Hello, everyone, I'm Wing This time we are going to bring you a comparison review of the latest generation of DDR5 memory and the previous generation of DDR4 memory recently we have a long-awaited price reduction of graphics cards and many people are also reconsidering building a PC Because of the great improvement of the 12th Core it's becoming one of the best choices for PC building At this point, the problem will occur to you whether to choose DDR5 and which frequency to choose and this also concerns the selection of the motherboard DDR5 is available for almost half a year but because of the expensive price, the increase of graphics cards price, and few people actually building PC the majority of users are still unfamiliar with it so there comes lots of rumors for example, the performance of DDR5 barely equals to DDR4 3000MHz when its frequency gets to 6000MHz one DDR5 has dual-channel,etc. this video should be able to clarify some doubts for you This project is actually decided during the review of the 12th gen Core however, I have been busy all around for so long until now There are already a number of reviews online now I was considering if we just skip this video for good but after a second thought, I realized that the tests and emphases of every review would be different the more data, the better reference why not just show it to you guys since we already have the results, am I right? as for the new features of DDR5 you can find some conclusions in the 12th gen Core review and there are many videos online which can give you an explanation, more than you need It doesn't matter what to test and how to test I'll just save my words here Most of the new features don't seem to have that much of an impact on users during actual use for example, the matter of moving the power supply to the memory you won't feel any difference also, how is the stability of the on-die ECC there is no way to tell if the overclocking settings are not adjusted properly, the error will still occur after booting The most worthy of concern feature is still "dual-channel in a single stick" DDR5 is changed from one 64 bit channel in a single stick in the DDR4 period to two 32 bit sub-channels in single stick in some earlier versions of CPU-Z if you insert two DDR5 sticks, it directly shows 4 channels on the latest version, it has been changed to 4x32bit you got more tunnels, but each one is narrower at the same time The reason for concern here is, after all, the price DDR5 is still expensive at present If a single stick has the performance of a dual-channel, is it possible to install only one first and then get a full configuration after the price is lower? So we also tested the performance of a single DDR5 specifically After the background stories, let's take a look at our test configuration RAM first we choose two DDR5-6000 C40 16GB Trident Z5 RGB of G.Skill It is regarded as a relatively high-frequency memory stick in the early DDR5 family We will set different lower frequencies for this pair of high-frequency RAM to test the performance of DDR5 at different frequencies after all, underclocking is much easier than massive overclocking Because of the workload, we selected several typical frequencies for testing 4800MHz, 5600MHz, 6000MHz respectively actually there should be 5400MHz between 4800 and 6000MHz however, we didn't find that among the standard frequencies of JEDEC we only have 5200MHz and 5600MHz Among the DDR5 memory available now any ordinary stick can reach a frequency of 4800MHz if one can reach the frequency of 5200MHz, that's a slightly better stick the price of which is much higher the price difference between 4800MHz and 5200MHz ones is relatively close to that between 5200MHz and 5600MHz ones when it comes to the frequency of 6000MHz, that is a high-end one and the price goes much higher than 5600MHz so we choose these three options in the end all the timings are set to C40-40-40-76 the frequency division mode is GEAR2 for the moment the reason why we did not add the timings as a variant is that when several permutations and combinations are multiplied together, the workload directly scares us to death meanwhile, if you look closely, you will find that The DDR5 sold on the market at different frequencies are basically at around C40 timings the better ones are C36, and the worse ones are around C40 These settings should still be closer to our daily use Then the DDR4 memory used as a comparison is the Trident Z Royal 4400 C19, which is also from G.Skill we also choose to downclock it to test some common frequencies as well we use two settings here for DDR4 one is that the Core before the 12th and 10th gen can be easily achieved 4133 C19-26-26-46 the other, because of the memory controller of the 11th gen Core and Ryzen is the more suitable and classic frequency, 3600MHz C16-18-18-38 both in the mode of GEAR1 Some people may say that the comparison is not fair