Windows 7 Benchmarks Show Little Improvement On Vista 369
snydeq writes "InfoWorld's Randall Kennedy examines Windows 7 from the kernel up, subjecting the 'pre-beta' to a battery of benchmarks to find any signs that the OS will be faster, more responsive, and less resource-intensive than the bloated Vista, as Microsoft suggests. Identical thread counts at the kernel level suggest to Kennedy that Windows 7 is a 'minor point-type of release, as opposed to a major update or rewrite.' Memory footprint for the kernel proved eerily similar to that of Vista as well. 'In fact, as I worked my way through the process lists of the two operating systems, I was struck by the extent of the similarities,' Kennedy writes, before discussing the results of a nine-way workload test scenario he performed on Windows 7 — the same scenario that showed Vista was 40 percent slower than Windows XP. 'In a nutshell, Windows 7 M3 is a virtual twin of Vista when it comes to performance,' Kennedy concludes. 'In other words, Microsoft's follow-up to its most unpopular OS release since Windows Me threatens to deliver zero measurable performance benefits while introducing new and potentially crippling compatibility issues.'"
Dupe? already? (Score:1, Informative)
http://tech.slashdot.org/article.pl?sid=08/11/10/1522246 [slashdot.org]
Re:What? (Score:5, Informative)
I have to agree with you here, mostly. Most of the tests make very little sense, and expecting W7 to be a rewrite is just stupid. Watching some of the W7-related PDC 2008 videos, I never got the impression that improving performance was their major priority, except perhaps for some tweaks for netbooks. Instead, most of the focus appears to be on other areas such as improved usability and power consumption. Not to mention that the M3 is a pre-beta build.
However, the OS can certainly have a significant impact on something like video encoding: differences in the scheduler or system calls/APIs can do that. Here's a somewhat outdated Vista vs XP [tomshardware.com] benchmark. The xvid and h.264 encoders are around 20% slower in Vista, and the impact is similar in some other cases, such as with WinRAR or UT2004. Differences of just a few percent can usually be ignored, but I find these significant. If somewhere between the release of Vista and W7 the maximum differences are lowered to around 5% compared to XP, whether with a service pack, new drivers or optimizations, I'd consider that good enough and possibly switch. After all, going from Win98 to XP also caused a drop in framerates, but was well worth it.
Re:MinKern anyone? (Score:5, Informative)
This explains nicely - http://blogs.zdnet.com/microsoft/?p=1707 [zdnet.com]
Short answer: mostly.
Re:Goodness me, what FUD (Score:5, Informative)
Vista is barely slower than XP on hardware bought within the last 2 years. It was fairly slower on RTM for many reasons, but vastly improved drivers & some colossal patches have put that to bed now.
When did this event occur? Last I tested Vista performance on this machine was with Crysis. That would be close to a year after Vista release. I got half the FPS compared to in XP. Half.
Apart from DX10 there is nothing in Vista that interests me that can't already be gotten for XP via third party applications. So far there aren't exactly a huge amount of DX10-only games, and unless the performance issue mentioned above has indeed been sorted it would be a moot point either way.
Re:What? (Score:1, Informative)
You are correct that video encoding is a cpu-bound process.
It would make absolute sense to have cpu-intensive, io-intensive and mixed-cycle processes used for benchmarks, not because the actual time it takes for the cpu to encode the video (or for the drive to read a sector) would be different, but the overhead imposed by the rest of the o/s and necessary processes.
It's this overhead that makes an o/s fast or slow, so the easiest way to measure it in a way that makes sense to the end-user, is to measure total wall-clock time for a cpu-intensive, an io-intensive and a mixed task.
Of course the cpu time required for the actual encoding will be the same (assuming the encoding libraries and the standard shared libraries e.g. msvc are the same, and assuming context switches have the same latency). But wall-clock time also includes the timeslices given to the o/s (kernel and processes) as well as the latency of the scheduler itself.
Re:so? (Score:5, Informative)
I've used XP (feels like I've used it forever), Vista (even longer), and Ubuntu (since 6.04).
