If this is your first visit, be sure to check out the FAQ by clicking the link above. You may have to register before you can post: click the register link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below. |
|
|
Thread Tools | Rate Thread | Display Modes |
#1
|
|||
|
|||
Does 10 TP seem slower to anyone? Running on VMware Player.
I have vmware player 7.0 and have a whole list of machines, XP,7,8.1,10TP,Mint17,Ubuntu.
All seem to run pretty good, windows 7 is pretty responsive, but 10TP just seems to be sluggish as all get out. I've got all the machines set with the same 4G memory and 2 processors and video settings. The only variable I can come up with is the OS itself. |
Ads |
#2
|
|||
|
|||
Does 10 TP seem slower to anyone? Running on VMware Player.
Big_Al wrote:
I have vmware player 7.0 and have a whole list of machines, XP,7,8.1,10TP,Mint17,Ubuntu. All seem to run pretty good, windows 7 is pretty responsive, but 10TP just seems to be sluggish as all get out. I've got all the machines set with the same 4G memory and 2 processors and video settings. The only variable I can come up with is the OS itself. What does your Task Manager show (both inside the VM and outside the VM) ? Is a core railed to 100% ? VirtualBox sometimes goes into a loop, with older OSes (the Win2K bug). If you run SuperPI both inside and outside the VM, does the bench conducted within the VM get to within 90 percent of the host performance ? x86 on x86 hosting should get pretty close to full performance. http://web.archive.org/web/200710261...pi_mod-1.5.zip That's a different issue than the one demonstrated here. In this example, the user has a quad core with two dual core silicon dies sharing the FSB, and cache coherency traffic is making a difference (the newer OSes don't migrate tasks carelessly, like the old OS did). The OS really shouldn't make all that much difference to vanilla code execution. The results would be closer on a single core CPU (maybe someone can test that for me :-) ). http://blog.testfreaks.com/informati...vs-vista-vs-7/ SuperPI-32m WinXP 18 minutes 1 second = 1081 sec Vista 18 minutes 4 seconds = 1084 sec Win7 17 minutes 43 seconds = 1063 sec That's a single threaded benchmark, with a cache dependency. Selecting "32m" is an attempt to nullify the impact of a CPU with a large L2 or L3. You make the "number of digits" selection large enough, so the cache won't provide a big advantage. That's only important when comparing two different CPUs (so the CPU with the small cache, isn't penalized, or the CPU with the large cache doesn't end up looking too heroic). If you want to use a smaller number of digits when testing inside and outside the VM, there would not be a "cheating factor" on cache to worry about. Both situations run on the same hardware. Once you've assured yourself the benches run inside and out at the same speed, you've looked at Task Manager to see if anything is rails... then your problem is a graphics issue. Make sure VirtualBox has the experimental 3D support turned on. HTH, Paul (who has not done *any* of these benches on Win10TP yet...) |
#3
|
|||
|
|||
Does 10 TP seem slower to anyone? Running on VMware Player.
Big_Al wrote:
I have vmware player 7.0 and have a whole list of machines, XP,7,8.1,10TP,Mint17,Ubuntu. All seem to run pretty good, windows 7 is pretty responsive, but 10TP just seems to be sluggish as all get out. I've got all the machines set with the same 4G memory and 2 processors and video settings. The only variable I can come up with is the OS itself. What does your Task Manager show (both inside the VM and outside the VM) ? Is a core railed to 100% ? VirtualBox sometimes goes into a loop, with older OSes (the Win2K bug). If you run SuperPI both inside and outside the VM, does the bench conducted within the VM get to within 90 percent of the host performance ? x86 on x86 hosting should get pretty close to full performance. http://web.archive.org/web/200710261...pi_mod-1.5.zip That's a different issue than the one demonstrated here. In this example, the user has a quad core with two dual core silicon dies sharing the FSB, and cache coherency traffic is making a difference (the newer OSes don't migrate tasks carelessly, like the old OS did). The OS really shouldn't make all that much difference to vanilla code execution. The results would be closer on a single core CPU (maybe someone can test that for me :-) ). http://blog.testfreaks.com/informati...vs-vista-vs-7/ SuperPI-32m WinXP 18 minutes 1 second = 1081 sec Vista 18 minutes 4 seconds = 1084 sec Win7 17 minutes 43 seconds = 1063 sec That's a single threaded benchmark, with a cache dependency. Selecting "32m" is an attempt to nullify the impact of a CPU with a large L2 or L3. You make the "number of digits" selection large enough, so the cache won't provide a big advantage. That's only important when comparing two different CPUs (so the CPU with the small cache, isn't penalized, or the CPU with the large cache doesn't end up looking too heroic). If you want to use a smaller number of digits when testing inside and outside the VM, there would not be a "cheating factor" on cache to worry about. Both situations run on the same hardware. Once you've assured yourself the benches run inside and out at the same speed, you've looked at Task Manager to see if anything is rails... then your problem is a graphics issue. Make sure VirtualBox has the experimental 3D support turned on. HTH, Paul (who has not done *any* of these benches on Win10TP yet...) |
#4
|
|||
|
|||
Does 10 TP seem slower to anyone? Running on VMware Player.
