Why I find Windows 8.1 more suitable than Windows 7

[xpclient]

They readjusted WEI in every release so it's not directly comparable.

@Noel I am glad too. Yes I couldn't believe it myself. There was a time when I loathed everything after XP. I had given Vista a thorough run and tried my best to migrate to it. I won a laptop as 1st prize in 2006 in a Microsoft Virtual TechDays contest. Came with Vista. Was awful. Went back to XP. Then when everyone was on Windows 7 and was praising how great it was, I used to hate it because it took away the Classic menu, ruined the Taskbar and had many Explorer regressions which they wouldn't fix because of their "our way or the highway attitude". I think while arguing about some minor feature about Explorer, you too told me a couple of times in Microsoft's forums that it was time to embrace change and that I was clinging to the past or something similar.  

I was fully aware of how much technically Windows 7 was better than XP for the most part. I like to learn, read and write about Windows a lot. Wrote 70% of Wikipedia's series of articles on Windows and its new or removed features. But I objected to the way Microsoft started to behave after XP - they seem to have started believing that they can never do anything wrong, and flat out refuse to fix things. It was only after Classic Shell, 7+ Taskbar Tweaker etc and other tools got created to make up for the user experience deficiencies of Win7 that I warmed up to it.

Microsoft did the same mistakes again with Windows 8 - lots of broken features, no proper continuity, omission of things which worked on 7 and forcing people to embrace a fresh start even if it was not justified at all. Being obsessive about the UI, I never believed certain things about Windows 8 could be fixed if ever. But they can indeed be fixed and I was just as surprised how much the system could be molded to make it behave I wanted.

When I got a new PC last year there was nothing holding me back performance wise from not moving to a newer version of Windows *if it was indeed better* for the most part. They always remove many useful features while trying to add something of value so you always have to make huge compromises but the community did all the hard work again. I realized that most regressions in Windows 8 were again fixed by the community and Microsoft had done nothing really to fix it.

I think if we can't modify the system and are forced to use it as it is, then I will go back to Windows XP. But because Windows 7 and 8.1 can be re-configured and heavily tweaked plus they will remain that way for as long as drivers exist, it is worth taking advantage of the genuine improvements Microsoft did make to 8.1 and try to fix the regressions ourselves, rather than straight out rejecting everything they have done.

With Windows 10, unfortunately I am finding nothing of value at all. And I am finding it extremely hard to fix anything and have it remain like that. It's wild and out of control. I have never seen software that treats its users so condescendingly. Seeing how awful it is, I think investing in 8.1, not 7 makes more sense for me at this time. I do keep Windows 7 and 10 installed, just to keep a check on how better or worse they are. There aren't 1:1 feature replacements for every feature 7 has in 8.1 but it's close with third party apps.

Edited August 12, 2016 by xpclient

[NoelC]

Well, if nothing else 8.1 will be supported the longest of any perpetually licensed version.  Maybe by the time end of support becomes an issue someone will have had time to invent a worthy successor.

-Noel

[NoelC]

10 hours ago, jaclaz said:

Hmmm.

Actual numbers tell a different story.

The Windows Experience in Windows 7 "assesses key components on a scale of 1.0 to 7.9"
The Windows Experience in Windows 8.1 (and 10)  "assesses key components on a scale of 1.0 to 9.9"

I need to dig out the detailed reports and we can look at the more detailed numbers, instead of trying to derive something from the overview numbers, because Gaurav's right - they're all jumbled up.

-Noel

[jaclaz]

4 hours ago, NoelC said:

I need to dig out the detailed reports and we can look at the more detailed numbers, instead of trying to derive something from the overview numbers, because Gaurav's right - they're all jumbled up.

-Noel

Yep , I only analyzed the data you posted, highlighting how they are - to say the least - deceiving.

Besides the "change of scale" (which as seen can be easily made comparable) the striking difference in the two graphics categories is really hard to be justified.

The only thing that comes to my mind is the DirectX 12 vs DirectX 11 vs DrirectX n which (not really my field of interest/knowledge) might well be "game changers" (if you pardon me the pun), but that hardly makes an impact on anything that is non-high-speed-games, and besides apart the usual MS hype on the matter, I haven't seem them praised so enthusiastically by the (admittedly few) gamers I know if not for specific games.

jaclaz

[xpclient]

I found an app called NovaBench that does a quick test of CPU speed, RAM speed, disk write speed and GPU rendering speed similar to Windows Experience Index. I ran it with identical setup on my Windows 7/8.1/10 on the same PC. 8.1 came ahead followed by 7 and 10 was slowest. But I think these scores can't be accepted because there is little to no explanation given behind how NovaBench calculates the score and how much weightage is given to each score.

