IOPS, that's only thing which matters for NVMe today for practical use in 95~99% of cases.
In those cases, PCIe 3.0 storage with higher IOPS eats PCIe 4.0 storage that may have double bandwidth.
IOPS, that's only thing which matters for NVMe today for practical use in 95~99%
Hmmm .... that's also questionable, can you tell me in what situation you would require 200K+ IOPS even RND 4K T64 on your game or work PC?
The biggest bottleneck (if you can even call it a bottleneck) still is 4K RND 1T, but that's mainly an OS restriction ... and not something that can be solved on an SSD end.
not saying intel's isn't gonna be faster,they gotta have at least some grounds for a statement like that. but I think those numbers are
a) very non-specific
b) very inconclusive
nvme 3.0 used to run faster on intel than x370 too,but the real world difference and synthetics are different things
in anvil it was as much as 19% in intel's favor,pcmark 21%
https://www.tweaktown.com/articles/8073/amd-ryzen-ssd-storage-performance-preview/index.html#Benchmarks-PCIe-Storage-Samsung-960-EVO-1TB-Intel-vs.-AMD
so intel probably isn't lying.but you gotta take these results and confront them with real world scenarios,where like Fox said, pci-e 4.0 isn't really faster than 3.0 for nvme drives
Hilbert Hagedoorn:
The biggest bottleneck (if you can even call it a bottleneck) still is 4K RND 1T, but that's mainly an OS restriction ... and not something that can be solved on an SSD end.
You can't compare two completely different technologies like that. Optane was good and showed a lot of potential, however, the technology is just far too expensive to become commercially available to the masses. Ergo the interim cache solution.
Hmmm .... that's also questionable, can you tell me in what situation you would require 200K+ IOPS even RND 4K T64 on your game or work PC?
The biggest bottleneck (if you can even call it a bottleneck) still is 4K RND 1T, but that's mainly an OS restriction ... and not something that can be solved on an SSD end.
I'll tell you where bandwidth clearly wins instead:
When I create custom windows image. It requires sequential load and when changes are done, there is sequential dump.
When I load/boot virtual images of some OS.
Where My NVMe's end up having high "Active time" while allowing only fraction of throughput?
Steam's patching of data files.
Those few games that do not use stupidly high compression that makes CPU decompression bottleneck.
Database depending applications.
Citrix VDI.
For usual home user, few have issue with any of current parameters, because software is limit. (Especially in games data loading.)
But as games catch up with theoretically available bandwidth by having compression that's easier to decompress or scales with more CPU cores, storages will start to hit 100% "Active Time" and each will do it on different bandwidth for same type of data being read/processed.
Low IOPS storages will hit 100% while providing lower volume of data per second.
When I bought Patriot Burst 1TB SSD (not NVMe), that thing did choke on game loading. Showing 100% "Active time" while reading ~150MB/s.
For It was barely good enough as storage for data I do not use. And I got rid of it shortly after I got it.
(And while picking bad example as hyperbole, it shows that storage may hit unexpected wall of low performance.)
And that's why I have no interest in PCIe 4.0 storage at premium prices. (And fact that I have no PCIe 4.0 is connected. I simply have no need for it with current devices and general software state.)
- - - -
Bit off-topic, but:
"Did anyone seen comparison of PS5 vs XSX loading times, resume, game switch, ... ?"
It's interesting how many tech site editors have gone to work for the big companies, Kyle Bennett, Scott Wasson and Anand Shrimpi.
Talking of Anandtech, why have they stopped doing video card reviews? Weird, at least it gives more room for HH and this great site
I hope you re not planning on running off to Intel, AMD or Matrox Hilbert ! 🙂
This is Ryan Shrout, here. This man cannot be trusted with regards to anything. I wouldn't even trust that man to buy me a stick of gum.
Always assume anything he says is a lie, because safer to operate from that position.
We can come to our own conclusions when independent reviewers get access to the hardware.
LMAO, USB disconnecting/WHEA Errors/ Heat issues and now low performance PCI SSD. AMD Sells faulty/incomplete/untested products and still people are buying overpriced their products.
The chart could mean that the Intel CPU wakes up faster from deep sleep (C-)states or that Speed Shift is faster than the equivalent of AMD CPUs (Smart Shift?), aka "autonomous mode".
In the end I would test with all C-states disabled first, then enable C-states and then enable autonomous mode.
