Understanding the X58 chipset Asus P6T BIOS Settings for the i7 Nehalem CPU

Very very interesting thread guys, it will help me a lot in my effort to OC i7 even if on a different Mobo!

EDIT: Clarification: My RAM is 3 x 1GB G.SKIL DDR3 1333 (PC3 10666)

I can keep it close to its default [DDR3-1269MHz] and choose tighter timings: 8-8-8-20
or
I can increase it close to DDR3 1600 (PC3 12800) speeds: [DDR3-1691MHz] but only if I increase timings: 10-10-10-24





...and now, memory tests. As long the difference between DRAM Bus Voltage and QPI/DRAM Uncore Voltage is less than 0.5 volts, we can go over the 1.65 volt limit labels warn us about.

In order to pass stability tests for my 3 x 1GB G.SKIL DDR3 1333 (PC3 10666) at [DDR3-1691MHz] I had to relax timings all the way down to 10-10-10-24.


Even for 9-9-9-24, the system would not boot unless I went all the way up to 1.84 DRAM Bus Voltage, the limit of safety. And by the way for 9-9-9-24 at 1.84 volts, here's what the score was:


_____________________
Intel i7 920 [211] BCLK x 19 = 4.0 GHz @ [1.4000] CPU Voltage & [1.35000] QPI/DRAM Uncore Voltage, Batch 3836A394
3 x 1GB G.SKIL DDR3 1333 (PC3 10666) [DDR3-1691MHz] 10-10-10-24 @ 1.64 DRAM Bus Voltage
ASUS P6T Deluxe v.1 [LGA 1366 Intel X58] BIOS 1102
Thermalright Ultra-120 eXtreme 1366 RT with 120mm Scythe S-Flex F fan
ASUS EAH4850 TOP Radeon HD 4850 512MB @ 680 MHz GPU & 2100 MHz Memory
Antec nine hundred case, two front 120mm fans, one back 120mm Fan, one top 200mm fan
Corsair CMPSU-750TX 750W

So even if CAS 9 was stable, the disproportionate amount of voltage needed and seeing as it would be at its maximum, was not justified.


Dropping the voltage to 1.64 (or maybe even lower if possible) & going with 10-10-10-24 is what I was left with but it raised an interesting question:

3 x 1GB G.SKIL DDR3 1333 (PC3 10666)
[DDR3-1269MHz] @ 8-8-8-20
or
[DDR3-1691MHz] @10-10-10-24



To find out, first SuperPi comparison:

And then x264 HD Benchmark:

Results for x264.exe v0.58.747

[DDR3-1269MHz] @ 8-8-8-20:

encoded 1442 frames, 94.95 fps, 3902.77 kb/s
encoded 1442 frames, 99.67 fps, 3902.77 kb/s
encoded 1442 frames, 101.31 fps, 3902.77 kb/s
encoded 1442 frames, 100.20 fps, 3899.40 kb/s
encoded 1442 frames, 34.00 fps, 3974.41 kb/s
encoded 1442 frames, 34.02 fps, 3974.41 kb/s
encoded 1442 frames, 34.09 fps, 3974.41 kb/s
encoded 1442 frames, 33.93 fps, 3975.15 kb/s


vs.

[DDR3-1691MHz] @10-10-10-24:

encoded 1442 frames, 101.64 fps, 3902.77 kb/s
encoded 1442 frames, 102.31 fps, 3902.77 kb/s
encoded 1442 frames, 101.64 fps, 3899.40 kb/s
encoded 1442 frames, 99.99 fps, 3902.77 kb/s
encoded 1442 frames, 34.11 fps, 3974.41 kb/s
encoded 1442 frames, 34.14 fps, 3974.41 kb/s
encoded 1442 frames, 34.00 fps, 3975.15 kb/s
encoded 1442 frames, 34.06 fps, 3974.41 kb/s



* * *



Results for x264.exe v0.59.819M

[DDR3-1269MHz] @ 8-8-8-20:

encoded 1442 frames, 105.23 fps, 3895.56 kb/s
encoded 1442 frames, 105.23 fps, 3894.90 kb/s
encoded 1442 frames, 105.23 fps, 3894.38 kb/s
encoded 1442 frames, 106.56 fps, 3893.72 kb/s
encoded 1442 frames, 38.04 fps, 3981.22 kb/s
encoded 1442 frames, 38.06 fps, 3981.57 kb/s
encoded 1442 frames, 38.06 fps, 3981.67 kb/s
encoded 1442 frames, 38.04 fps, 3980.99 kb/s


vs.