Edit: Scores removed because as Jaclaz points out in the next post, how NovaBench assigns those points is not clear. The benchmark weightage for disk drive was not purely indicative of performance.

Edit 2: Actually, digging deeper into their scores, I found out that all the scores except the disk scores are fair for comparison purposes. The disk score takes into account the partition size  which is #FAIL. Size is definitely not an indicator of performance, although for SSDs, larger SSDs perform better. You could still compare the write speed though.

Edited August 14, 2016 by xpclient

[jaclaz]

@xpclient

Those seem some different (but still largely meaningless) metrics.

I quickly checked the novabench site but there is nothing there explaining how they assign those *random* points and the relative weight on the total.

And they do look a lot random.

Are you telling me that with Windows 8.1 hard disk is 19.5/12.5-1=56% faster than in 7 (and as well 19.5/13-1=50% faster than in 10)?

Come on  it is entirely possible that some optimized drivers can increase a disk speed by 5% or 10% but 50%?

And NoelC experienced actually the opposite, with Windows 8.1 being slower when it comes to disk:

(which is actually consistent with the WE data before)

jaclaz
 

[NoelC]

Over time I've come to realize that the file system in 8.1 just feels a little, well, different than in 7.  It's about what you'd expect from a system that's had a little more maintenance and a few more bugfixes.

Some operations - e.g., searching for a file on the disk with a tool like Stefan Kung's grepWin - can be faster in 8.1, while enumerating file info in an Explorer Properties dialog can be much slower.  And accessing files across Windows Networking can be quite a bit faster in 8.1 than 7. 

There's not really one overall trend, but I'd say there's some additional optimization apparent in some parts of 8.1.

And maybe it's because of all the years of bugfixing, but IMO 8.1 handles "fringe" situations better.  As an example, over the last few days I've been doing serious stress testing of some of my software, and have caused things like "out of RAM" situations (which is impressive on a 64 bit system with 48 GB of actual RAM and a 64 GB pagefile).  I haven't seen any instability.  Music (e.g., Pandora.com) hasn't even stuttered.

-Noel

Edited August 14, 2016 by NoelC

[jaclaz]

The data you posted (created by MS own utility, remember) is saying that "Primary Hard drive" is performing VERY NOTICEABLY worse, whilst "Gaming graphics" (whatever it is) is performing VERY NOTICEABLY better.

Real world experience tells us instead that they are "different" but not so dramatically so.

You have to choose, you either believe yourself and your observations or the MS metrics, since results are so diverging.

If you choose your own experience (as it would be logical) it is clear that the WE metrics are largely meaningless, and are not valid comparing tools between different OS's on a same (virtual or real) machine.

As a matter of fact outside of "my WEI is bigger than yours" kind of discussions WEI is totally meaningless.

jaclaz

Edited August 14, 2016 by jaclaz

[NoelC]

My only intent was to spark discussion.  I like to post pictures. 

I should probably do some more head to head objective testing with the newest Win 10 and Win 7 and 8.1 with all the updates.  That's a lot like work, though. 

-Noel

Edited August 14, 2016 by NoelC

[xpclient]

Well according to the developer of NovaBench, the tests include floating point and integer operations, MD5 hashes to test CPU performance, RAM speed (MB/s), 3D graphics speed (FPS), Drive write speed (MB/s) and partition capacity because NovaBench was designed to benchmark not purely performance but value as well!  Partition size is definitely not an indicator of performance, although for SSDs, larger SSDs perform much better than smaller ones. The developer is planning to remove the disk partition size from the disk rating score so comparisons across Windows versions are more valid. I suggested him to include read speed test instead. We really need a simple, free, lightweight, fast utility for quickly benchmarking Windows performance and get consistent results which can be compared across Windows versions. Most of the benchmarking apps are heavyweight, full-featured commercial apps (e.g. SiSoft Sandra) or specialized to test only 1 performance area of the Windows OS (3D Mark, Passmark) etc.