enough for DDR5, 4800MHz is quite an old version of frequency and the suitable match should be 2400MHz for DDR4 while the problem is, at the price of DDR4, 2400MHz you cannot afford the same capacity of DDR5 at 4800MHz the RAM of 3600MHz, C16 are all over the street while DDR5 at 7000MHz to 8000MHz is beyond ordinary users two pieces may cost more than the CPU and motherboard so in the end, we choose this consumer-oriented comparison way next, CPU Currently, if you need a processor which can support both DDR4 memory and DDR5 the 12th gen Core desktop processor is definitely the one here we use the highest-end i9-12900K when running the test The gaming performance is also at the top level, which won't become a bottleneck during the test after CPU, let's move on to the motherboard because among all the motherboards which can drive i9 when we ran this test at first, no single piece of motherboard could support both RAMs so we used two motherboards this time One is ROG's MAXIMUS Z690 HERO, which is DDR5 and the other is the same ASUS Z690-P, DDR4 as well although there are some differences between their product positioning they have a relatively good power delivery there is no problem driving the i9-12900K without overclocking after our real-world test at least in the comparison of memory performance the influence is not too much, and the difference can be seen The rest of the parts are basically the same as our review of the 12th gen Core no more detailed explanations here If you are interested, you can watch our previous video the operating system must be Windows 11 for sure So let's take a look at the real-world performance The first is the test of theoretical bandwidth we use AIDA64 as our testing software This table can be divided into two parts part one, the bandwidth The overall bandwidth of DDR5 memory is much higher than that of DDR4 indeed A single DDR5 4800 can already reach the read speed of 4000MB/s, and the write speed of 3500MB/s This bandwidth has exceeded that of the DDR4 2400 dual-channel already and is close to that of the DDR4 2666 C17 dual-channel However, if you look carefully, you will find that the ratio of bandwidth increase is actually similar to the ratio of frequency increase no matter DDR4 or DDR5 I think the DDR5 has such a high bandwidth is mainly due to the frequency rather than the so-called "dual-channel in a single stick" At the same time, you can also find that after two RAMs form a true dual-channel the bandwidth is not completely doubled there is still a slight loss part two of this results, latency Because of high timings and Gear2 Even with DDR5 of 6000MHz the latency is still higher than DDR4 of 3600MHz in the settings of two DDR4s because the timing of that 4133MHz one is higher so when the frequency is higher, the latency does not drop Even the latency of 3600MHz C16 is a bit lower Then DDR5 single-channel latency is actually lower than that of dual-channel Next, let’s go directly to the gaming part, which should be concerned about by most people The first one, CS:GO, 1080P, minimum settings in a word, a complete victory for the DDR4 group D5 6000 C40 also failed to beat D4 3600 C16 but even with single-channel DDR5 of 4800MHz, this game can reach more than 800fps which is just a few dozen frames away from DDR4 of 3600MHz you won't feel much difference during everyday gaming we also tested PUBG, which is CPU single-core performance and RAM performance demanding 1080P resolution and the mostly used triple extreme settings in a word, a complete loss for the DDR5 group D4 3600 can reach 5~10 frames higher than D5 6000 on average In the two DDR4 settings, because the 4133 timings are also pulled a little higher the FPS is very similar to the 3600 when the RAM bandwidth reaches a certain number it depends more on latency while latency is determined by frequency and timings both and the FPS of the single-channel DDR5 of 4800Mhz is obviously affected while the single-channel DDR5 of high frequency is less affected compared with the dual-channel, the FPS is not much reduced the third game, Total War: Three Kingdoms 1080P, low graphics quality, again The FPS gap between DDR5 6000 dual-channel and DDR 4 4133 is about one frame that's barely a tie The DDR5 single-channel will also be affected during this game DDR5 of 4800MHz drops about 6.3% of FPS compared to that of high-frequency dual-channel one The other two high-frequency single-channel settings drop a little less The DDR5 6000 single-channel performance on FPS is even slightly better than that of the DDR5 4800 dual-channel our last game is the relatively new Far Cry 6 1080P, low graphics quality, the third time DDR4 dual-channel high frequency has one or two more frames than DDR5 dual-channel high frequency barely a tie again But the DDR5 single-channel is more affected during this game DDR5 4800 has over 15% less FPS than D4 dual-channel high frequency Because of the