Paul wrote on 1/15/2015 8:19 PM:
Big_Al wrote: I have vmware player 7.0 and have a whole list of machines, XP,7,8.1,10TP,Mint17,Ubuntu. All seem to run pretty good, windows 7 is pretty responsive, but 10TP just seems to be sluggish as all get out. I've got all the machines set with the same 4G memory and 2 processors and video settings. The only variable I can come up with is the OS itself. What does your Task Manager show (both inside the VM and outside the VM) ? Is a core railed to 100% ? VirtualBox sometimes goes into a loop, with older OSes (the Win2K bug). If you run SuperPI both inside and outside the VM, does the bench conducted within the VM get to within 90 percent of the host performance ? x86 on x86 hosting should get pretty close to full performance. http://web.archive.org/web/200710261...pi_mod-1.5.zip That's a different issue than the one demonstrated here. In this example, the user has a quad core with two dual core silicon dies sharing the FSB, and cache coherency traffic is making a difference (the newer OSes don't migrate tasks carelessly, like the old OS did). The OS really shouldn't make all that much difference to vanilla code execution. The results would be closer on a single core CPU (maybe someone can test that for me :-) ). http://blog.testfreaks.com/informati...vs-vista-vs-7/ SuperPI-32m WinXP 18 minutes 1 second = 1081 sec Vista 18 minutes 4 seconds = 1084 sec Win7 17 minutes 43 seconds = 1063 sec That's a single threaded benchmark, with a cache dependency. Selecting "32m" is an attempt to nullify the impact of a CPU with a large L2 or L3. You make the "number of digits" selection large enough, so the cache won't provide a big advantage. That's only important when comparing two different CPUs (so the CPU with the small cache, isn't penalized, or the CPU with the large cache doesn't end up looking too heroic). If you want to use a smaller number of digits when testing inside and outside the VM, there would not be a "cheating factor" on cache to worry about. Both situations run on the same hardware. Once you've assured yourself the benches run inside and out at the same speed, you've looked at Task Manager to see if anything is rails... then your problem is a graphics issue. Make sure VirtualBox has the experimental 3D support turned on. HTH, Paul (who has not done *any* of these benches on Win10TP yet...) 90% of that is over my head. And I don't have Virtual Box. VMware Player as stated in the OP. |
#5
|
|||
|
|||
Does 10 TP seem slower to anyone? Running on VMware Player.
Paul wrote on 1/15/2015 8:19 PM:
Big_Al wrote: I have vmware player 7.0 and have a whole list of machines, XP,7,8.1,10TP,Mint17,Ubuntu. All seem to run pretty good, windows 7 is pretty responsive, but 10TP just seems to be sluggish as all get out. I've got all the machines set with the same 4G memory and 2 processors and video settings. The only variable I can come up with is the OS itself. What does your Task Manager show (both inside the VM and outside the VM) ? Is a core railed to 100% ? VirtualBox sometimes goes into a loop, with older OSes (the Win2K bug). If you run SuperPI both inside and outside the VM, does the bench conducted within the VM get to within 90 percent of the host performance ? x86 on x86 hosting should get pretty close to full performance. http://web.archive.org/web/200710261...pi_mod-1.5.zip That's a different issue than the one demonstrated here. In this example, the user has a quad core with two dual core silicon dies sharing the FSB, and cache coherency traffic is making a difference (the newer OSes don't migrate tasks carelessly, like the old OS did). The OS really shouldn't make all that much difference to vanilla code execution. The results would be closer on a single core CPU (maybe someone can test that for me :-) ). http://blog.testfreaks.com/informati...vs-vista-vs-7/ SuperPI-32m WinXP 18 minutes 1 second = 1081 sec Vista 18 minutes 4 seconds = 1084 sec Win7 17 minutes 43 seconds = 1063 sec That's a single threaded benchmark, with a cache dependency. Selecting "32m" is an attempt to nullify the impact of a CPU with a large L2 or L3. You make the "number of digits" selection large enough, so the cache won't provide a big advantage. That's only important when comparing two different CPUs (so the CPU with the small cache, isn't penalized, or the CPU with the large cache doesn't end up looking too heroic). If you want to use a smaller number of digits when testing inside and outside the VM, there would not be a "cheating factor" on cache to worry about. Both situations run on the same hardware. Once you've assured yourself the benches run inside and out at the same speed, you've looked at Task Manager to see if anything is rails... then your problem is a graphics issue. Make sure VirtualBox has the experimental 3D support turned on. HTH, Paul (who has not done *any* of these benches on Win10TP yet...) 90% of that is over my head. And I don't have Virtual Box. VMware Player as stated in the OP. |
#6
|
|||
|
|||
Does 10 TP seem slower to anyone? Running on VMware Player.
Big_Al wrote:
90% of that is over my head. And I don't have Virtual Box. VMware Player as stated in the OP. Get the benchmark http://web.archive.org/web/200710261...pi_mod-1.5.zip Try it inside the VM environment. Try it outside the VM environment. The VM environment runs at about 90% of the speed of native execution, which makes room for some instruction emulation calls. Most of the guest OS instructions, execute without interference. But some must be trapped, and emulated. So the guest cannot go at 100% speed. Doing a simple minded benchmark like SuperPI, is a simple way to assure yourself that things are working as expected. Only if the results weren't even close (the run time in the VM is ten times slower than the host), would you assume there is some problem. ******* If the OS running inside the virtual machine does not get the graphics hardware support it wants, it uses software rendering. This makes OSes like Ubuntu, dog slow. It's not the OS which is slow, it's the desktop GUI which takes forever to finish. When the OS issues a graphics call that is supposed to run in hardware, a lengthy software routine takes its place. This is why, you want a virtual machine environment with 3D support (directx2d/directx3d). As an example, VPC2007 or Windows Virtual PC hosting softwares don't have hardware graphics acceleration, so every OS with hardware acceleration needs, will behave poorly. I don't know whether VMWare has graphics support. VirtualBox does have it. VirtualBox graphics support is labeled as "experimental support", meaning they implemented enough of it to keep a few things happy. There are various levels of graphics support 1) Frame buffer emulation. This is the least amount of work you can do. VPC2007 does this. It emulates an S3 video card. The guest OS writes to the frame buffer, and the writes end up redirected to the host frame buffer in some way. No other graphics operations are supported. No programmable shaders. No game support. You can't play Crysis in the guest machine. 2) Partial support for an API. VirtualBox supports some portion of DirectX 2D and DirectX 3D calls. I don't know exactly how it does this. Maybe the odd game (like Crysis), could be run in the Guest OS. 3) Full support. Hyper-V (available in Win8, with a CPU that supports SLAT/EPT), should be able to pass more of the guest graphics calls directly to the hardware. I understand it's a bit slow. The best implementation I've run into, was how a graphics card was handled on the Macintosh. Back in the day, the Mac had a basic frame buffer card, and using VGA connectors for passthru, you could add a "gamer" card so 3D games would work. It was the fact there were two cards, allowing one card to be owned by the host, and the other card by the guest, that makes this possible (cleanly). CPU --- frame_buffer_card --- 3DFX_card --- computer_monitor VGA VGA One of the commercial virtual machine softwares there, gave the "gamer" card entirely to the guest. So the guest had full control, and the game image was an overlay (wrote over top of) the regular screen. So that idea violates the "insulation" aspects of virtual machines, but it does allow graphics acceleration to work properly. I think I played Quake II on that thing, at 1 FPS (frame per second). It was like a slide show. And that was the fault of the pitiful processor and the X86 to PowerPC instruction conversion. Your VMWare on the other hand, is X86 on X86, so runs at full speed (90% performance level). ******* If you find VMWare sucks, try VirtualBox (Oracle/Sun). Good luck, Paul |
#7
|
|||
|
|||
Does 10 TP seem slower to anyone? Running on VMware Player.