[DDR3-1691MHz] @10-10-10-24:
encoded 1442 frames, 107.07 fps, 3895.10 kb/s
encoded 1442 frames, 106.69 fps, 3895.56 kb/s
encoded 1442 frames, 105.59 fps, 3895.56 kb/s
encoded 1442 frames, 105.83 fps, 3894.50 kb/s
encoded 1442 frames, 38.06 fps, 3981.50 kb/s
encoded 1442 frames, 38.12 fps, 3981.22 kb/s
encoded 1442 frames, 38.09 fps, 3981.22 kb/s
encoded 1442 frames, 38.15 fps, 3981.58 kb/s

Finally Everest:

[DDR3-1269MHz] @ 8-8-8-20 vs. [DDR3-1691MHz] @ 10-10-10-24:

Memory Read: 15104 MB/s vs. 16591
Memory Write: 14965 MB/s vs. 14975
Memory Copy: 17363 MB/s vs. 18171 MB/s
Memory Latency: 52.0 ns vs. 50.8 ns
CPU Queen: 36897 vs. 36895
CPU PhotoWorxx: 40836 vs. 42630
CPU ZLib: 135108 KB/s vs. 134731 KB/s
CPU AES: 31953 vs. 32010
FPU Julia: 17243 vs. 17248
FPU Mandel: 9386 vs. 9386
FPU SinJulia: 7493 vs. 7494


_____________________
Intel i7 920 [211] BCLK x 19 = 4.0 GHz @ [1.4000] CPU Voltage & [1.35000] QPI/DRAM Uncore Voltage, Batch 3836A394
3 x 1GB G.SKIL DDR3 1333 (PC3 10666) [DDR3-1691MHz] 10-10-10-24 @ 1.64 DRAM Bus Voltage
ASUS P6T Deluxe v.1 [LGA 1366 Intel X58] BIOS 1102
Thermalright Ultra-120 eXtreme 1366 RT with 120mm Scythe S-Flex F fan
ASUS EAH4850 TOP Radeon HD 4850 512MB @ 680 MHz GPU & 2100 MHz Memory
Antec nine hundred case, two front 120mm fans, one back 120mm Fan, one top 200mm fan
Corsair CMPSU-750TX 750W

Excellent data here, thanks a lot

So now here I am, the CPU is about 25% faster than the fastest CPU "money can buy" today as things go... I'm in a hurry and here's what I was doing and here's how things went: I'm playing music, I'm rendering video files to make a DVD and I have browsers open and I'm trying to open another video file at the same time to take snap shots of it for the DVD menu thumbnail, and all of that at the same time...


What a perfect test. While doing all of this at once is unimaginable on my old Athlon XP, the 4 GHz i7 is struggling. It's doing it but I'm exhausting resources somewhere along the way.

So what would make doing all of this snappier?

Switching to a 64-bit OS? Doubling the RAM from 3GB to 6GB? Raw CPU speed? I'm not talking improvements, I'm talking the hypothetical situation of no stuttering whatsoever while doing all of this simultaneously... How would things change with 64-bit OS vs. more RAM vs. more raw CPU speed vs. maybe HD switch from 7200 RPM to Solid State...



_____________________
Intel i7 920 [211] BCLK x 19 = 4.0 GHz @ [1.4000] CPU Voltage & [1.35000] QPI/DRAM Uncore Voltage, Batch 3836A394
3 x 1GB G.SKIL DDR3 1333 (PC3 10666) [DDR3-1691MHz] 10-10-10-24 @ 1.64 DRAM Bus Voltage
ASUS P6T Deluxe v.1 [LGA 1366 Intel X58] BIOS 1102
Thermalright Ultra-120 eXtreme 1366 RT with 120mm Scythe S-Flex F fan
ASUS EAH4850 TOP Radeon HD 4850 512MB @ 680 MHz GPU & 2100 MHz Memory
Antec nine hundred case, two front 120mm fans, one back 120mm Fan, one top 200mm fan
Corsair CMPSU-750TX 750W

sounds like the HD not able to keep up with that kind of demand. that is when i would look into a 4-6 raid-0 setup with a card that has 128-256mb cache.

Hm. I was working between two physical hard drives... so you think the HD is the cause here? I hate to go RAID and increase the risk of loosing it all if one fails. (I had a HD go belly up last month.)

Well... would solid state drive help significantly, what did we say about them when we discussed them, when are the ones w/o delays and probs coming out?

If I bought just one SSD, would doing this work on one improve things vs. doing this between two regular ol' HDs?

Well first off how much ram was in use when you where doing that? If its not hitting say 2.5Ghz your probably still safe with 3Gigs. But as mentioned, I could defiantly see that the HDD holding you back if your hitting it hard especially with videos. How is the CPU usage when this happened as well?

If leary about raid you could do a Raid 0 and have a Raid 1 for backups or if you want to go full blown, a Raid 0+1 (4 drives), where you'll get the benefits of speed, yet have redundancy automatically built in, course its more expensive option.

Otherwise that i7 should be kicking butt and taking names especially in multi threaded apps.

well with the kind of multi tasking he was talking about most single SATA drives depending on the size. will have 2 - 4 heads for reading/writing compared to scsi they will have twice that amount or there abouts. which is why they are better in that reguard but with a higher cost. i only suggested SSD since no moving parts and with a raid setup will get mad thru-put with .1-.4ms access time. like death was saying, but i think its raid 10 that does what he is talking about. though im prolly wrong its been a while since i looked into the raid levels.

the quick here is though you asked one guy to move 1000lbs. of course its going to take him a while, if you had say 2 things would be easier but for this job you really need 4 guys. not really a direct relation to how many drives to get but more in the work needing to be done by the HD.