Now ignoring the flawed disk rating of NovaBench, the write speeds on my PC were as follows: Windows 8.1: 98-109 MB/s,    7: 88-94 MB/s,     10: 98-100 MB/s.

[jaclaz]

14 minutes ago, xpclient said:

Now ignoring the flawed disk rating of NovaBench, the write speeds on my PC were as follows: Windows 8.1: 98-109 MB/s,    7: 88-94 MB/s,     10: 98-100 MB/s.

Ah well, then it's ok, we can continue the discussion using some more meaningless metrics.

I feel then free to add a reference to my (completely twisted) TES or Tweaking Effort Scale  :

jaclaz

 

[NoelC]

Personally I think PassMark PerformanceTest does a pretty good job of measuring the performance that matters.  That's what I used to compare systems before.  I've always felt that higher numbers e.g., in their Advanced Disk Test section correlated well with a system being more responsive with disk access.

Regarding how important your TES is, jaclaz...

I tweaked Win 8.1 once (though of course I have been learning about it and refining things all along).  I'm running the same install that I put in at the end of 2013.  It's still awesome.  I don't consider that a lot of ongoing effort as compared to the utility I got out of it.  

Plus I wrote it all down, so if I actually DO need to do it again, it's pretty easy.  I've been through VM setups with my own guide and I can get a system fully functional in an hour or two.

What matters more, I think, is how available the system is for actual use.

My setup runs 24/7 without fault, providing a secure, rich computing environment I can rely upon to be able to do my work and run my business without worry.  A couple of nights ago I quit working at 3am, and I just left everything open and went to bed.  When I regained consciousness everything was just where I left it, and the system had done all its backups and maintenance activities.  Plus I had worked out the answer to the problem I was struggling with.

-Noel

[NoelC]

OK, I've taken some objective measurements...

Conditions:

VMs in VMware with Win 8.1 host, all hosted on the same RAID 0 SSD array, each has 4 logical processors, 8GB of virtual RAM, 1920 x 1200 display size, one C: partition of 100 GB.  Each OS within the VM is updated, similarly configured (generally leaned-down), and provides a similar desktop experience (though I haven't got Aero Glass back working on the Win 10 system yet).

The summary of all the tests is shown at the upper-left, and the advanced disk test is shown at bottom.  I set up a mix of read and write file system operations with normal caching, in order to simulate actual disk operations as experienced by a user.

Win 7:

Win 8.1:

Win 10:

We could debate the credibility of testing VMs, but I say it's a practical way to ensure the configured "hardware" is similar.

I suggest that the variances shown imply actual differences the level of efficiency in the implementation of the operating system.  Note the significant drop in cached disk performance in the Win 10 Anniversary Edition.

I don't claim this is an end-all set of numbers, but it illustrates the differences I'm seeing here.

-Noel

Edited August 14, 2016 by NoelC

[jaclaz]

Good, some more random numbers, which are put into the attached table, with some short comments.

Those numbers (just like the previous ones) simply make no sense whatever, I guess that everyone here is saying (or was saying) that overall there are some slight differences perceived in the working of the different OSes, which would be normal BTW, as there has been no particularly sensational changes introduced in the OS, just some little adjustments.

Here it seems like some subsystems object of the test are like years ahead while some are centuries behind,

jaclaz

 

[NoelC]

The only thing I can offer to explain the differences between Win 7's Disk Mark test and that of the other VMs is that the Win 7 VM is set up with a virtual SCSI controller while the others have a virtual IDE controller.  Perhaps that has influenced PassMark's test.  I do trust those graphs in the Advanced Disk test, though.  THEY mirror very well what it feels like when one actually uses the file system to do real things.

I can believe that the kernel tries to step out of the way more in the newer systems, e.g., the runs between interrupts might be longer and thus CPU-intensive activities run longer without the system going off and doing something else. 

And I could easily believe that the Windows 8 and 10 desktops are negatively affected because of all the BS Microsoft has added to be able to handle Metro / Modern / Universal Apps.

I can't explain the disparity in the Memory Mark test.

But what you see is what I saw, and you'd be right to doubt the consistency of the testing based on just those results.

I'm open to suggestions for gathering more "real world" test data.  Maybe I should copy a big video file to the TEMP area, play it, and track CPU usage?

-Noel

Edited August 15, 2016 by NoelC