Please report all spam threads, posts and suspicious members. We receive spam notifications and will take immediate action!
Page 1 of 3 123 LastLast
Results 1 to 10 of 22

Thread: tRD "Performance" bios bug?




  1. #1
    Join Date
    Apr 2008
    Posts
    942

    Exclamation tRD "Performance" bios bug?

    As many of you will know, the memory setting in the bios tRD has the biggest single affect on memory performance.
    Using MemSet from Tweakers.fr for "on the fly" memory tweaking and say Everest for memory benchmarking you can see the differences yourself as explained in the Anandtech articles below:-

    AnandTech: ASUS ROG Rampage Formula: Why we were wrong about the Intel X48
    AnandTech: ASUS ROG Rampage Formula: Why we were wrong about the Intel X48

    However, during my recent tweaking I believe I may have found a bug in my bios (see sig) that may be present on other Gigabyte boards/bios version combinations.
    Normally this wouldn't be a big deal, but as it does have a significant performance hit on memory throughput (by setting the bios incorrectly), I thought i'd share it.

    Scenarios and the relavent settings that highlight the problem:-
    (Cpu Host frequency 333Mhz)

    Setup a
    MIT Section-------------------------------->Windows MemSet 3.5

    System Memory Multiplier = Auto
    (800Mhz)

    tRD (7) Auto-------------------------------->tRD = 7

    Setup b
    MIT Section-------------------------------->Windows MemSet 3.5

    System Memory Multiplier = 2.40
    (800Mhz)

    tRD (7) Auto-------------------------------->tRD = 9

    Setup c
    MIT Section-------------------------------->Windows MemSet 3.5

    System Memory Multiplier = 2.40
    (800Mhz)

    tRD (7) 7*---------------------------------->tRD = 7

    *manually set to the same value as Auto

    1) So basically with the settings in question set to Auto, there isn't a problem. The bios shows the tRD value of 7 and so does MemSet in Windows - hoorah!

    2) However if you decide to set the System Memory Multiplier manually, ie in my case to 2.40 (333 x 2.40 = 800), without altering the tRD value from it's Auto Value (which is displayed in the bios correctly) - something causes that value (tRD) to change as you proceed into windows! Which is confirmed by Memset (and benchmarking to rule out a Memset reporting error).

    3) But if you Set tRD in the bios to the value of 7 Manually (as per the value Auto suggests) effectively "locking" it, Memset will report the same (ie 7) when in windows.

    So for some reason, if you choose to set The System Memory Multiplier Manually, but leave as per you might/would normally do with other memory bios settings to Auto - it might (as with my setup) be changing settings that you might not neccessarily think were connected to it.

    There is a reasonable Memory performance hit by tRD being raised from 7 to 9, so I think this is worth noting.
    See Benchmark:-
    AnandTech: ASUS ROG Rampage Formula: Why we were wrong about the Intel X48
    Last edited by VorLonUK; 10-22-2008 at 07:58 AM.
    GA-P35C-DS3R Rev2.0 F11 bios, E8200 (@3.0Ghz), OCZ DDR3 PC3-10666 Reaper 4GB (@1200Mhz), Xonar D1, 8800GTS 512, Corsair HX520 (Single 12volt line, Max 40A), WDC 3200aaks/5000aaks in AHCI mode, Vista 64 Premium.

  2. #2
    Lsdmeasap's Avatar
    Lsdmeasap is offline GIGABYTE Guru
    Join Date
    Jan 2008
    Location
    PCB Island
    Posts
    25,940

    Default Re: tRD "Performance" bios bug?

    I have always said Auto is not always what you get in windows. What you see for Auto on the left can be several things which even at the same settings can vary anytime you look in the BIOS. Auto (or what you see for auto on the left) can be what is being used Right then in the BIOS, Can be what you have used in the past, can be what is found in the SPD/XMP chip on the ram, and yes sometimes can be what you get in windows.

    Main thing to keep in mind I think is that this is not a Bug, Auto will change on the fly to what it wants, depending on several things, and will not always show you the correct values on the left in the BIOS.

    So what you see changed from 2.00 to 2.40, was this on the same ram speeds? I see you noted that it was 800, just wanted to ask to be sure. If it for sure was the same ram speed with a lower FSB that may be why Auto chose 9 instead, a lower FSB would/could mean a tighter NB Strap that would be set Automatically so Auto may have decided 7 would be too tight for that strap to run stable. Yes, you may find it stable but when the board is Guessing that is all it can do is guess (Especially with 2x2GB or more)

    All in all, Auto is not always what you see/think it is using. So best, if you know about a Setting and it's effects to manually set it.

