GNOME Circle Officially Announced For Letting More Apps/Libraries "Join GNOME"

[mos87]

GNOME Circle J*rk?

[mppix]

A few unlucky people who have put in the wrong gnome shell extentions have be on the receiving end of mega leaks. Part of the reason why they are needing proper quality control and that should come in gnome 4.0.

.....

This has been only true some of the time. When things are working right. Currently people can find themselves out of ram if they start playing around with gnome shell extensions due to bugs in them.

Shell extensions can be developed by anyone. Yes they have arguably too much power [hence the new development] but users should also be reasonably responsible for what they install on their machines [this is true on any OS].

What you have missed hibernation puts a upper limit on how big you can in fact go before you start really losing its advantage. Part of the problem is the time to hibernate. More ram used longer the time to hibernate the more power you use to hibernate and unhibernate there does truly come a point where its better to shutdown than hibernate. Remember shutdown and boot again means you have lost where you were working in most applications so this does come with a productivity disadvantage.

The objectives of hibernation is:
1) to save power and save boot time with laptops and the like there is a upper limit with ram usage when this ceases to work to save time.
2) maintain the state of what you were doing so you don't have to start from scratch. Of course you lose this if you shutdown.

Sleep you still have the ram active eating into your battery this is also a factor why you hibernate instead of sleep is so you are not eating battery.

Hibernation advantage (vs sleep) is power consumption when off.
Storing/restoring memory with pcie SSDs has become faster but I think any of my [non server] systems boot faster (usually measured in 1-3s).
Anyhow, pcie5+DDR5 will speed things up further.

Yes its true modern desktops can support up to 128GB of ram. But that is not the limit where inefficiencies start causing trouble. Having hibernation work well with laptops has a limited tolerance for inefficiencies. Apple M1 processor soc only having 16G of ram is not for no reason in a portable you don't really want above 32G of ram heck 16G of ram is better when you are going to hibernate or sleep to extend battery life/save power.

128GB ram system in sleep mode pulls down roughly 8x the power of a 16GB ram system in sleep. There is really not plenty of room for scaling up with the inefficiencies between hardware that has locked memory sizes and the power issues of portable devices. There is a need if possible to keep the desktop under 512 megs of memory to operate.

Hard point the is just because a desktop computer can support 128GB of ram does not mean it should so it should not be held out as excuse for bad memory management. Also 128GB of ram as a requirement is a high cost to pay if we let it get to that point.

Gnome has been having some serous problems in their gnome extensions and a few other places with memory management. Memory leaks and memory miss management in desktop environments does need to be taken as serous problems because Memory in a machine is a finite resource without the owner spending more money and sometimes spending money is not option really to fix it because more ram will result in issues like reduced battery life of the device.

Memory will continue to grow. Expect 64GB and then 128GB to become standard for workstations some time after DDR5 hits.

Also, I used Mac laptops for more than 10y. I don't think I ever tried hibernation, just sleep.

[oiaohm]

Shell extensions can be developed by anyone. Yes they have arguably too much power [hence the new development] but users should also be reasonably responsible for what they install on their machines [this is true on any OS].

Not exactly. You don't have to put the shell extensions in the key core part so cannot be properly contained. The idea that this is the users problem alone means the one that decided to add extensions did not have responsibility to put containment around them.

Hibernation advantage (vs sleep) is power consumption when off.
Storing/restoring memory with pcie SSDs has become faster but I think any of my [non server] systems boot faster (usually measured in 1-3s).

That is exactly the problem The difference between normal booting and hibernation has been getting bigger as the amount of ram goes up.

Anyhow, pcie5+DDR5 will speed things up further.

https://www.techspot.com/review/1893-pcie-4-vs-pcie-3-ssd/

One problem the speed of SSD are not doubling with pcie version changes. PCIe 4.0 is by bus double the speed of PCIe 3.0 but SSD only when up by 30 percent. Its like how graphics cards that support PCIe 4.0 are not finding any more advantage having more than a 8x slot but with PCIe 3.0 with GPU you could get some advantages out to PCIe 3.0 x16 yes this is crossing the threshold. PCIe 5.0 if nothing changes with GPUs should see a 4x slot be suitable for GPU the fun of the threshold effect

SSD/Flash and GPU turn out to have a max speed anything more than that provides no more advantage. pcie5 is likely to speed up Hibernation restore less than PCIe 4.0 did in fact the forecast is absolutely nothing will be gained in performance for hibernation with PCIe 5.0 . PCIe bus is basically not the bottleneck for hibernation with PCIe 4.0 ssd the flash storage is. This is why increasing ram in machines you wish to hibernate are not going to help things..

Memory will continue to grow. Expect 64GB and then 128GB to become standard for workstations some time after DDR5 hits.

That is work stations plugged into wall. Laptops are portables are not growing at the same speed at all.

Also, I used Mac laptops for more than 10y. I don't think I ever tried hibernation, just sleep.

Mac laptops do have auto modes around hibernation. So will in fact at time do hibernation when you asked for sleep.

[mppix]

Not exactly. You don't have to put the shell extensions in the key core part so cannot be properly contained. The idea that this is the users problem alone means the one that decided to add extensions did not have responsibility to put containment around them.