workload, only these games were tested this time kind of a group of common types If you want to see more game tests please go to the gaming performance review of DDR4 vs DDR5 by 51972 very detailed review by the way, the video he spent so much time making only has 200,000 views if you don't give him enough support, it will really discourage his enthusiasm The overall conclusions he got were similar to mine except for several projects the DDR5 of more than 6000Mhz cannot beat DDR4 of high frequency in gaming We went on to test some common professional software after all, people who buy desktops don't always use them to play games First, not our old friend Cinebench for sure it was known for sure long ago This type of CPU hungry software are basically RAM-independent PhotoShop test of PugetBench comes first dual-channel DDR5 defeated DDR4 a little bit overall PhotoShop seems to have a certain demand for RAM bandwidth but it's not a big difference Correspondingly, the DDR5 single-channel has been greatly affected Even a single-channel RAM with a frequency of 6000MHz still has over 10% less benchmarking scores than the dual-channel DDR4 PugetBench's Premiere test comes second in this test, as long as it is dual-channel it is stronger than a single channel, but there seems to be little difference between different frequencies and this software is quite random you can find that some low-frequency results are better than high-frequency ones the difference of several tenths of scores can be calculated within the margin of error in the end, the four familiar tests of SPECWorkstation which has three render tests plus one compression test In general, the high-freq DDR4 is equal to DDR5 6000MHz Only DDR4 is slightly better in LuxRender there are some differences between single-channel and dual-channel DDR5, but it is relatively tiny the influence of the frequency is rather bigger the last one, 7Zip, is greatly affected by SSD so the overall difference is not too big OK, we are coming to the end, let's draw a conclusion although the results are shown one by one quickly, seems not so many there are a great number of permutations and combinations to be tested it's really time-consuming and we basically have to run two or three times for each test we take the highest or average score to rule out errors or something don't forget to give a free like and support for us if you stay so long here after all the tests If you bought your PC for gaming primarily at this stage, it's better to choose DDR4 with high frequency and low timing theoretically at least D5 6500 has the potential to catch up with D4 3600 C16 for professional use especially those students who will use some software that clearly demands RAM bandwidth while if you have enough budget, you can also choose DDR5 of course, if you see a branded machine with a particularly attractive price or a motherboard that attracts you so much but it can only support DDR5, then there is no choice After all, it depends on demand and price in fact, the gaming performance of DDR5 can be predicted just by looking at the bandwidth and latency tests after all, HEDT and these four-channel, six-channel, and eight-channel server platforms have similar situations of high bandwidth with high latency if higher RAM bandwidth comes with better gaming performance then everyone already has bought a server to play games the problem of "dual-channel in a single stick" I don't agree with this DDR5 itself has a much higher frequency than DDR4 so a single stick can also have a relatively high bandwidth In fact, I did a similar test on the DDR4 the influence of single-channel DDR4 on gaming is similar to the data we have on DDR5 so if you want to have a better gaming and professional software performance with DDR5-6000 you still need two of them but as mentioned earlier if you have a machine or motherboard which can drive DDR5 very well it can still be okay to use just one stick at first The influence of many projects is actually not that big what's more, graphics cards are mostly demanded by games needless to say, I still recommend you to use higher frequency as much as possible In addition, I recently heard that the overclocking ability of Hynix DDR5 die is particularly impressive bare bars of 4800MHz can easily overclock to over 6000MHz I have already bought the RAM of KLEVV, which comes from the same group as Hynix later we will have some tests if it's true then at this stage, DDR5 can become much more cost-effective Finally, of course, I hope that the price of DDR5 can come down sooner and the frequency can go up sooner as well hope everyone be able to embrace the latest generation of technology as soon as possible Okay, so this is the end of the video If you like it, hit the like button and comment below If you are new to the channel, hit the subscribe button so that you won't miss any content afterward I'm Wing, and I'll see you next time. Byebye~
