75% performance you say? Bullshit. You double the frequency and double the latency and you achieve nothing. Ddr5 with high frequency and high latency is marketing bullshit.

bluedevil:

75% performance you say? Bullshit. You double the frequency and double the latency and you achieve nothing. Ddr5 with high frequency and high latency is marketing bullshit.

you get more memory bandwidth, if your workload is memory bound it does increase your performance linearly. latency hasn't really improved since the ddr2 days, you could buy ddr2 800 cl3 kits which offer comparable latency to the lowest latency kits today. similar to a 3600 cl14 or 4000 cl15 kit

bluedevil:

75% performance you say? Bullshit. You double the frequency and double the latency and you achieve nothing. Ddr5 with high frequency and high latency is marketing bullshit.

If this were still DDR4 then you'd be right. But DDR5 works a bit differently in the way the CPU accesses it, where to my recollection, the CPU is getting more data per clock cycle. In either case, I'd still like to see latencies that are less than doubled.

user1:

you get more memory bandwidth, if your workload is memory bound it does increase your performance linearly. latency hasn't really improved since the ddr2 days, you could buy ddr2 800 cl3 kits which offer comparable latency to the lowest latency kits today. similar to a 3600 cl14 or 4000 cl15 kit

I was referring to games but we can test your theory that bandwidth matters more then bandwidth over latency ratio. Just run a benchmark for your app then go into bios and up the 4 primary ram timings as far as they can go and run the benchmark again. It's the ratio between latency and bandwidth that really matters.

Here are the Jedec standards for DDR 5. Please note that there are three levels of cas for each speed. They are Fast Standard: Common Standard: Loose Standard: . You will notice that the bandwidth starts to expand as the speeds go up, but the latency stays almost constant no matter what speed you choose. So in the end it is faster even at higher cl numbers. https://media1.giphy.com/media/ap6wcjRyi8HoA/200.gif?cid=ecf05e47pwf15j9uuy3m71pbfo7gaye4keicoslujwxzywv3&rid=200.gif&ct=g

Omg. I can't seem to make this clear enough. Ex ddr at speed 2400mhz cas 10 offers exactly the same damn performance as 4800 MHz at cas 20. The damn bandwidth does not matter if you double the latency also. Ddr5 3200 cas 22 gives you more performance then ddr5 6400 cas 46 because ddr5 6400 has higher cas at 46. It would have offered the same performance as ddr 5 3200 if it had cas 44.

bluedevil:

I was referring to games but we can test your theory that bandwidth matters more then bandwidth over latency ratio. Just run a benchmark for your app then go into bios and up the 4 primary ram timings as far as they can go and run the benchmark again. It's the ratio between latency and bandwidth that really matters.

depends on the game, though most games these days don't push that much data, so probably won't help much, but on the other hand, if you do use something like photoshop, or other applications that involve media manipulation and trans-coding ,It should help a fair bit

bluedevil:

Omg. I can't seem to make this clear enough. Ex ddr at speed 2400mhz cas 10 offers exactly the same damn performance as 4800 MHz at cas 20. The damn bandwidth does not matter if you double the latency also. Ddr5 3200 cas 22 gives you more performance then ddr5 6400 cas 46 because ddr5 6400 has higher cas at 46. It would have offered the same performance as ddr 5 3200 if it had cas 44.

According to your theory then all anybody needs is 2400 mhz at 10 and that's it lol. Bandwidth does matter if you use your computer for real work and not just the gaming argument with a high end gpu. Gaming using integrated graphics, file compression, photo editing, video encoding and transcoding all respond to higher bandwidth than cas. Everyone is going crazy over the DDR5 without even seeing what the actual numbers are using Alder Lake and a 690 motherboard. Then when those numbers come out, let the fun begin!

NewTRUMP Order:

According to your theory then all anybody needs is 2400 mhz at 10 and that's it lol. Bandwidth does matter if you use your computer for real work and not just the gaming argument with a high end gpu. Gaming using integrated graphics, file compression, photo editing, video encoding and transcoding all respond to higher bandwidth than cas. Everyone is going crazy over the DDR5 without even seeing what the actual numbers are using Alder Lake and a 690 motherboard. Then when those numbers come out, let the fun begin!

It's the bandwidth over latency ratio that matters and the real frequency of the ram needs to run on sync with the cpu to ram link else you lose performance. This is why for 11th gen you need 3600 ram. And if you read my posts in this thread I did clarify that I was talking about gaming. Integrated graphics for gaming? Stop kidding yourself. Also it's not my theory it's how ram performance works. Bandwidth over latency ratio. I don't know about your so called real work but I don't see any reason why even in that scenario bandwidth would matter more then the ratio. Care to provide some benchmark numbers for your statement? Just increase the primary ram timings by a lot and keep the frequency the same then run the bench and see if anything changes.