In Ubuntu I rarely had any hardware problems. Ubuntu 8.10 recognizes all hardware without ANY problem. In Windows (same hardware!), I have to install at least 5 different hardware drivers. Mind you that this was not on cheap or obscure hardware.
The way I see the hardware issue is: a fresh Windows installation needs half a dozen drivers to be installed manually by the user. Finding drivers is usually pretty easy, especially for newer hardware. In Linux, you have two scenarios:
1. It Just Works (TM).
2. You have driver issues: in this case, you're better of having problems with older hardware that is more likely to be supported by some third party driver.
Office software: OpenOffice.org? It fits my needs (but I do not use it in a professional context so YMMV)
Games: agreed, this is Windows turf.
General acceptance: someday... (one can hope)
Re:Isn't this kind of the point? (Score:4, Informative)
Re:Goodness me, what FUD (Score:3, Informative)
Besides, they're testing a version of Windows 7 that is not even a beta drop. As such, it has yet to get its full code optimization, and by the time Windows 7 finally ships at the retail level expect substantial performance increases.
So, why isn't a service pack then? (Score:3, Informative)
If it is only minor improvements, then why is it not a service pack, or like 95 and 98 SE versions?
Why do you feel these small chances are worth another full price release?
Re:so? (Score:4, Informative)
Shouldn't you use LaTeX for writing your dissertation anyway?
Word always gave up on me on large documents with a lot of content.
Re:At least it is not slower (Score:3, Informative)
I've used Linux from 1999 until July this year when I finally gave up and bought myself a Macbook
No one asked if there's another operating system, the OP only mentioned that Ubuntu 8.10 is slower than previous versions.
Re:So, why isn't a service pack then? (Score:3, Informative)
Why a consumer would actually pay for it is another matter.
Perspective (Score:4, Informative)
General acceptance: someday... (one can hope)
Well, as with several other stuff, it's just a matter of perspective.
If by general acceptance, you specifically restrict to PC compatible computers. Yes, there aren't many Linux installation around (well except if you work in a Linux-oriented shop, like research, academics, etc.) Just, like Intel has a quasi-monopoly on CPUs for these machines.
But if you extend your definition to the more broad concept of linux being executed on an electronic device, the situation is completely different : you'll suddenly realise that the Penguin is already everywhere. ... most brands of routers run Linux.
Just take DSL routers : there's currently one in almost each house here around. Linksys, D-Link, Netgear,
In several european country, DSL ISPs are even bundling own-branded "{name_of_ISP}-Box" routers for VoIP / IPTV and Internet running embed Linux.
Yes currently Mac OS X and Linux only account for less than 30% of the market share, leaving more than 70% Windows Box. But the 100% total of those are connected to the net using boxes which 99.9% of the time run Linux.
Same goes for lots of the Media box connected to your TV set. Unless you built your own Windows Media Center HTPC, chances are, you bought a ready-to-use box.
In the USA, that is most likely a TiVo. Which runs Linux. Here in Europe, you probably bought from MediaMarkt one of those countless dead-cheap miniITX-based "add your own harddisk" noname asian box. Which most probably runs Linux too.
Same in an enterprise : the desktops will be probably running XP. The servers could be running Server 2003. But the routers, the cheap RAID/NAT box, the noname small network-to-printer bridges, and lots of other small electronic gizmo are running some form of embed linux.
On the desktop, Linux is facing strong competition from Windows and Mac OS X. On the other hand, in the embed market Linux is only facing what is basically a big mess of hundreds of small ad-hoc firmwares, with no clear leader, and that lot of manufacturer are dumping in favor of Linux, simply because it offers them a much better, more coherent and easier to maintain platform to work with.
Currently if you want to build some network-enabled gadget, either you re-invent the wheel and built your own solution. Or you just slap Linux with some micro server on it.
Trolls are still waiting for "the year of the Linux Desktop". They just missed that "the year of the Linux gizmo" has already happened long before.
If you look at electronics at a whole, Linux is suddenly a much stronger leader.
Just as, if you look at electronics at a whole, the battle for CPU dominance has long ago been lost to ARM & MIPS.
(with a bunch of PICs occupying a significant place for an even broader definition of electronics)
--
Beside....
Finding drivers is usually pretty easy, especially for newer hardware. In Linux, you have two scenarios:
1. It Just Works (TM).
2. You have driver issues: in this case, you're better of having problems with older hardware that is more likely to be supported by some third party driver.
And in lots of distribution, its just a matter of adding a new repository with additional drivers.
With some distro like openSUSE, that's basically just clicking on a ".ymp" link at the end of the explanation page on their wiki, and everything (adding the repository, installing the packages, etc.) is handled automagically.
That's it. Info page -> Click -> Installed.
Re:It may boot faster thanks to another photocopy (Score:1, Informative)
Windows has been performing parallel boot of device drivers and services since Windows XP RTM in 2001. Who broke out the "photocopier" again?
Re:so? (Score:3, Informative)
I too have a bit of a problem with OOo3. I am currently deployed, and my leadership puts out a newsletter to families back home every month to let them know what we're up to, how we're doing, etc. They write it in word, as some sort of a publication format (not written in publisher, but similar style). These newsletters NEVER format correctly in OOo3. Now this is not really OOo3's fault, so much as microsoft and their propriety, but it still illustrates the existing compatibility problems. I hope this type of thing is fixed in the future, but for now, it just doesn't work for me for everything.
Re:Time for Microsoft to [Start] over (Score:2, Informative)
Nobody was using Apple *to begin with* when they introduced OSX. They had less than 2% market share at the time, and it would have been a moot point to rewrite the handful of applications that worked on the Mac.
Quark? Photoshop? Final Cut? That's about the extent to which Apple was useful pre-OSX. The idea that people rewrote their applications is ridiculous -- a new spawn of programmers propped up enthralled with the new OS and started developing for it. Go ahead and Google it -- find the big applications that are used for the Mac nowadays -- they are all from NEW developers that never bothered to write a line of code for OS9 or 8.
That said, Microsoft's money maker is the fact that MOST applications maintain compatibility while giving more technological advantages over the previous OS. Windows XP for corporations, introduced desktop policy setting, active directory, centralized user management and control, and more. Don't think of the OS as important -- it's a delivery mechanism for their technologies. And some corporations are very happy with what they have in Windows XP/2003 Server for active directory control, user control and creation, etc.
Vista now gives you 'preferences' which allows you to change EVEN MORE in a centralized manner. People don't realize this because they all use it as a desktop OS, but in a corporate world these changes are valuable and useful. The backward compatibility is also with their OWN stuff, so while it's a nice idea to drop everything, create a VM for the apps etc -- it's unrealistic and would let Microsoft die in a fire. I'm sure many people here would love to see that, but it's not realistic to think that.
Apple's are great PCs for certain tasks, but there's nothing you cannot do in a Mac (or Linux) that Windows can't do. Now whether it does it BETTER is a different situation, and I'll fully admit I'd rather (and DO) have my website hosted on Apache with a RHEL backend and MySQL database because it's STABLE and it's FREE. I'd rather do video editing and audio editing on a Mac because the applications are better suited for that task.
But I can do all of that on Windows too. And that's why it's going to continue to sell, and they won't change a working formula. They are only building up the pieces AROUND Windows that make it a success -- and that includes Office, Sharepoint, System Center, BizTalk, and more all to tie your organization in, to use MS technologies that all leverage one another, and in comparison to some competitors -- are actually cheaper too.
It's not just Windows. People really need to get that.
Re:It may boot faster thanks to another photocopy (Score:3, Informative)
Re:So, why isn't a service pack then? (Score:1, Informative)
I keep wondering the same thing regarding OSX. Every few months they charge a few hundy for a point-release update, full of new bugs and failure. And yet, the Apple community is HAPPY to pay for it.
Win7, if you had bothered to read ANYTHING about it, will be substantially different in many areas. The underlying code is built on Vista, but a great many components are rewritten. It's sufficiently different enough that it will be a new version.