C-states always had an impact on SSD performance and even on noise coming out of your computer while SSD during SSD access (with laptops also depending on whether they run on battery or plugged in). Intel's RST driver included a performance feature that (optionally) disabled deep sleep states for better performance. In the past I used that to dynamically disable C7 from within Windows while keeping C3 enabled (or disable both), without having to go into the BIOS.
You can't compare two completely different technologies like that. Optane was good and showed a lot of potential, however, the technology is just far too expensive to become commercially available to the masses. Ergo the interim cache solution.
im showing you pcie 4 ssds doing 60mb/s as well as some doing above 90mb/s.
LMAO, USB disconnecting/WHEA Errors/ Heat issues and now low performance PCI SSD. AMD Sells faulty/incomplete/untested products and still people are buying overpriced their products.
While Intel's complete hardware is beautiful and doesn't have more exploits than the original release of World of Warcraft.
There is certainly an issue with USB, but it's not as widespread as forums would have you believe. Microsoft would be the first ones to know due to the telemetery from Windows 10. On the other hand, speaking about a platform as a whole, Intel is barely catching up a year later, and it will still be behind in core counts and thermals.
This "benchmark" is also meaningless, unless independent reviewers check it. We have no numbers, and nothing specific, even if you check their "performance index" website, they say nothing about actual numbers or settings.
[spoiler] [/spoiler]
Also who in their right mind would get a 8/16 CPU over a 12/24 one, for a slight storage performance increase, even if true.
My 3.5 years old x299 does ~ 300MB/s 4k random read @ QD=1 with Optane 900p.
Optane 900p on my x570 and 5900x does about 230MB/s 4k random read @ QD=1. Using Optane 900p for OS and programs for both plattforms.
My threadripper 1950x server does 200MB/s on Optane 900p. Still way better than normal ssd's.
Nothing, it's completely in their bragging rights. Everyone wins if performance really improves, everyone loses if figures are massaged to look better than what you get for real.
LMAO, USB disconnecting/WHEA Errors/ Heat issues and now low performance PCI SSD. AMD Sells faulty/incomplete/untested products and still people are buying overpriced their products.
I have 2 AMD platform : X570+5900x and B550+3800xt : NO USB disconnecting/NO WHEA Errors/ NO Heat issues and NO low performance PCI SSD
I have 1 Intel platform : Z370+9900k stock : Heat issue on heavy Load task and sometimes WHEA Erros in obs windows.....
I have 2 AMD platform : X570+5900x and B550+3800xt : NO USB disconnecting/NO WHEA Errors/ NO Heat issues and NO low performance PCI SSD
I have 1 Intel platform : Z370+9900k stock : Heat issue on heavy Load task and sometimes WHEA Erros in obs windows.....
LMAO, USB disconnecting/WHEA Errors/ Heat issues and now low performance PCI SSD. AMD Sells faulty/incomplete/untested products and still people are buying overpriced their products.
Cool story bro. I have been on intel for good 10+ years. First time amd since Athlon xp.
I found zero issues, perfect performance out of the box. Usb problem is not big as you might think. It is blown out of proportion.
My case : in usb 2.0 port logitech g314 wasn't properly recognized by logitech software - but keyboard works.
In usb 3.0 works just fine.
Also fixed with new bios
My 3.5 years old x299 does ~ 300MB/s 4k random read @ QD=1 with Optane 900p.
Optane 900p on my x570 and 5900x does about 230MB/s 4k random read @ QD=1. Using Optane 900p for OS and programs for both plattforms.
My threadripper 1950x server does 200MB/s on Optane 900p. Still way better than normal ssd's.
I use Optane in all of my systems now. In particular my workstation (4X 905P) and mobile workstation (M.2 22110 905P) use Optane and everything from boot times, to install times to app load times to VM launch times is super snappy.
Hilbert Hagedoorn:
Hmmm .... that's also questionable, can you tell me in what situation you would require 200K+ IOPS even RND 4K T64 on your game or work PC?
The biggest bottleneck (if you can even call it a bottleneck) still is 4K RND 1T, but that's mainly an OS restriction ... and not something that can be solved on an SSD end.
If you ran a 905P for a couple of months you might feel differently.
I'm waiting on these stupid PCIe4 Intel chips to come out so they will finally release the P5800X. Without question these are going into my next workstation, which will be AMD based for the first time in my life.
[/spoiler] Also who in their right mind would get a 8/16 CPU over a 12/24 one, for a slight storage performance increase, even if true.
:D