That is exactly the problem The difference between normal booting and hibernation has been getting bigger as the amount of ram goes up.

https://www.techspot.com/review/1893...vs-pcie-3-ssd/
One problem the speed of SSD are not doubling with pcie version changes. PCIe 4.0 is by bus double the speed of PCIe 3.0 but SSD only when up by 30 percent. Its like how graphics cards that support PCIe 4.0 are not finding any more advantage having more than a 8x slot but with PCIe 3.0 with GPU you could get some advantages out to PCIe 3.0 x16 yes this is crossing the threshold. PCIe 5.0 if nothing changes with GPUs should see a 4x slot be suitable for GPU the fun of the threshold effect

SSD/Flash and GPU turn out to have a max speed anything more than that provides no more advantage. pcie5 is likely to speed up Hibernation restore less than PCIe 4.0 did in fact the forecast is absolutely nothing will be gained in performance for hibernation with PCIe 5.0 . PCIe bus is basically not the bottleneck for hibernation with PCIe 4.0 ssd the flash storage is. This is why increasing ram in machines you wish to hibernate are not going to help things..

That is work stations plugged into wall. Laptops are portables are not growing at the same speed at all.

Mac laptops do have auto modes around hibernation. So will in fact at time do hibernation when you asked for sleep.

Well, some pcie4 SSDs are now nearly doubling the (peak sequential) read/write performance of pcie3. However, few consumer hardware can actually sustain the load for long times (didn't check if a full ram copy would go beyond the capabilities of typical drives; it certainly exceeds the buffer/RAM size of most drives).
For pci5, I think we will see drives that peak at twice the pci4 speeds even in the consumer space. However, we will have to see if SSDs will actually get faster for real-world useage (random read/write; sustained read/write) moving forward or just cheaper (QLC etc..)

This brings us to the question whether hibernation -as is- actually works or if we need a better approach/implementation.

[oiaohm]

Well, some pcie4 SSDs are now nearly doubling the (peak sequential) read/write performance of pcie3.

Its absolutely not read write.

https://www.anandtech.com/show/16087...e-4-ssd-review

Sequential Read		6400 MB/s	6900 MB/s	7000 MB/s
Sequential Write	SLC	2700 MB/s	5000 MB/s	5000 MB/s
TLC	500 MB/s	1000 MB/s	2000 MB/s

Yes read has been able to be doubled this time around. Write is not doubled. Do note the TLC section so if you run out of SLC space and have to go TLC while doing hiberantion you are going to be slower with the new PCIe 4.0 devices vs the old PCIe 3.0 devices.

SLC Write Cache Size	Min	4 GB	4 GB	6 GB	TBD
Max	49 GB	94 GB	114 GB

This bit is also important. This high write speed is limited capactity. 128G of ram is on the wrong side of it. Amount of usable SLC shrinks as the drive fills and can be filled by other things the user has been doing.

However, few consumer hardware can actually sustain the load for long times (didn't check if a full ram copy would go beyond the capabilities of typical drives; it certainly exceeds the buffer/RAM size of most drives).

The write side has a limited amount in consumer drives that can do sustained speed until they run out of the SLC. 32G is basically the upper limit unless the amount of SLC increases a lot. To be sure ram will fit in the SLC with the above drive you need to be under 4-6G of ram.

For pci5, I think we will see drives that peak at twice the pci4 speeds even in the consumer space. However, we will have to see if SSDs will actually get faster for real-world useage (random read/write; sustained read/write) moving forward or just cheaper (QLC etc..)

QLC and TLC are really not fast on the write. SLC is fairly much capped out on write speed as well.

This brings us to the question whether hibernation -as is- actually works or if we need a better approach/implementation.

Most likely hibernation broke. We cannot keep on expanding ram and doing hibernation as we were. Changing from HDD to SSD allowed us to expand hibernation ram size without major performance problems for a far while. SSD has stopped being the gold goose laying the gold eggs of performance for hibernation to keep on working if we keep on increasing ram.

That max there of 114GB might sound fine to be doing a 64G hibernate lets say you had been encoding a video or something like when you hit hibernate over half 114GB could have been used by what the user had been upto or reduced due to the drive being filled with stuff.

Past 32G of ram is where hibernate issues really start showing themselves with SSD drives. Next generation of drives we cannot expect to be much better. We are trading away performance of SSD for capacity in consumer drives by a large amount.

[mppix]

That max there of 114GB might sound fine to be doing a 64G hibernate lets say you had been encoding a video or something like when you hit hibernate over half 114GB could have been used by what the user had been upto or reduced due to the drive being filled with stuff.

Past 32G of ram is where hibernate issues really start showing themselves with SSD drives. Next generation of drives we cannot expect to be much better. We are trading away performance of SSD for capacity in consumer drives by a large amount.

How about compression?

[oiaohm]

How about compression?

Horrible compression is most likely not going to be a work around. Problem is auto-hibernation. You put your laptop to sleep when the battery gets low its auto set to come out of sleep and hibernate before it runs out of battery hopefully. This is also why time comes critical and write speed is critical. Few more seconds may be a few more seconds of runtime the battery does not have in a auto-hibernation. Running CPU though heavy process to compress is also going to eat more battery.

Something that could help is marking applications that can be straight up terminated in a hibernation event but that does not help all the applications you need state of are not storing their own state. Another option of course is cap memory to like 16G-32G when on battery then it comes why did you buy 128G for a portable.

Thing to remember the key objective of hibernation is to save users state so they don't lose where they are in workflow. Reboot does not solve that problem. Person put laptop into sleep mode not thinking they would lose the state that why you have auto hibernation and hybrid sleep hibernation.

The old hybrid sleep workaround is in fact abusive on SSD. This is where hibernation is attempting to be in background writing the ram state out all the time to keep the amount needs to be written out when the final hybrid sleep down comes this was a hack for low performance HDD. SSD write limits not the wise idea.