tested anyway. The in-game benchmark reported a 9 fps higher minimum fps for 5600X system with no difference in the average fps. Gameplay was visibly smoother though. This won't be included in the summary and was just for curiosity's sake.
Another frame-capped game, Need For Speed: Hot Pursuit 2010 (not remastered one) was also tested as it was already installed. We see no difference in average frame rate due to the cap but we do see a 18% improvement in 1% low fps. This also won't be included in the summary.
This is a surprising one. Usually open-world games tend to be more CPU demanding owing to a large game world, as we saw in GTA 5. Testing Witcher 3 showed no change in the average fps while 1% low fps increased by a mere 4 fps. The results were consistent enough across multiple runs to not chalk it up to some error.
Destiny 2, being an online only game is a difficult one to benchmark properly. The online nature of the game introduces variables beyond control and thus no two runs is likely to be same. This test was performed by exploring a lost sector in Trostland.
While average fps was close enough with both the processors to call it a tie, 1% low frame rate was 42% higher with 5600X.
Compared to 1055T, 5600X system was 23% faster in average frame rate and 61% in 1% low frame rate. Again, we see 1% low fps with 5600X almost matching the average frame rate with 1055T and averages around the 60 fps mark which results in a much better playing experience.
Coming to our first esport title, Valorant sees the highest gaming performance uplift in out entire gaming test suite. Average frame rate soared to 146 fps from 84. That's a massive 74% increase in performance. 1% low gain was even more impressive at 117%!
In competitive tactical shooter games such as Valorant, stuttering can affect a player's performance significantly. One might say high 1% low fps in games such as these is more important than high average frame rate. An average of over 200 fps will do nothing for you if the game experiences dips every now and then. That frequent stutter will put you at a significant disadvantage and leave you with a janky feeling in a highly competitive game like this.
Without even upgrading the GPU, my gameplay experience improved significantly courtesy of the 1% low gains.
CS:GO, another competitive tactical shooter doesn't see as much improvement as Valorant does but significant nonetheless. CS:GO running on source 1 engine is known to be more CPU demanding than GPU. The 5600X was 42% faster in average frame rate than the 1055T, pushing over 150 fps on average. 1% low fps gain was at 39% over 1055T. +40% improvement is nothing to sneeze at and this game too sees a substantial improvement in playability. Noticeably, the occasional stutters were almost non-existent in 5600X.
Note: CS:GO and Valorant tests were done in online deathmatch mode on same map each time. Yes, introducing multiplayer element adds a degree of randomness to the results and can't be consistently reproduced. But testing online performance is a closer indicative of actual gameplay experience since majority of the time spent in these games will be in multiplayer mode.
To support this point, we take a look at the next graph which involves scripted benchmark in CS:GO.
We take a look at a scripted benchmark test(found here) in CS:GO, which shows a 8% and 12% increase in average frame rate and 1% low frame rate respectively. This doesn't at all indicate the performance improvement found in actual online gameplay.
We saw the scales tip from being CPU bottlenecked with the 1055T to now completely GPU bottlenecked. This is probably the absolute best Radeon R7 260X can give and upgrading to an even faster processor, say a Ryzen 9 5950X (ambitious, I know) will give diminishing returns with regards to the GPU performance. Some improvements may yet be extracted by upgrading the memory.
The graphics card's full potential has only been unleashed 8 long years after being bought. That too an entry level one. Imagine the bottleneck with a faster GPU. How sad is that?
That is why it is imperative to carefully evaluate the type of workload your CPU will be crunching and whether the CPU will last the upgrade cycle you have planned(plan? what plan?). Of course, upgrade cycles are usually much shorter than my decade long one.
The following graphs gives an overview of the results we arrived to in this section.
First up, review of the synthetic benchmarks.
Nothing to be said here that I haven't already talked about. Clearly 2D graphics sees the most improvement. Where does this 2D graphics really come into play you say? No its not side scrolling games. Everything you do when you are not gaming(or multimedia workloads and such). Browsing though stackoverflow to fix that nasty bug, scrolling though the word document you had to make up for that pesky project of yours, checking out that trailer of the upcoming sequel to the movie nobody can appreciate but you. You know, normal stuffs.
The GPU switches to a lower state as these are too easy for the GPU to even call a "task". No, really. switching to lower clocks lets the GPU conserve power (consequently, lower temperature) and only heavier workloads such as games or rendering a scene does the GPU crank up to its maximum.
Synthetic benchmark results often do not correlate to similar results in games. So, we move on to review the gaming benchmarks.
Notice that the 1% low frame rate improved much more than the average frame rate and that is to be expected from a CPU upgrade. While GPU upgrade gives you raw performance to push those fps to new heights, a fast CPU smooths out the process. A fast enough CPU can issue commands to GPU consistently without much overhead while also have enough headroom for other tasks such as streaming or recording.
In the next section, we will some extra benchmark results with an even slower CPU added to the mix, the Phenom II X2 550, first processor I ever had.
Comments
Post a Comment