Jump to content
InsanelyMac Forum

Dilemma Prison

Members
  • Content count

    72
  • Joined

  • Last visited

About Dilemma Prison

  • Rank
    InsanelyMac Protégé

Profile Information

  • Gender
    Male
  • Location
    Wisconsin
  1. I am kinda reluctant to post my config files mess until it isn't a partially working mess. I end up running afoul of too much of that around here. But for the sake of edification, Edit your hardware and issues into your post.
  2. Dilemma Prison

    New Driver for Realtek RTL8111

    More or less. But unless someone is doing SSD boot, auto-login and really quickly looks at their network status they won't even notice. Either way this is more zippy than the lnx2mac driver, at least on my x58 boards. Just a quick and dirty compare here with black magic speed test going to the same FreeNAS via AFP. 8111E (lnx2mac) W 91MB/s R 89MB/s 8111E (mieze) W 102.5MB/s R 110MB/s Intel 1000/PT Dual port (w/ single link up) 113MB/s R 110.5 MB/s Marvell Yukon 88E8053 W 109.4MB/s R 110.1MB/s
  3. Dilemma Prison

    Permanently Banned From Tonymacx86

    Yeeeeah. Agreed. Subject shift. So, according to public records Tonymac86's public tax stuff says they made ~ $180,000 last year. Given an presumed level of fudging on that I wonder how much they are rolling in yearly.
  4. Dilemma Prison

    New Driver for Realtek RTL8111

    Nice. Thanks for the hard work on this Mieze. Version 2.0.0d2 is working more or less perfectly with the 8111E on my x58a-ud3r so far. 110MB/s read write. Yosemite 10.10.3 Reliable thus far and doesn't hit the CPU as hard as some other drivers I've used for the chipset previously. Worked w/ injection via clover. I noticed that it is slightly slow to come up to a full link after startup and login. The interface is red for roughly 5-10 seconds after my desktop pops up. Could be some quirk of my setup but I figured I would mention it.
  5. To do an initial boot or install of 10.9 / 10.10 using clover optionally w/ out a DSDT to start with: Clover: -For testing install to USB flash i formatted MBR/FAT32. (Start with this; saves tons of time) -For install to HDD I did not have to pre-format the EFI as a msdos partition. Not sure why it persists as a recommendation in some places. Let the Clover installer handle things. Currently running Clover r3202. Installer settings I used are as follows : -Install Clover in the ESP -Bootloader : Install boot0af in MBR -CloverEFI : CloverEFI 64-bits SATA -Themes : BootCamp -Add HFSPlus.efi driver in drivers64 and drivers64efi/ in place of the vbox version. (not sure if it is needed in both) -Add kext HWSensors + FakeSMC 6.11.1363 to */EFI/Clover/kexts/10.9 and /10.10/ respectively. Obviously it works installed in S/L/E as well. -Add kext for 8111E Ethernet : I used Mieze's Realtek8111.kext version 2.0.0b2. Worked well installed in appropriate /kexts/10.x/ for injection. Go to http://www.insanelymac.com/forum/topic/287161-new-driver-for-realtek-rtl8111/ for instructions. Current version 2.0.0b2 was here : http://www.insanelymac.com/forum/topic/287161-new-driver-for-realtek-rtl8111/page-46?do=findComment&comment=2127590 I was getting 110/MB/s read/write w/ brief testing. Slow to pickup DHCP initially. Gets you a device w/ a MAC address for en0 and should allow for app store to work as long as the OS sees it as built in. It can be checked later using CLI ioreg or IOJones or similar; if I recall ethernet device should be GIGE if DSDT is properly patched or the lanfix2000 is enabled in clover. Depending on configuration it may be best to remove all other network devices prior to your first boot. This is what I did and may not be best-practices. A functional app store login seems to require a proper MAC address, evidently on a built-in connection. I cannot corroborate the claims that it has to be the MAC of an Apple ethernet or airport device for 10.10 stuff. It worked for me with my realtek as en0/built-in on both 10.9 and 10.10 app store. -Tried 2 different intel 1000/PT dual port worked as well but is freezing up (just the NIC) under load. I presume there is some fix involving swapping a kext or id remap, but didn't pursue. -My Marvell Yukon 88E8053 single port (both sonnet tech branded and no-brand cards from taiwan) run perfectly with no changes on 10.9.x and 10.10.x -I initially commented out nearly everything in the installed config.plist (appended # before plist entities) This initial settings combo got me up and running w/out my DSDT in place : Boot flags : -v and kext-dev-mode=1 to be safe. My cloned install didn't required it for some reason (perhaps pre-relaxed sandbox settings) (I have thus far not needed npci=0x2000 or dart=0. Keep seeing dart=0 brought up as a fix for things and insofar as I can tell it is for VT-D issues; possibly only for AMI bios. These boards are Award. The x58 chipsets technically support VT-d but according to Gigabyte the ud3r v1 doesn't have VT-d support in the bios and the v2 version of the board can only enable it using beta bios versions from some hidden bios panel.) Graphics: (same behavior w/ 10.9.2 thru 10.10.3) For GTX 770 disable nvidia inject. Leaving it on will result in the 0mb VRAM cosmetic issue and single display only operation. -Doesn't run at full performance oob. Core clock never hits full speed. (need AppleGraphicsPowerManagement patch or equivalent workaround). -w/ 10.9.2 display port wasn't running 4k resolutions above 30hz. For Radeon HD 6850/70 enable ati inject. -was able to run 3440x1440@30Hz via HDMI (in 10.10) which is beyond printed spec for the card. -issues w/ 10.9 where screen would boot black w/ only DVI. Display on miniDP or just a active adapter with no display would work around issue. SMBIOS: The initial automatic smbios setup generates a smbios for imac11,2 ; not a more appropriate 1366 Mac. The iMac SMBIOS didn't interfere with getting things booted initially. I use MacPro5,1 and regardless of which is chosen it is probably best to set the ProductName string under SMBIOS in config.plist before first boot. Doesn't seem to mess up startup either way but certain SMBIOS keys need to be present for iCloud and similar to work. In my case I set no other keys and I was able to re-login to the app store and iCloud keychains after migrating my existing user directory from a 10.9.5 Chameleon hack to a 10.10.3 Clover hack. Not sure if this side-steps some pitfalls with getting iCloud to work properly. Misc DSDT / Clover Fixes : USB settings : Inject, Add ClockID and FixOwnership all enabled. (not sure if these are all required; but I experienced boot issues during testing that may have necessitated 'Inject' being enabled) ACPI / DSDT settings / patches : (** denotes required for GA-x58 as best as I can tell) FixDisplay ; FixHPET ** ; AddDTGP ; FixHDA ; FixLAN ** ; FixAirport ; FixShutdown ; FixFirewire ** ; FixIPIC ; NewWay ; FixRegions ** ; FIX_RTC ** ; AddIMEI ; DeleteUnused ** ; FIX_ADP1 ; AddPNLF ; FIX_S3D ; FIX_ACST ; AddHDMI ; Apple RTC ** Checked SSDT Enable : GeneratePStates ; GenerateCStates ; EnableC6 Drop tables : SSDT string : Cpu0Ist SSDT string : CpuPM In my case leaving the fixes activated after putting a DSDT in place did not cause noticeable issues. My DSDT lacked the equiv FixEthernet / FixFirewire due to age All of this gets a vanilla mavericks and yosemite up and running enough to get things configured.
  6. Agreed. Even out of date info can be useful :/.
  7. Mega delayed reply here. My guess was that regardless of whatever modifications have been made to the sata kexts we are using (I forget how many kexts I had tried that claim to be recompiled or tweaked for 64bit use) none of them fix the use of what I presume is a 32bit DMA memory map or something along those lines. One would think that it shouldn't have an effect on file transfers between two drives not connected to the nforce sata bus, but it does. Something related to the way the root volume gets used, I suppose. I never did test it but the issue may resonate in a fashion similar to 32bit memory issues ala windows XP; with the VRAM of the graphics card being part of the 4gig max for memory that can be mapped (which could be why we see so many varying maxmem totals. My system was running a 512mb card). I frankly have not kept up on how moving to 64bit architectures effects things. Someone with the appropriate programing skills and some familiarity with authoring iokit drivers can probably fix the issues (Not in my immediate skillset, sadly.) This problem seems to afflict a whole bunch of ata and sata kexts that were probably coded originally for use with 10.4/10.5. I've seen threads here and elsewhere where folks have updated a few of them; none for nforceata though. Digging up info on all of this was a royal pain when I did it; all of the kexts floating around are poorly labeled and a number of the folks who rebundle other people's stuff have renamed/relabled kexts at various times much to my annoyance. As to setting the flags to limit CPU cores (cpu=x, etc) I can't speak as to why I had no negative impact running both my cores on my particular software combo (My 10.6 and 10.7 installs have been as bare-bones as possible). Someone needs to spit out a updated kext
  8. Thats going to be hard (or require some annoying hardware swapping); I have a 930w/1333ram in the v1 and a 950w/2000 in the v2, along with a lot of other different hardware. I doubt the changes in the v2 have a substantial effect on raw performance. I suppose I shall post about any variances I encounter when I finish pulling the 10.7 together for the v2. Maybe the realtek nic in the v2 won't run like {censored}. I had sluggish performance with the nic on the v1. While it was under heavy file transfer loads most of my other network communication would stall out. I assumed it was the result of questionable drivers. I put a sonnet pcie card in to avoid using the onboard. (AFP capped @ 35-40MB/ss throughput w/ realtek Vs 95-100+MB/s with the marvel chipset). We shall see ;P
  9. As of this morning I now own a X58A-UD3R v2.0 in addition to the v1.0 board I have in an existing hackintosh; I'll be building a dsdt and pulling my software bits and pieces together sometime this week to get this second system up and running. The question; is anyone following this thread interested in me posting a diff comparison of my dsdts of both boards, virgin and patched etc, or would I be repeating info already posted elsewhere.
  10. Could this be that old issue of some stock apple kext loading and interfering with devices that have the nvidia vendor ID? ; wasn't there scenarios where you zapped some of the default kexts to prevent that on some mobos?
  11. Just voted Well; I ran about 2tb of data through various drives on the machine; at this point I'll consider it stable; or at least not experiencing that particular issue. I'm going to keep pushing maxmem up to see what the exact memory limit is; see if I can figure out what exactly isn't working properly.
  12. Ah, the much complained about i680 memory issue. A lot of users ran into issues w/ just running 4 chips, over clocked or otherwise. Some resolved problems w/ bumping voltages, but most of what I read involved doing so in tandem w/ overclocking. A lot of people were blowing their ram too. I seem to recall there being lots of nebulous problems with memory combos, oft discussed on the overclockers boards. I'm running only a pair of 2gb 800mhz chips in this system at present, no overclock on the ram. (1.9v i think ) At previous times I had 1/1-2gb 1,1,2,2-6gb and 2,2,2,2-8gb of ram in the thing; 8 gigs back when I was using 10.6 server and the large FIS port replicated hard drive arrays. Ah well; I set the cpu/maxmem values down and in the process of beating the {censored} out of the storage at the moment. *EDIT* Well; it's definitely an issue with either the cpu/mem values. cpus=1 maxmem=2048 results in kernel_task consuming a steady 3.5% of one core. This is happening when duping files on drives attached to my add-on card and when duping files on the nforce sata. Looks like it's possibly some more nebulous nforce related issue: apple's stuff not liking chipset or config or ahci implimentation or something along those lines. I'm going to start bumping settings upward until the problem comes back. *EDIT2* Using 2 cores is not triggering the massive CPU usage. It must be related to the memory limit. maxmem=3456 w/out the cpu flag still shows the expected 3-4% overhead for disk to disk file transfers. I guess I will leave it there and thrash/reliability test my disks now. Digging through the vague {censored} on other forums I find instances of other folks dropping maxmem for stability reasons (but nobody actually mentioning anything about their CPU utilization). Some of these are on AMD nforce systems, some Intel nforce systems and others just on core2 intel chipsets. For my own system, when I have my CPU maxed by my sata + ethernet overhead (redlining in activity monitor) for an extended period of time I seem to eventually crash (w 4gb+ ram). Bonus side effect : transfer rate for AFP is pretty much holding around 95mb/sec now instead of spiking around, no longer CPU limited due to the previous issue.
  13. Thanks for the reply verdant. That explains the consistent behavior. Previously; I had 8 500gig drives attached to this system w/ port replicator case + card and never encountered instabilities doing bulk file transfers (and that case/card still works w/ 10.7 in a stable fashion). However, I also was using the nforce ethernet and didn't typically see file transfer rates above 30-40 mb/sec. I installed a marvel pro nic recently and am finding that sustaining throughput up in the 90-95 mb/sec range seems to crash the machine. I thought it might be the nic but doing file transfers back and forth between my newer, faster sata drives seems to also cause crashes. In the interest of further testing last night I put a highpoint card w/ the same disks I tested on built in sata on my system and did a bunch of bulk file transfers between the volumes. (it was my hope I could avoid the issue by using a different sata chipset. The results were not promising. The CPU utilization was relatively high; maybe 2/3rds of what it was when the drives were on the nforce sata. On my i7 system this card generates only 2-3% utilization on one core with the same transaction on the same drives. This is hitting a wolf dale core for 50%? Something isn't right here. I am wondering if this is being caused by some other issue. I suppose I'll do all these tests over w/ the chameleon flags and see what happens. I might throw a usb stick on there and see if I can get the system to fire up w/ the os attached to the raid card. .and not load the sata driver at all. maxmem=2048 simply hides anything beyond the first 2gig? Would slapping a pair of 1gig sticks in provide the same effect?
  14. Howdy! Just got around to upgrading my i680 SLI system up to 10.7. I ended up encountering the [PCI configuration Begin] error and using the npci=0x2000 as indicated in your first solution fixed the issue. No other issues; Thanks for your massive documentation / troubleshooting / information herding efforts verdant! The guide is fantastic. Just tossing a question/observation out about the nature of the AppleNForceATA driver for the nforce sata. When doing file transfers I get some fairly substantial cpu utilization (the process kernel_task). When moving data at 80-90 mb/ sec I can expect kernel_task to eat up between 40-70% of one core, which seems to be far more substantial than other machines running the standard driver. I never payed much attention to the system utilization and noticed that my 10.6.8 install does the exact same thing with the 32 bit driver (anv version, I believe) Anyone else get high system usage w/ sata io? I don't know if I am kicking a dead horse documented elsewhere with the question or It's just my hardware/software combo doing this.
  15. Dilemma Prison

    PCI Express 16x Card in 1x Slot

    A distinction needs to be pointed out here. The Mac pro has slots of a variety of speeds but those slots all still have the 16x electrical socket. Most motherboards have slots sized for the pcie interconnect they are capable of. 1x for 1x, 4x for 4x, 8x for 8x, 16x for 16x, etc. When this is not the case it tends to be the 8x (signal) slots with 16x electrical sockets; a motherboard with 1x sockets will not have full 16x length plugs on them. Socket converters exist that allow a 16x/8x/4x card to be placed in a 1x slot. Depending on the application though the performance may be terrible and the adapter w/ a graphics card may not fit in a standard case. http://www.startech.com/PCI-Express-x1-to-...dapter~PEX1TO16 There are extended length wire-ribbon versions of this sort of thing as well. . . this stuff is cheap enough that I can't imagine why someone would want to cut sockets or pins off of their graphics cards / mobos. . .doesn't take much of a miss-aligned insertion to zap for hardware.
×