What do you guys use to monitor as many things as possible to get an idea about what's maxing out?

HDTach to understand what the threw put of the HDD is at least, then I can use Vista's resource monitor to find out how much HDD activity there is.

Other then that task manager for CPU/Ram Resources being used.

Guess an easy indicator is if the program is using 100% of a said core or cores, and when you do other things dedicate them to there own core or cores and the CPU usage falls off then there is a bottleneck probably in either the ram (ran out of memory), if not then its the HDD speeds.

As for the raid, it might be Raid 10, or Raid 0+1 which ever its called, 4 HDs', which really act like 2 in Raid 0 but redundant.

I do have an XP/Vista dual boot but have finished my extensive Vista testing with a conclusion that it is not suitable for people who like to do things a fraction of a second faster if they can, so I am again using Windows XP. After all, that's what this topic is all about.

So, this HD Tach you say will allow me to isolate if HD is the true bottleneck? I start it up, then assault the system with several multi-tasks and we can interpret the graph to see what's up?


All right, so help me set up the rest of it in Windows XP, what do we need, a RAM monitor and four core CPU monitor, what else and what do you recommend to use?

HD Tach will give give you an idea of how many Megs/sec at least on read (will be less on writes) your drive will give across the disk. Its a benchmark.

As for Ram and CPU Monitor, the task manager is all you need nothing else. Think there is some better programs out there but not sure what to even recommend.

To read your activity on the HDD of how much information is going where, Vista Resource managaer is the only thing that I know that does this, so it can't be done from within XP that I know of but again its a ball park figure, but if your hitting your high points from say HD Tach then its a HDD issue.

If you want might want to give Win7 a shot, little better than Vista in the performance department. Or turn off superfetch in Vista and see if that speeds up your experience.

Interrupts, bad drivers, PCI bus grant times, DRM, Drive IO?

c627627 said:

So now here I am, the CPU is about 25% faster than the fastest CPU "money can buy" today as things go... I'm in a hurry and here's what I was doing and here's how things went: I'm playing music, I'm rendering video files to make a DVD and I have browsers open and I'm trying to open another video file at the same time to take snap shots of it for the DVD menu thumbnail, and all of that at the same time...

the 4 GHz i7 is struggling. It's doing it but I'm exhausting resources somewhere along the way.

So what would make doing all of this snappier?

Switching to a 64-bit OS? Doubling the RAM from 3GB to 6GB? Raw CPU speed? I'm not talking improvements, I'm talking the hypothetical situation of no stuttering whatsoever while doing all of this simultaneously... How would things change with 64-bit OS vs. more RAM vs. more raw CPU speed vs. maybe HD switch from 7200 RPM to Solid State...

Click to expand...

There are many things to consider, you might have something pathological going on, creating your stutter, however keep some of these things in mind because none of your suggestions alleviate the things I will discuss, and in fact under exotic circumstances make things worse.


1> CPU-wide Interrupts might be turned off for more than 15 milliseconds (the universally accepted limit for the last 20 years in programming). Programmers turn off interrupts for a variety of lazy design reasons, but commonly do so to service an I/O interrupt and keep it help too long, or to stop people from possibly debugging their software security code protecting DRM of encrypted code or encrypted DATA. On Macs and on Windows, anything to do with a real DVD video has code that by contract law HAS to be protected from Debuggers. In fact, amusingly Nero full suite wont load with some debugger tools loaded, and even Motorola+Apples OS debugger ( MacsBug ) merely being loaded will make the DVD player of the OS refuse to play. Other solutions employed by other DVD players, allow SoftICE to be loaded, or other tools, but do skanky games to protect the key exchange of BluRay or DVD. You also might have too many interrupts in general. You can monitor the load with a variety of tools. Microsoft has web pages discussing some of their tools ot read the stored statistics they maintain. Anything past 2000 interrupts per second is beyond pathologically sick on a modern machine unless for some reason you are performing 2000 disk i/os per second using 10 controllers, not one, and have the driver for each controller set to COALESCE 32 or 16 I/O completions per interrupt to reduce load. Coalesce ramp up time waits up to 5 milliseconds maybe 15 , before giving up on the pending delayed interrupt, and generally uses a load counter and avoids dead ramp measuring before going into SEMI-POLLED high load mode. You only need your sound buffer to hold 20 milliseconds of sound. But many have 4 Kilobytes or 8 Kilobytes of buffer in the hardware to reduce echo delay in live performance. 8 Kilobyte at 16 bit sample per channel times 2 channels * 44.1 Kilohertz = 176,400 bytes per second = 4410 bytes needed if interrupts turned off for 25 milliseconds, thus the historic 15 millisecond rule, allowing a 4096 byte hardward buffer to not underrun, with 44.1Khz 16bit 2 channel load of 2646 bytes. Sickeningly, Apple reduced HW buffer to 2048 briefly a decade ago and I had to counteract it in special raid drivers. More sickeningly, Apple disabled interrupt MORE THAN 25 milliseconds in one system release update briefly for one particular model that also had a MIDI driver and MIDI support and took out MIDI on occasion. They killed their own crap. But I am digressing. I smell interrupts being disabled for too many milliseconds by SOMETHING in your system if you have audio stutter. Or your have a APPLICATION creating audio and feeding audio without using a privileged driver. Apple in OS 8.1 playing QUICKTIME general midi files comes to a screeching stop if applications get starved of time by a user merely holding down mouse button forever in a menu. Basically they needed to make the entire chain of audio interrupt driver via timer fed drivers. 90MHz Pentiums stutterred on non interrupt fed mp3 players at 160kbps mp3, but the same 90MHz Pentiums could play two streams at a time with interrupt fed players of the time, immune from stutter starvation. But a top end modern system doing a lot of DVD ripping using "MS Direct Show" intercepted vid driver tricks and WinDVD or PowerDVD rather than a hacker tool, will indeed stutter from DRM turning off interrupts too often, in fact playing a mp3 using crappy WinAmp 4 would stutter in my experience, while ripping using WinDVD. it had nothing to do with disk I/O, or CPU, or RAM, it was all interrupts masked crud by the DRM of the DVD player. ANOTHER REASON TO ONLY RUN HACKER TOOLS TO PLAY DVDs.


2> Installing VISTA will only slow you down for a wide variety of latency bennchmarks involving opening and closing files, creating new tiny files, deleting files, expanding files and writing to endpoints, launching apps and quitting apps rapidly, playing a 11 millisecond 2-frame (one silent frame), mp3 blip file from disk in a loop, and other benchmarks that time how long to start and stop almost any small task you want to measure. Installing 64 bit pure OS ( whether Mac OS X kernel user-compiled to pure 64 bit, or XP 64 ) will show you the sad sad sad result of slow code because many things have 32 bit integers straddling 32 bit boundaries, and use very very little 64 bit math, and the os, and even benchmarks, have little benefit from a ram window larger than 4 gigabytes per task, or 2 gigabytes. Windows 7 MIGHT help, only in that it might treat your 8 threads on your 4 cpus with proper respect. But that will only help exotic massively parallel benchmarks.

3> More than 3GB RAM ?
Nahh. Personally I have 8 gigabytes min in all my machines, and 16 GB in one, but I am a fool with too much money to blow. No discussion of warez, SEE BELOW - IMOG, I boot into Window XP corp ed. exclusively. No VISTA anymore ever. No 64 bit and its 3rd party bugs ever. I stick with whats fast and popular. Ironically you lose 1.5 gigs of IDLE ram on a semi busy machine just by switching to VISTA, so you really only get 1.5 of the 3 you want to add.
If you have a need to enjoy 16 gigabytes of RAM, Apple has been shipping such machines for 4 or more years, because their BSD Unix OS exploited with 8 ram slot, tweaked to run on eight PowerG5 64bit cores, or eight XEON intel recent years, and triple slot pci-X 133Mhz (no pc ever offered 3 equal speed gigabyte per second cards, even afterward, until boards with THREE pciExpress 2.0 16x lanes started shipping in 2008, many many years later). But many mac people are equally happy with 8 or 4 or even gasp! 2 gigabytes. The reason is that virtual RAM works well if your disk drives are reasonably fast and you do not tax the system (hidef video edit). In fact Apple STILL in 2009, after many disappointing OVERALL benchmarks, refuses to ship their OS configured for 100% 64 bit computing, though you can recompile it form their open sources that way. Apple uses all the RAM, but uses PAGE REMAPPING at high speed to juggle data between 4 gigabyte windows. A single app can launch with more than a 4 Gigabyte window.
In 10.6 (soon) Apple will enable 64 bit all the time system wide and slow everyone down a few percent. Third party drivers and cards have been 64 bit clean for over 4 years so it will cause no hiccups. Just sadder benchmarks.

4> Disk I/O SATA and not IDE on boot drive, AHCI
Running SATA and not IDE on boot drive and also running in AHCI BIOS mode should help a lot and allow proper scatter gather and head seek elevator algorithms and multiple pending I/Os. As per wikipedia and 5 years experience : Enabling AHCI in a system's BIOS will cause a 0x7B Blue Screen of Death STOP error (INACCESSIBLE_BOOT_DEVICE) on installations of Windows XP and Windows Vista where AHCI/RAID drivers for that system's chipset are not installed; i.e., boot failure. And also make all repair boot disks fail to load, including all lovingly created minimal XP repair USB dongles and Cds. You NEED to install AHCI support on boot cd tools and in main OS to use the fast and clean ACHI mode for your sata boot drive.