    Did I explain that understandably? If not, let me know what I said confusing and I will try to better answer it

  3. #3
    Join Date
    Apr 2008
    Posts
    942

    Default Re: tRD "Performance" bios bug?

    So what you see changed from 2.00 to 2.40, was this on the same ram speeds?
    I'm not sure if I see how.....

    It looks like Auto had already set 2.40 "behind the scenes" as my RAM was showing 333 x 2.40 in the bios - aka 800Mhz.

    It's when that Value was manually "Fixed" to 2.40 by me aka 333 x 2.40 = 800Mhz , ie the exact same thing, that Another Value tRD which was set to Auto Changed.

    Throughout all, my RAM stayed at 800Mhz, but tRD was being changed, simply by changing another setting (memory multiplier) from an AUTO 2.40 to a FIXED 2.40!

    Surely if the memory multiplier was switching between 2.00 and 2.40, I would see my ram changing from 667 to 800Mhz in both the bios and in windows - neither which I have noted.
    GA-P35C-DS3R Rev2.0 F11 bios, E8200 (@3.0Ghz), OCZ DDR3 PC3-10666 Reaper 4GB (@1200Mhz), Xonar D1, 8800GTS 512, Corsair HX520 (Single 12volt line, Max 40A), WDC 3200aaks/5000aaks in AHCI mode, Vista 64 Premium.

  4. #4
    Lsdmeasap's Avatar
    Lsdmeasap is offline GIGABYTE Guru
    Join Date
    Jan 2008
    Location
    PCB Island
    Posts
    25,940

    Default Re: tRD "Performance" bios bug?

    Likely when you choose 2.40 you choose a different strap on 2.40 then Auto did.

    Having two values in Auto you can never see what the BIOS is really setting, mainly the Memory multiplier/strap combo. If it, or you, choose a Tighter strap and have tRD in Auto it is going to raise the tRD for sure. If you choose a looser/higher strap or none it may or may not be affected.

    Likely when you had Auto for Memory Multiplier it was choosing 2.40D which would allow a looser tRD, and when you set it to 2.40 you choose 2.40B or plain 2.40 which Auto thought needed a higher tRD to boot. Since we cannot see what it is choosing we cannot be sure, but I think that is why.

    I assure you though, it is not a bug. It is just one of the many reasons why Auto is not the best choice, it often will set a value higher then need be to try to assure that the system will boot

  5. #5
    Join Date
    Apr 2008
    Posts
    942

    Default Re: tRD "Performance" bios bug?

    Likely when you had Auto for Memory Multiplier it was choosing 2.40D which would allow a looser tRD, and when you set it to 2.40 you choose 2.40B or plain 2.40 which Auto thought needed a higher tRD to boot. Since we cannot see what it is choosing we cannot be sure, but I think that is why.
    Are you saying there are various versions/profiles of say the 2.40 Multiplier, ie x,y,z (when set to Auto). Yet the user has access to setting just one 2.40 Multiplier profile when it is setup manually?
    GA-P35C-DS3R Rev2.0 F11 bios, E8200 (@3.0Ghz), OCZ DDR3 PC3-10666 Reaper 4GB (@1200Mhz), Xonar D1, 8800GTS 512, Corsair HX520 (Single 12volt line, Max 40A), WDC 3200aaks/5000aaks in AHCI mode, Vista 64 Premium.

  6. #6
    Join Date
    Sep 2008
    Location
    Canberra, Australia
    Posts
    206

    Default Re: tRD "Performance" bios bug?

    i believe this is a simple SPD issue (if you can call it that)... PC2-6400 DIMMs generally have an SPD which sets the default values according the the JEDEC standards for DDR2-667.Essentially, the "Normal" values are the default timings as outlined in the SPD at default clock, so when it says 7 beside Auto, that is the timing which will be set at stock speeds (which, strangely, for DDR2-800 is 667MHz... i believe this is because of the nature of the sales of RAM, they make 667s and if the pass at 800, they'll sell em at 800), so i'm wondering, when you set the multiplier to Auto, are you sure its at 2.40 and not just 2?

    if you use Auto with a higher clock speed, your BIOS will read the SPD and change the timings accordingly (usually, this is just a rule from experience, not concrete at all), therefore it is always best to manually set your timings, as this will ensure you always know what your timings are being set at.
    NZXT Lexa BlackLine Case with Red LEDs
    Corsair TX-750 Modular PSU
    Gigabyte EP45-DS3P (BIOS F7)
    Intel Core 2 Quad Q9300 @ 3.73GHz (downclocked for the summer )
    OCZ Vendetta 2 CPU Heatsink
    2x 2GB Kingston HyperX RAM PC2-6400 @ 996MHz (4-5-5-15)
    3x Seagate Barracuda 7200.11 500GB HDD in RAID5
    2x HIS Radeon HD 4850 @ 745/1090 in CrossFireX
    2x ThermalTake DuOrb GPU Heatsink
    D-Link Super G PCI WLAN Adapter
    Asus Lightscribe 20x SATA II DVD-DL Burner
    Creative X-Fi XtremeGamer Fatal1ty Pro
    2x 22" LG W2252TQ-PF Widescreen LCD Monitor (YAY!! Finally got the second one!!)

