Jump to content

modnar

Member
  • Posts

    295
  • Joined

  • Last visited

  • Donations

    0.00 USD 
  • Country

    Slovenia

Everything posted by modnar

  1. Thank you for your tests, @NotHereToPlayGames I guess I'll reset it back to 2MB. Imgur.com does seem to like 1MB better, less choppy picking through stuff.
  2. First thank you for new release, roytam1. Today I found a new setting to experiment with, javascript.options.main_thread_stack_quota_cap. Now I don't really know exactly what it does, it just seem good to me to halve it, so I set it to 1048576 instead. Would someone please tell me if this modification is useful? Browsing seems more smooth with less memory being spent on it.
  3. You might want to try the 475.14 (latest) driver with this github workaround, if you feel like it.
  4. @roytam1 Please make the next build of Serpent with threads / 2 or 1 (thread) for 1 core CPU's. I believe it will be faster than current n-1 (or 1) solution.
  5. Of course I compared the 4-25 build with 4-18 and others. Here's how it was: 4-18 and earlier builds were all slow then 4-25 was out and it was fast compared to before (but I had to revert to 4-18 because 4-25 kept freezing / not-responding). This new build 4-30 or 5-2, while faster than 4-18 and earlier builds, is slower than 4-25. On my CPU Serpent's thread count now returns 3 which I observed in VM's is a no-go. Slower than 2 or 4 threads. So better stick to even number of threads, but not all of them, so 2. At least for Core2 Quads that are really 2x2 CPU's (shared L2 between only 2 cores) and not 4x1 CPU.
  6. First thank you for a new build, roytam1. From using this new Serpent 52.9 build 4-30 (or 5-2) I observe slower performance and I suspect it's because of the threads setting. Imgur.com site though heavy is good for testing things. Also tomshardware.com, also heavy. For number of threads I would propose CPU cores / 2 or 1 if 1 core. I have a four-banger here, and while 4-25 Serpent stopped responding on imgur, while it was functional it was faster than version 4-30. I came up with half of threads because I observed with VMWare if I set the number of cores to odd (3) virtual machines perform worse as they do on even number of cores (2 or 4). 4 though also are not recommended for anything but Linux because there aren't any left for the host then (I suspect poor optimisation on Windows guests). Also - my Core2 Quad is not really 4 core CPU, it is 2x 2 core CPU.
  7. Has anyone here succeeded installing e.g. 3com EtherLink XL 10/100 PCI TX NIC (3C905B-TX) network card or similar 3com card in Windows "7"? I can't seem to find any driver for the NT_6-1. No other LAN card comes close to 3com's.
  8. XP SP2 version? Is older and doesn't have all functionality, so even worse. 2003 SP2 (kernel 5.2) isn't XP (kernel 5.1), so not applicable.
  9. I also downgraded to 4-18 release of Serpent 52.9. On imgur.com it just stopped responding after going through a few posts. BSOD that some mentioned could also be because of excessive fragmentation the browser causes with its cache on disk. I remember long time ago (2013) I had the same problem with fireferret.
  10. I only noticed MailNews downloading very slowly (20KB/s) in the morning, but in the afternoon (~3.5 h back) it went fast. Thank you for your fine work!
  11. @Dietmar Would you/could you make a better fltMgr for XP? The one that is added by SP3 (build 5512) is a p***poor W2k3 port and I believe deliberately held back (slows down with many minifilters installed).
  12. What if you used MailNews (I use MailNews, by the way)?
  13. An odd occurrence: yesterday evening I closed Supermium (ungoogled); a while later I ran it again and it wouldn't show the window (v144 R1).
  14. Thank you for a good explanation. I guess it's Microslop™ at work again turning even good things into, well, slop. Sigh.
  15. Thank you both for your answers; I wonder is there any other compiler that would be more flexible? Need you really use Microslop™ compiler?
  16. Thank you @basilisk-dev, @roytam1. Wonderful work! I wonder - now in "Troubleshooting Information"/"CPU capabilities" states SSE2 here - would the browser be any speedier if also SSE3 and SSSE3 were supported when found. I ask this because Intel CPUs by themselves aren't fast (FPU), compared to AMD if you do not use their special marketable instruction sets like MMX and especially SSE. To back my claims, here is HWInfo32's Benchmark results (taken with version 6.12 some time ago): AMD Phenom 9600 (quadcore 2.3 GHz) CPU: 107698 FPU: 139478 MMX: 83891, Xeon 3065 (or Core 2 Duo E6550 2.33 GHz) CPU: 68245 FPU: 43843 MMX: 80748 (MMX score as high as 4 core AMD).
  17. As I said it's a CD, so yes, the original. Not DVD. It is the right version, 0.9c. It used to be so that you had to have XPSP3 CD to create the Hiren's iso/CD. http://www.hiren.info Sorry, Astroskipper for the off topic.
  18. You can easily fix registry mistakes by booting into Hiren's Boot CD (15.2); in "Registry" there is Registry Editor PE. Works well.
  19. Just a heads-up: There are two security holes which are also being exploited that Google reportedly fixed a few days ago: CVE-2026-3909 & CVE-2026-3910. The new v144 probably hasn't have the fixes included...
  20. Thank you a gob and a bunch.
  21. List here. Also: it is not really advisable running Win2003 fltmgr on XP (different kernels) but it would be great if someone would port Win10 fltmgr.sys and fltmc and fltlib to XP. That one is really powerful.
  22. @MilkChan Would you also please also make a KB for 2003SP2 original fltMgr (5.2.3790.3959)? While it has a local vulnerability, it's the fastest for XP.
  23. I'm now testing two new changes that are in FireFerret_ESR 115.32.1 and not in Serpent 52.9: browser.cache.disk.max_chunks_memory_usage = 40960 (instead of 10240) browser.cache.disk.max_priority_chunks_memory_usage = 40960 (instead of 10240) It does help some, smoother, but at least on imgur.com there still seems to be memory leak happening or crazy scripts, I don't know...
  24. @roytam1 Wanting to get better performance with imgur.com site I followed inspiration and disabled devtools.performance.ui.enable-framerate. Please, tell me if this is advisable, because I am observing better browser (Serpent 52.9) performance (smoother scrolling and responsivness) on all "heavy" sites (and number of them increases). Upon further investigation I reset the upper variable and enabled memory.free_dirty_pages; it is not found even in Firefox 115.32.1 (Windblows 7), it's enabled by default. Update: Apparently I went out of bounds with these; restored to defaults: neither of these prove to be good for Serpent 52.9 (further slows down browsing).
  25. I did not see it happen here (with Serpent 52.9).
×
×
  • Create New...