4> Disk I/O Vibration
A case that VIBRATES ultrasonically will slow down head seeks over 40%. Use acoustic ear foam for gunnery range wedged in both sides if a drive is insert rail mounted. That always restores the missing 40% I/O/sec on Raptor drives for me. The ultrasonic lateral vibration from other head seaks causes a landing miss greater than 1 track, and the drive goes into a different safety head seek mode to increase head seek guess accuracy. You will not read about this secret anywhere because I never mention it. A head VP of Gigabyte showed me the trick and it worked.

4> Disk I/O tech ?
SCSI would help but thats only because the SOFTWARE drivers are not based on hacks by monkeys and inept coders. The hardware also tends to not use the inner hub and outer platter area for faster seeks and faster I/O all the time on average, but more money. SCSI has been 15,000 RPM from 7 vendors for over 5 years. No SATA is ever 15,000 RPM, and tops out usually at 10,000 RPM Raptor). SCSI sometimes also uses more platters (more heads) for speed. Also one time SCSI had TWO HEADS PER DRIVE PLATTER !!!! TWO !!! The Seagate ST-12450W double head drive. (I owned 6 of these 18 head drives 14 years ago : www.seagate.com/support/disc/manuals/scsi/8900b.pdf). Seagate stopped making double head per surface drives because slight shaking of them in shipping ruins them. Incredibly fragile.

Your idea to try solid state is a bad one. SCSI 320 eats solid state drives for breakfast. Solid state RANDOM WRITE TEST using a list of 500,000 RANDOM 512 byte sectors, pending as many as desired (typically 800 at a time) is TEN TIMES FASTER THAN SOLID STATE DRIVES.

TEN TIMES SLOWER in 2009 for tiny massive amounts of random writes using solid state (read speed is not as affected).

Ironically it should be FASTER. NO heads to move for solid state. The truth is stomach churning and revolting and would just make you puke. Basically SOLID STATE DRIVES are a tree of replicated pocket ram drive dongles with 2048 or 1024 identical FLASH RAM driver parts. Its the same slow slow slow part for writes replicated 1024 or 2048 times in silicon. Its still barely 2 megabytes per second in 2009 for the one 512 byte sector to complete.
RAID may make things worse for you, but good raid-0 will probably help.

5> Integerated Wifi or PCI WiFI card ?
Drooling borderline retards write the driver code for these and not only waste CPU, they disable interrupts a lot.
Belkin 802.11g PCI card consumes 40% of CPU on some setups (in periodic bursts) Hardware Disable (BIOS or pull card, or other driver hibernate trick) drops Belkin 802.11g PCI card to 0% yielding back 40% wasted CPU. I am not claiming ALL people have this issue, but SOME do.

5> PAGE FAULTS?
IF YOU SEE a need for more than 3 GB, than more ram would help IMMENSLEY! You just need to monitor your true page swap load.

In summary, I feel bad for you, because you have the one machine everyone here wishes they had, even me, and you are discovering that poorly written crap can make other poorly written crap have audio stutters when playing music.

If it means that much to you, an old (ancient) version of WinAmp will play mp3 using interrupt feeds and will NOT STUTTER on your machine. I forget the version. Certainly 2.x anything. And maybe a few versions of 3.x

Or maybe you can config your PCI audio card to use larger low water mark in hardware buffer. It might have 32K of lag buffer in hardware but set to use 4K for low low echo delay. Or ensure audio card is NOT getting 7 channels all the time instead of 2 you need, or is 16 bit per sample not 24 bit, or just get a PCI-Express audio card. Or use no audio card and use a USB 15 dollar D/A headphone/lineout adaptor as a side test. (a test avoiding your sound card). Or reconsider EVER adding ANY PCI to this machine at all and sticking with pure pciExpress (PCIe).

If you audio stutters its a SYMPTOM of other problems, but sometimes the best advice is to bury the symptom and make it disappear.

Dont forget to look out for other monitoring tools that show all the lower level statistics of the OS. MS has a few I believe that work in XP.

Now that you achieved this miracle of OC, you should spend a day and see if you can get Mac OS X 10.5.6 humming perfectly on it. It’s a huge community, and they will help if a particular integrated device needs a driver tweak.

Try it, its still possible to run Windows XP under 10.5.6 OS X in three different virtual machines, one of them is free (virtualbox.org from Sun) the other two are OSX VMWare and Parallels. Parallels is better for GPU gaming, but normally people reboot into Windows XP if doing a lot of solid gaming.

Two things, what an awesome post and welcome to the forums.

Although others may engage you in a debate about the detailed technical aspects, you drive an important point home which is that simply increasing the speed of one piece of hardware may not be the overall solution as many of us assumed.



2.

driverguru said:

No VISTA anymore ever. No 64 bit and its 3rd party bugs ever.

Click to expand...

For every day use, I've reached the same conclusion after using Vista on my dual boot but could you post a list of reasons that made you personally say no Vista any more.

3. You say idle RAM is lost in Vista but did they not say that this is a common misconception about Vista and that RAM loss was not what it seemed as the RAM is immediately released as soon as it is needed in Vista?


4. Yup Blue Screen of Death with AHCI instead of IDE enabled. When you say "install AHCI support," I'll google it but is the procedure mobo specific?


5. Re the part about "slow down head seeks over 40%" first off, I don't have raptors just regular 7200 RPMs will that help me? Second, what!? up to 40% increase just by inserting what where? Pics please or links but this is major, why is this not mentioned everywhere? So I'll be definitely following up on that jewel:
1. Tell me what to use to measure this.
2. Help me understand what exactly I am supposed to do to set up this experiment & I'll post pics & results.


6. Thanks for SCSI info, it makes sense, let me ask you: what do you see as far as the Solid Sate Drive roadmaps go, what do you see in the near future for storage size & speed?


7. Re WIFI cards, funny post. Stay away from PCI WiFI then?


8. Please post on method of monitoring true page swap load and re

driverguru said:

You can monitor the load with a variety of tools.

Click to expand...

I would love to do this, I can even replicate the assault on the same machine in Vista since I have a dual boot XP/Vista but what all should I use to set up this experiment and measure it all just for info purposes?


9. Re audio-only stutters, unless I posted up there that I noticed them, I'm trying to think if I did, but I'll definitely pay attention to see if I hear them - then implement your suggestions to address them. I am also on a lookout for a good deal on a PCIe sound card since I can't have *any* PCI cards in my machine because ATI 4850 video cards cannot run stable under extreme loads if there are any PCI cards installed since they block air flow to the ATI 4850 which runs hot as designed.


10. Yeah sure I can get Mac OS X installed on my system, I was just looking at how someone was doing it the other month, the only reason I did not look into details is because I could not think of a reason as to why I would want to for personal use... if I can read or hear a useful reason to do so personally, I'd give it a shot. In other words: Mac... OS X... what is it good for?



P.S. Just to clear things up to others reading this, I believe Windows XP Corp is identical to Windows XP Pro apart from the activation mechanism.


P.P.S. I got one for you: At the Windows XP boot screen when the blue bar is running under the logo, I hear *click* then *click* again then it continues... at every boot but only on XP when the logo shows up and not after that at all... any idea on what's that all about?

reply to replies of my post on possible slowdowns

c627627 said:

3. You say idle RAM is lost in Vista but did they not say that this is a common misconception about Vista and that RAM loss was not what it seemed as the RAM is immediately released as soon as it is needed in Vista?

Click to expand...

Vista users are the ones with delusions when it comes to denying others observations.
www.crn.com/software/207001890
stopwatches and ram measuring tools do not lie.
I do not mind 1.5 gig of ram consumed for VISTA if I have 8 gigs or 16 gigs, I was only mentioning to you that for your case, adding 3 gig more and switching to VISTA from an XP flavor would net you only 1.5 truly 100% available free gigs of RAM.
MS KB940105 Hotfix gave back 800 megabytes of lost RAM under Vista, but not all the lost RAM. Example application for the 800 MB = "Company of Heroes mission 6".
But I am talking about wasted RAM not related to disk prefetch or disk cache, but rather truly allocated RAM held in swap. And Vista slurps down nearly 1.5 gigs of it from what I recall reading several times. Some alarmists claim 2GB. Its just a fact that VISTA is a slow resource wasting bloated fiasco when compared to XP.

The only bad thing is that in a few years some massive games might desire 64 bit app workspace (well over 2 gigabytes for certain), and 32 bit versions of XP will have to be eventually be left behind. Also total RAM, total number of cpu cores and cpu chips, and other factors mean a short life for XP on big boxes. OSX might take up some of the slack on the high end.

c627627 said:

4. Yup Blue Screen of Death with AHCI instead of IDE enabled. When you say "install AHCI support," I'll google it but is the procedure mobo specific?

Click to expand...

mobos use controller chips from very few SATA supplying companies, and the companies make drivers that are very backward compatible, and therefore its not hard to slipstream in , or add , AHCI drivers you need onto your bootable media. Some can be installed under windows before you reboot and switch to AHCI.

c627627 said:

5. Re the part about "slow down head seeks over 40%" first off, I don't have raptors just regular 7200 RPMs will that help me? Second, what!? up to 40% increase just by inserting what where? Pics please or links but this is major, why is this not mentioned everywhere? So I'll be definitely following up on that jewel:
1. Tell me what to use to measure this.
2. Help me understand what exactly I am supposed to do to set up this experiment & I'll post pics & results.

Click to expand...

I used ioMeter from Intel (www.iometer.org), which you can compile yourself but typically is downloaded precompiled. then I setup ioMeter to only use a 20 gigabyte range of blocks to be fair to disk head movement of mission critical database simulation, and used outer edge of disk platters of 4 raptor drives with latest firmware in one of the fastest 1-U rackmounts ever sold at the time with two of the fastest cpus, and the worlds fastest SATA cards at the time going to the 4 drives. The ioMeter was set to do READ of tiny blocks in a random pattern, and io per second was being measured to see if each IO under 7.5 millisecond or so. (15K RPM SCSI 320 drives are under 5) (Random Read IOPS) For over a week I could show full speed on any ONE drive but not with drives NEAR EACH OTHER. I tried suppressing EMI radiation, radio, then assumed magnetic, or other cross noise, power problems , etc etc etc. NOTHING got drives to go full head seek speed. Then I was told to cram two small pieces of accoustic ballistics ear foam on the rails ramming the drives in tight but cushioned by the foam. Thnese were removably mounted drives. All 4 drives got 40% more head seeks per second sustained immediately. Not all drives are probably that crappy. But WD740ADFD Raptors were. I could not use SCSI 320 because I was working for a SATA controller company at the time.

By the way, a cheapie 2003 Maxtor 15K RPM SCSI 320 8C018x0 is 3.2 milliseconds seek time. Hitachi, Fujitsu, Seagate, etc had even faster 15,000 RPM drives in 2003. Now 6 years later, still not one 15,000 RPM drive for SATA. SATA is for people who do not care about how fast their system runs or fetches back data from virtual RAM.

YOU DO KNOW THAT YOU DO NOT NEED 10,000 RPM RAPTOR for large file I/O speed. In fact many 7200 RMP sata drives area faster in sustained speed. Such as Seagate Barracuda 7200.10 750GB 7200rpm SATA 3G. refere to these benchmarks of cramming 4 of each of 4 brands into a standard mac with standard controller chip and apples free raid back in 2006 : www.barefeats.com/quad07.html Four Seagate SATA 750 GBs = 294MB/s READ, 284MB/s WRITE

As for ANY drive getting faster ios per second using acoustic foam, OTHER than sensitive Raptors, its possible, but they were the only ones I saw that had such problems with ultrasonic noise. AUDIO low frequency noise does the same thing to all normal drives (vibration), as can be seen in this hilarious datacenter example of a guy yelling into the middle of a RAID bank at a data center at Sun :
www.youtube.com/watch?v=tDacjrSCeq4

Wow what a video.

c627627 said:

6. Thanks for SCSI info, it makes sense, let me ask you: what do you see as far as the Solid Sate Drive roadmaps go, what do you see in the near future for storage size & speed?

Click to expand...

Digital photography lovers kept predicting that a digital camera would be BETTER in resolution than a 50 dollar 35 millimeter camera loaded with Kodachrome 25 or 64 (en.wikipedia.org/wiki/Kodachrome) taking a picture of a resolution line test at high noon. They predicted Digital would be better than analog film by 1998, then 2000, then 2002, then 2004, then 2009. it still never happened. 4 THOUSAND alternating black and white thin horizontal lines can be photographed AND COUNTED from one side of a negative to another at any reasonable angle of rotation of the camera. In 2009 the highest resolution DIGITAL camera in the world is still the $22,995 Super 10K-HS www.betterlight.com/superModels.html. It counts green pixels twice and uses RGBG per 4 pixel cel . or rather 10200 x 13600 pixels / 4 (RGBG) = 5,100 x 6,800 REAL resolution in greyscale. If I rotate the target resolution chart a few degrees, the 5,100 lines is NOT ENOUGH to photograph and count the 4000 lines of resolution I can capture on a camera from 1950 clone using 1950s film for 50 dollars. $22,995 in 2009 IS STILL FUZZIER for line count edge to edge than a 50 dollar analog camera. The LIARs never stop.
The LIARs love to lie. The liars say "resolution" (resolving thin lines) is not so important, and that one day REAL SOON a digital camera will be shaper than a 50 dollar 35 millimeter cameras standard film stock. Sure the magical fairies from fantasy land will make it all true. Plus analog cameras can use negatives far far far larger than 35 millimeter !!!!! Some Analog use 4"x5" negatives!. Digital photography will not surpass analog photography at 4"x5" for RESOLUTION probably another 20 years based on the track record of lies.

The track record of lies regarding solid state devices is astounding. Almost legendary. I wrote drivers for the first thumb USB drives, as well as the worlds fastest Fibre channel PCI- cards. I would like nothing more than SSD not be a pack of sick lies.

But it is. Its so slow for a executing a giant list of pending 512byte random disk writes over a small database file compared to SCSI, no matter what budget you are speaking of from 1 thousand to $50,000, that I never will know when SSD will not be SLOWER THAN SCSI for reasonable partitioned zones. Maybe in 2 years, maybe in 20. Who knows? But it never happened YET. Just like photography in the last 10 years of failed predictions and lies.

The SSD designs use either 1024 or 2048 discrete individual slow controllers for flash writing and each is slower than crap for writing. When dumping far far more than 2048 disk blocks in a random write benchmark the SDD all come to a screeching halt on graphs and look like floppy disk drives. I am no SCSI 320 bigot, I am just passing along the truth.

c627627 said:

9. Re audio-only stutters, unless I posted up there that I noticed them, I'm trying to think if I did, but I'll definitely pay attention to see if I hear them - then implement your suggestions to address them.

Click to expand...

If you had no audio stutters then most of what I discussed does not apply. You seemed to imply your machine was not keeping up.
If you want a faster zippier machine, maybe you need to explore an OS that will waste SOME of your CPU but keep your cores busier on average. Mac OS X, uses a MACH multi cpu design through and through. BSD is a layer over MACH. And MACH and other parallel additions are all about the future of people having 16 core machines and more. http://en.wikipedia.org/wiki/Parallel_computing

Your GPU is an example of such a resource. Apple even uses the GPU for non video computation : FFT analysis and much much more, all callable by preexisting libraries for programmers. True, Apples CORE IMAGE GPU card library has 100 graphics plugins as part of OS : http://en.wikipedia.org/wiki/Core_Image , but they also have other non-2d graphics libraries.

Apple in the next system version (under testing now) offers Open Computing Language (OpenCL). OpenCL allows programmers DIRECT CONTROL OF EVERY CPU CORE AND DIRECT CONTROL OF GPU CARD using a new c-like language, and VM is expanded to 16TB of PAGED RAM, and 42 bit hardware addressing as the last 5 years of mac motherboards.

Apple also turns off some less busy CPU cores and jacks up clock speed on busy cores on its 8 core designs. (Apple "Turbo Boost" 12% clock speedup).

All I am suggesting is that if you are going to use extremely powerful hardware, you might want to tinker with installing some other OSes. OS X, Linux, anything. You might feel less sluggish doing some of the stuff you discussed.

c627627 said:

10. Yeah sure I can get Mac OS X installed on my system, I was just looking at how someone was doing it the other month, the only reason I did not look into details is because I could not think of a reason as to why I would want to for personal use... if I can read or hear a useful reason to do so personally, I'd give it a shot. In other words: Mac... OS X... what is it good for?
/QUOTE]

Personally, I've seen most people that switch to using Mac OS X , tend to only want to ever boot into windows 15% of the time or less, and say the same thing you said but the other way around. If you had a mac buddy with a huge pile of software, and you saw enough demos of stuff you'd see what all the excitement is about. Besides you can still have XP windows intermixed with mac windows on the same screen using a Virtual machine such as Parallels. When you double click on a document it launches either a mac or a windows context to run your document in the correct application.

DO NOT DISCUSS WAREZ HERE! This is a bannable offense, but this is only a warning. Thank you for your understanding - IMOG

Click to expand...

you can install AHCI after the initial install. Vista disables the drivers it doesn't use durring installation, so you have to go in to the registry and enable it http://support.microsoft.com/kb/922976

I'm on limited time, so please pardon my cherry picking. There are a couple items I feel should be addressed here for the sake of technical accuracy. Much of the input is interesting and may very well be accurate. The parts I've had time to look at and discuss here are NOT accurate.

driverguru said:

Vista users are the ones with delusions when it comes to denying others observations.
www.crn.com/software/207001890
stopwatches and ram measuring tools do not lie.
I do not mind 1.5 gig of ram consumed for VISTA if I have 8 gigs or 16 gigs, I was only mentioning to you that for your case, adding 3 gig more and switching to VISTA from an XP flavor would net you only 1.5 truly 100% available free gigs of RAM.
MS KB940105 Hotfix gave back 800 megabytes of lost RAM under Vista, but not all the lost RAM. Example application for the 800 MB = "Company of Heroes mission 6".
But I am talking about wasted RAM not related to disk prefetch or disk cache, but rather truly allocated RAM held in swap. And Vista slurps down nearly 1.5 gigs of it from what I recall reading several times. Some alarmists claim 2GB. Its just a fact that VISTA is a slow resource wasting bloated fiasco when compared to XP.

Click to expand...

This memory comment is inaccurate. Refer to this article specifically the Memory Priorities and Superfetch sections. These are things that most people don't understand about Vista, and its why your comment is inaccurate:
http://technet.microsoft.com/en-us/magazine/2007.03.vistakernel.aspx

Your specific comment about the hotfix is just one bug that was addressed. That article I linked explains why memory management is not as bad as people like to claim in vista.

I am a Gentoo user. I don't care for Vista either, but memory management is not an area where Vista should be criticized in comparison to XP. It's not bad, just misunderstood by epic proportions.

4> Disk I/O Vibration
A case that VIBRATES ultrasonically will slow down head seeks over 40%. Use acoustic ear foam for gunnery range wedged in both sides if a drive is insert rail mounted. That always restores the missing 40% I/O/sec on Raptor drives for me. The ultrasonic lateral vibration from other head seaks causes a landing miss greater than 1 track, and the drive goes into a different safety head seek mode to increase head seek guess accuracy. You will not read about this secret anywhere because I never mention it. A head VP of Gigabyte showed me the trick and it worked.

Click to expand...

VPs are business men. Businessmen are notorious for misunderstanding and misquoting technical facts. I'd put more stock in it if it came from an engineer with statistics and reproducible methods.

While vibration can effect drive performance, a 40% impact on IO/sec due to vibration is demonstrably false on all but the most extreme cases. In an extreme case, sure. In typical setups this cannot be reproduced reliably.

A single allegory is not proof of fact. If you get those results EVERY time, you are doing something very wrong.