PEnns
Reputable
- Apr 25, 2020
- 702
- 747
- 5,770
- Mar 16, 2013
- 176,623
- 20,420
- 184,590
Building a PC today is vastly easier than in days of old.
Having to manage the specific order in which drivers install, just to get it to boot up with the available RAM...that was not fun.
himem.sys, config.sys, etc, etc....
And of course this ignores the multitudes who simply buy whatever from BestBuy/Amazon/Newegg.
But many of us would have pivoted to something else.
Photography, cars, UAVs....
- Mar 16, 2013
- 176,623
- 20,420
- 184,590
And most of the "problems" we see here are people who ignore the best practices, and try to strike out on their own.
newtechldtech
Respectable
- Sep 21, 2022
- 476
- 199
- 1,860
actually it was easier in older days , because you knew exactly what was happening while booting up .. and step by step you would know which hardware is making problems for you ... each hardware had its drivers diskettes , and no guessing , no screen of death ... the only annouing thing was setting up the right IRQs correctly without conflicts .
- Mar 16, 2013
- 176,623
- 20,420
- 184,590
You knew, and I knew...
But people that have a hard time today, with extensive information via the innertubes and forums like this...would have been completely lost
Darkbreeze
Retired Mod
- Jun 24, 2014
- 81,495
- 7,926
- 176,790
Besides which, what you see on these forums is maybe one out of every fifteen or twenty systems being built. Especially if the person is at least moderately intelligent. There is no way to stop random failures or problems from happening because we are dealing with electronics and there is no way to stop people who honestly lack the required skills and intellect to properly follow directions and complete a build from attempting to do so anyhow despite the fact that they clearly should not. And we get both types here, and elsewhere, and in both cases it is probably only a small margin of the overall number of people who ARE building their own systems.
And even for those of us who knew what we were doing back in the day, it was not easier. It was a lot more difficult and a lot less streamlined than it is today, especially when it came to configuring that hardware. No way it was "easier" back in the days of old.
And even for those of us who knew what we were doing back in the day, it was not easier. It was a lot more difficult and a lot less streamlined than it is today, especially when it came to configuring that hardware. No way it was "easier" back in the days of old.
M2 connections are much faster than SATA 3 connections. MUCH faster.
An M2 PCIe 3x4 connection can move up to 4000 MB/s.
An M2 PCIe 4x4 connection can move up to 8000 MB/s.
A SATA 3 connection can move up to 6Gb/s, which is 750 MB/s.
- Jan 20, 2010
- 17,504
- 8,842
- 79,540
It should be noted that those are the upper limits of the interfaces. Actual drive performance varies, particularly if we're talking low/mid-market or older SSD models. The best advice is to look for good benchmarks of the drive in question.
Incidentally, I don't recall ever seeing a SATA SSD break the 600 MB/s barrier. The mid-500's seems to be where they max out (presumably, due to protocol overhead):
Source: https://www.tomshardware.com/reviews/samsung-870-evo-sata-ssd-review-the-best-just-got-better/2
How much one will actually notice the difference between SATA and NVMe depends a lot on what you're doing. My web browsing PC just has a SATA SSD and it's quite responsive. However, if you're loading huge games or copying large files (video files, VM images, etc.), then NVMe can really pay off.
Last edited:
Make sure it a PCIe M2 drive. There are also SATA M2 drives that will fit the same slot. They're no faster than comparable 2.5" SATA drives.
It should be noted that those are the upper limits of the interfaces. Actual drive performance varies, particularly if we're talking low/mid-market or older SSD models. The best advice is to look for good benchmarks of the drive in question.
Incidentally, I don't recall ever seeing a SATA SSD break the 600 MB/s barrier. The mid-500's seems to be where they max out (presumably, due to protocol overhead):
Source: https://www.tomshardware.com/reviews/samsung-870-evo-sata-ssd-review-the-best-just-got-better/2
How much one will actually notice the difference between SATA and NVMe depends a lot on what you're doing. My web browsing PC just has a SATA SSD and it's quite responsive. However, if you're loading huge games or copying large files (video files, VM images, etc.), then NVMe can really pay off.
Thus up to 4,000 or 8,000 MB/s for PCIe M2 NVMe drives and up to 750 MB/s for SATA 3 SSDs..
The same thing applies to both PCIe and SATA 3 connections. Neither ever hit theoretical limits. The ratio in the real world when using high quality drives for both is about the same. The fastest PCIe 3 M2 drives are about 7X faster than the fastest 2.5" SATA drives. The fastest PCIe 4 M2 drives are about 14X faster.
- Jan 20, 2010
- 17,504
- 8,842
- 79,540
Important: only some M.2 slots are compatible with SATA M.2 drives. Even if it is, these slots sometimes share resources with a motherboard's standard SATA connectors, meaning some of them are disabled when you use a SATA M.2 drive. Check your motherboard manual, for details.
Fortunately, most M.2 SATA drives are no longer sold. So, unless you find a cheap deal on an old M.2 drive, it's almost certainly not SATA. More importantly, if it's advertised as NVMe, then it's definitely not SATA.
Also, I expect most motherboards of the current generation no longer have SATA-compatible M.2 slots, but I'm not sure about that.
This is true. I'd hoped they would run at PCIe speeds, but it's still the old 600 Gbps of SATA-3.
- Jan 20, 2010
- 17,504
- 8,842
- 79,540
Like I said, I've never seen a SATA SSD achieve faster than mid-500's of MiB/s. If you know of a faster one, please link. Otherwise, I assume that's the practical limit. At roughly 77% of the theoretical speed, that's not great. I wouldn't expect there'd be nearly so much protocol overhead.
It seems NVMe drives can get a lot closer, though.
- Feb 21, 2020
- 2,595
- 3,930
- 17,570
I believe it's due to encoding overhead? I'm not positive on this, but I know SATA as an example has a 6Gbps interface. 6Gbps = 750 MB/s, but it uses 8b/10b encoding, so for every eight bits of data, 10 bits of bandwidth gets used. That means 6000 Mbps = 4800 Mbps of real throughput, which is 600 MB/s. But that's also Megabytes, not Mebibytes. If you do the 1024 math, 600 MB/s is 572 MiB, and then a bit of protocol inefficiencies would typically result in real-world performance topping out around 550 MiB/s.
PCIe 3.0 and later use 128b/130b encoding, which means it's 98.5% efficient use of bandwidth compared to SATA's 80% efficiency. And the NVMe protocol was specifically designed for SSDs, so it's supposed to be lower overhead. Still, a PCIe 3.0 x4 link can theoretically do 3.938 GB/s, or 3.667 GiB/s. 4.0 ups that to 7.877 GB/s, or 7.336 GiB/s. I think in testing we generally use GB/s and MB/s for our SSD reviews, not GiB/s and MiB/s, but there's still some overhead so I don't think we reach the theoretical limit.
And since I'm doing the SSD testing now... I can see that the best total throughput results we've seen from PCIe 4.0 x4 SSDs are around 7150 MB/s. Which means, based on the practical speed limit of 7877 MB/s, the M.2 PCIe stuff basically does worse than SATA in terms of reaching the theoretical max? I don't know, I probably need more sleep. LOL
- Mar 16, 2013
- 176,623
- 20,420
- 184,590
In theory, there is no difference between theory and practice.
In practice, however....
- Feb 21, 2020
- 2,595
- 3,930
- 17,570
There's a math joke about "close enough for all practical purposes..." 🙃
- Jan 20, 2010
- 17,504
- 8,842
- 79,540
Yeah, we really should use a better term. Probably something like "nominal bandwidth", since "theoretical" implies we actually did the math to account for all the various overheads, in order to work out the actual best-case scenario, when we mean nothing of the sort.
pixelpusher220
Distinguished
- Jun 4, 2008
- 227
- 112
- 18,760
TRENDING THREADS
-
-
News AMD says Intel's 'horrible product' is causing Ryzen 9 9800X3D shortages- Started by Admin
- Replies: 87
-
Discussion PSU recommendations and power supply discussion thread - Tom's hardware
- Started by Darkbreeze
- Replies: 2K
-
Discussion What's your favourite video game you've been playing?- Started by amdfangirl
- Replies: 4K
-
Question Question about DDR3 1600MHz and 1333MHz compatibility - Help with minidump files and BSOD
- Started by dimmel
- Replies: 3
-
Facebook
Twitter
Instagram