Big_Al wrote:
90% of that is over my head. And I don't have Virtual Box. VMware Player as stated in the OP. Get the benchmark http://web.archive.org/web/200710261...pi_mod-1.5.zip Try it inside the VM environment. Try it outside the VM environment. The VM environment runs at about 90% of the speed of native execution, which makes room for some instruction emulation calls. Most of the guest OS instructions, execute without interference. But some must be trapped, and emulated. So the guest cannot go at 100% speed. Doing a simple minded benchmark like SuperPI, is a simple way to assure yourself that things are working as expected. Only if the results weren't even close (the run time in the VM is ten times slower than the host), would you assume there is some problem. ******* If the OS running inside the virtual machine does not get the graphics hardware support it wants, it uses software rendering. This makes OSes like Ubuntu, dog slow. It's not the OS which is slow, it's the desktop GUI which takes forever to finish. When the OS issues a graphics call that is supposed to run in hardware, a lengthy software routine takes its place. This is why, you want a virtual machine environment with 3D support (directx2d/directx3d). As an example, VPC2007 or Windows Virtual PC hosting softwares don't have hardware graphics acceleration, so every OS with hardware acceleration needs, will behave poorly. I don't know whether VMWare has graphics support. VirtualBox does have it. VirtualBox graphics support is labeled as "experimental support", meaning they implemented enough of it to keep a few things happy. There are various levels of graphics support 1) Frame buffer emulation. This is the least amount of work you can do. VPC2007 does this. It emulates an S3 video card. The guest OS writes to the frame buffer, and the writes end up redirected to the host frame buffer in some way. No other graphics operations are supported. No programmable shaders. No game support. You can't play Crysis in the guest machine. 2) Partial support for an API. VirtualBox supports some portion of DirectX 2D and DirectX 3D calls. I don't know exactly how it does this. Maybe the odd game (like Crysis), could be run in the Guest OS. 3) Full support. Hyper-V (available in Win8, with a CPU that supports SLAT/EPT), should be able to pass more of the guest graphics calls directly to the hardware. I understand it's a bit slow. The best implementation I've run into, was how a graphics card was handled on the Macintosh. Back in the day, the Mac had a basic frame buffer card, and using VGA connectors for passthru, you could add a "gamer" card so 3D games would work. It was the fact there were two cards, allowing one card to be owned by the host, and the other card by the guest, that makes this possible (cleanly). CPU --- frame_buffer_card --- 3DFX_card --- computer_monitor VGA VGA One of the commercial virtual machine softwares there, gave the "gamer" card entirely to the guest. So the guest had full control, and the game image was an overlay (wrote over top of) the regular screen. So that idea violates the "insulation" aspects of virtual machines, but it does allow graphics acceleration to work properly. I think I played Quake II on that thing, at 1 FPS (frame per second). It was like a slide show. And that was the fault of the pitiful processor and the X86 to PowerPC instruction conversion. Your VMWare on the other hand, is X86 on X86, so runs at full speed (90% performance level). ******* If you find VMWare sucks, try VirtualBox (Oracle/Sun). Good luck, Paul |
#8
|
|||
|
|||
Does 10 TP seem slower to anyone? Running on VMware Player.
"Big_Al" wrote in message ... I have vmware player 7.0 and have a whole list of machines, XP,7,8.1,10TP,Mint17,Ubuntu. All seem to run pretty good, windows 7 is pretty responsive, but 10TP just seems to be sluggish as all get out. I've got all the machines set with the same 4G memory and 2 processors and video settings. The only variable I can come up with is the OS itself. I had VMP 6.0.4 installed on my Win7 HP x86 desktop, and installed Win10TP x64 in it. It was fairly slow to start, but once it was up, I found it to be almost as quick/responsive as my Win7 installation or Win8.1Pro x64 that I have on my laptop. The problem I ran into was with the memory limitation of x86. I originally assigned 1GB of RAM to the VM, and both OSs would run OK for a while, then VM would start to bog down. I upped the VM usage to 1.5GB and that helped some, but then after running both Win7 and Win10 for a while, Win7 would bog down. Even after shutting down the VM, Win7 was still sluggish until I rebooted. I think if I had a 64-bit OS as the host and more RAM, it would have run quite well. I ended up installing Win10 on a second HDD in my laptop, and selecting either Win10 or Win8.1 at boot. That's a much smoother scenario for all OSs involved :-) Overall, I like Win10, and am eager to see how it is in the final release. -- SC Tom |
#9
|
|||
|
|||
Does 10 TP seem slower to anyone? Running on VMware Player.
"Big_Al" wrote in message ... I have vmware player 7.0 and have a whole list of machines, XP,7,8.1,10TP,Mint17,Ubuntu. All seem to run pretty good, windows 7 is pretty responsive, but 10TP just seems to be sluggish as all get out. I've got all the machines set with the same 4G memory and 2 processors and video settings. The only variable I can come up with is the OS itself. I had VMP 6.0.4 installed on my Win7 HP x86 desktop, and installed Win10TP x64 in it. It was fairly slow to start, but once it was up, I found it to be almost as quick/responsive as my Win7 installation or Win8.1Pro x64 that I have on my laptop. The problem I ran into was with the memory limitation of x86. I originally assigned 1GB of RAM to the VM, and both OSs would run OK for a while, then VM would start to bog down. I upped the VM usage to 1.5GB and that helped some, but then after running both Win7 and Win10 for a while, Win7 would bog down. Even after shutting down the VM, Win7 was still sluggish until I rebooted. I think if I had a 64-bit OS as the host and more RAM, it would have run quite well. I ended up installing Win10 on a second HDD in my laptop, and selecting either Win10 or Win8.1 at boot. That's a much smoother scenario for all OSs involved :-) Overall, I like Win10, and am eager to see how it is in the final release. -- SC Tom |
Thread Tools | |
Display Modes | Rate This Thread |
|
|