So if you have such a problem with it, start on Apple. If you succeed in getting them to "switch" their fraudulent business practices, I'll join in when you start questioning Microsoft and their motives.
Re:Goodness me, what FUD (Score:3, Informative)
I don't see how you contradict anything I said, yet the way you respond seems to indicate you disagree with me.
1) I did not say DX10 was of general interest. I said it was of interest to me (only reason for that is the possibility of games at some point requiring it). Are you saying I don't know what I'm interested in?
2) I didn't say it hadn't been improved. I said *the last time I tested* it was complete crap compared to XP in the *one* game I tested it with.
I did not say anything about hating Vista.
Re:What? (Score:3, Informative)
I don't know about launching apps, but boot time is supposedly faster in 7: http://lifehacker.com/5082336/windows-7-vista-and-xp-bootup-benchmarks-updated [lifehacker.com]
Re:The article is worthless (Score:3, Informative)
Do you really think that counting threads and memory footprint will give you any sort of indication of a systems performance? So, whatever those threads are really doing is not useful information?
By design Windows uses as much memory is available, as unused memory is of no value.
Unused memory is quite valuable. For instance, on a server where it needs to be able to quickly allocate memory to process a given request. Using all available memory would thus require paging stuff out to free memory for (for instance) a web server process to finish a script or similar request.
Very inefficient.
A performance indication would be to measure how much actual pagin is there when physical memory is exhausted by running process. Counting used memory is worthless. And counting threads and processes? Come on! What sort of analysis is this? Even if it were based on the final product (instead of a pre beta version), this analysis doesn't tell absolutely nothing.
Actually, counting threads, based off numerous more "techie types" knowledge of how Windows handles thread and process management, is a quite valid approach. Overloading a system with one of the worst thread schedulers on the PC is definitely not an approach that leads to performance benefits. Nor is assuming that everyone will have the latest and greatest hardware to make up for such a poor implementation - especially since MS continuously hypes their "latest and greatest" OS as something everyone should upgrade to. Anyone remember the Vista Upgrade Advisor - and how what it thought was a Vista Capable machine oft times was not? So... following a similar scenario, the use of tons of threads as the norm, on an OS that MS expects the world to upgrade to, is a design/implementation flaw that will just cause angst to all the poor unsuspecting end users who upgrade their XP machines finally to W7.
Not that I would expect that Win7 uses fewer resources that Vista. It would be a great thing if, coming a few years later, it used the same level of resources (meaning it should be able to run in machines over five years old) but expecting it to consume fewer resources is delusional.
Why? Because it is Microsoft? Or is there a technical reason? Keep in mind the far greater resources that Vista needed over XP or their server line (2003 and earlier) did not translate into any meaningful benefits for the end user. There was no reason for the increased bloat and resource usage at such a level. Vista should have required more resources - but not nearly as much more as the final product did.
Thus, if Microsoft were to actually get this one right, it should use less resources than Vista. More than XP? Yes. But still less than Vista.
Of course, that is unlikely to be the case, as is evidenced by their latest attempts to make the OS appear to be faster instead of actually making it faster (ie: tweaks to the UI to give the appearance of a snappier response, while all the "real work" still takes the same amount of time or more).
Performance today has much less to do with resource usage than with responsiveness and proactivity anyway.
Again, I beg to differ... but then again, I like running numerous things at a time... and knowing I have available memory (without having the need to wait for stuff to be paged out) when I am running things. Everything you argue for makes no sense - except in the respect that Windows needs as much as possible to perform as "adequately" as possible.
Re:so? (Score:3, Informative)
Just don't take your dissertation somewhere to have it printed and expect it to come out looking like it does on your computer.
When I worked at a small print shop/service bureau, we had a what we called "The Word Disclaimer" form that stated we could not guarantee the quality of any output from a Word file. It was created after many problems with clients who were irate when the Word documents we printed for them didn't look the way they expected them to look. Anyone submitting a MS Word document to us had to sign a copy before we would agree to print it up.