  7. #7
    Join Date
    Apr 2008
    Posts
    942

    Default Re: tRD "Performance" bios bug?

    so i'm wondering, when you set the multiplier to Auto, are you sure its at 2.40 and not just 2?
    But then surely after saving the bios, rebooting and entering the bios again, the memory speed would be displayed as 667, ie 333.xx x 2.00 - which it's Not.
    Both the bios and CPUz indicate 800Mhz (333 x 2.40) so surely the Multiplier has been selected under "auto" as 2.40 - correctly as per the SPD table.
    GA-P35C-DS3R Rev2.0 F11 bios, E8200 (@3.0Ghz), OCZ DDR3 PC3-10666 Reaper 4GB (@1200Mhz), Xonar D1, 8800GTS 512, Corsair HX520 (Single 12volt line, Max 40A), WDC 3200aaks/5000aaks in AHCI mode, Vista 64 Premium.

  8. #8
    Lsdmeasap's Avatar
    Lsdmeasap is offline GIGABYTE Guru
    Join Date
    Jan 2008
    Location
    PCB Island
    Posts
    25,940

    Default Re: tRD "Performance" bios bug?

    Yes, Auto may choose 2.40x or just 2.40, and the 2.40x has four variables (2.40A/B/C/D Or 2.40+/~/#/? (not sure of p35 400 strap symbol)) that may be used depending on CAS and what the FSB is and what the board thinks may be stable with all other settings and voltages.

    And no, you may choose any of the above variables when you set it manually, depending on your CPU and FSB Speed. Here I laid out some Strap tables for someone here >>>
    http://forums.tweaktown.com/f69/gott...27/#post263166

    Those are available to you, you may choose 2.40 with no strap letter/symbol, or you may choose any of the ones available to you with your CPU and FSB Speed

    It is not a glitch or anything at all, for sure.

  9. #9
    Join Date
    Apr 2008
    Posts
    942

    Default Re: tRD "Performance" bios bug?

    Quote Originally Posted by Lsdmeasap View Post
    Yes, Auto may choose 2.40x or just 2.40, and the 2.40x has four variables (2.40A/B/C/D Or 2.40+/~/#/? (not sure of p35 400 strap symbol)) that may be used depending on CAS and what the FSB is and what the board thinks may be stable with all other settings and voltages.

    And no, you may choose any of the above variables when you set it manually, depending on your CPU and FSB Speed. Here I laid out some Strap tables for someone here >>>
    http://forums.tweaktown.com/f69/gott...27/#post263166

    Those are available to you, you may choose 2.40 with no strap letter/symbol, or you may choose any of the ones available to you with your CPU and FSB Speed

    It is not a glitch or anything at all, for sure.
    I understand what you are saying LsdMe, but if "Auto" mode can choose from a selection of 2.40 variants why can't I. I have simply the One 2.40 variant option to select manually, ie basic no suffix 2.40.
    GA-P35C-DS3R Rev2.0 F11 bios, E8200 (@3.0Ghz), OCZ DDR3 PC3-10666 Reaper 4GB (@1200Mhz), Xonar D1, 8800GTS 512, Corsair HX520 (Single 12volt line, Max 40A), WDC 3200aaks/5000aaks in AHCI mode, Vista 64 Premium.

  10. #10
    Lsdmeasap's Avatar
    Lsdmeasap is offline GIGABYTE Guru
    Join Date
    Jan 2008
    Location
    PCB Island
    Posts
    25,940

    Default Re: tRD "Performance" bios bug?

    You should for sure be able to, no wonder we were both not getting what the other had to say. Show me a pic please

    Ya man, you should have at least 2.40+/~/# to choose from. See here you can see the 4.00+ one in this review >>>
    Gigabyte P35C-DS3R DDR2/DDR3 (Intel P35) mainboard review - BIOS SCREEN SHOTS @ OCWORKBENCH

    Do you not see the strap info like shown here in post one's images >>>
    http://forums.tweaktown.com/f69/gott...ore-try-28027/

    You should see that, here is a larger image of how it should look at certain FSB's >>>


    And another example

Page 1 of 3 123 LastLast

Thread Information

Users Browsing this Thread

There are currently 1 users browsing this thread. (0 members and 1 guests)

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •