As always, an RTX 3090 Founders Edition from Nvidia is used for the gaming tests, with maximized temperature and power consumption limits, for the lowest possible GPU bottleneck. The performance data is recorded with Nvidia Frameview 1.2, based on the open-source software Presentmon. Like last time, there are again three relatively different game titles, Counter Strike: Global Offensive (CSGO) as an example of a more latency-sensitive title, Cyberpunk 2077 and Shadow of the Tomb Raider with a balanced preference and a slight preference for bandwidth, respectively. Cyberpunk has recently received a larger patch, but since the performance has decreased overall and also for better comparability, we are still testing with version 1.31.
Gaming 1440p
In 1440p or QHD, DDR5 is almost too fast to see any real differences in the FPS. This should also explain why we do not test in 4K or UHD at all. But at least a few extra frames can still be squeezed through the GPU’s bottleneck with the fastest RAM configurations.
The picture looks similar for the frame time variances, with the exception of CSGO, where the faster configurations actually still lead to measurably smoother scrolling.
Gaming 1080p
The differences between the configurations are still in the integer range in FullHD and 1080p, but the manual overclocking makes a maximum of 15 FPS on average compared to the JEDEC Config of DDR5. And even with the 1% low FPS, we are still in the range of the measurable, but probably not perceptible, even in 1080p.
The differences are most obvious in the frame time variances in 1080p, where the manual overclocks with their low latencies pay off. Sure, you could also test at even lower resolutions or settings, but would that still be representative of most gaming systems with a weaker graphics card and thus even tighter GPU bottleneck?
5 Antworten
Kommentar
Lade neue Kommentare
Urgestein
Urgestein
Neuling
Veteran
Mitglied
Alle Kommentare lesen unter igor´sLAB Community →