|
Comment #21
Dear Adrian,
Thanks for all your efforts about BIOS setup. Can You assist me with my problem?
Our OS is NT 4.0 running at 100Mbps (hardware is PII-400MHz, 128MB PC100
SDRAM) for our
point of sale application. I have 2 types of hardware terminals: Cyrix 133MHz CPU (16MB)
and Pentium II-350Mhz CPU (32Mb) I noticed that the 586 is running and accessing the
server much faster than the Pentium II.
By the way, I have RAID 1 (hardware is Adaptec AHA133 on the NT4.0 server with mirrored
9Gb Seagate Cheetah SCSI disks).
Shall appreciate your immediate reply, if you cannot give adequate info, do you know of
some sources on networking that can advise me on networking.
We're hurrying up the implementation of our project. Thanks and God bless.
Rey Del Rosario
Hello Rey,
I'm terribly sorry but this kind of setup is way out of my league. Worse still, I have
no idea where you can look for information on how to rectify your problem. :-( However, please check in here once in a while. Hopefully,
someone knowledgeable will answer your query. When they do, I'll be sure to post it here.
Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/
Comment #22
Adrian
This is in reference to "PnP OS: (yes)/(no)"
A good friend of mine was recently piecing together a computer. He could not get the
Win98 to recognize his CD drives (a CD-RW, and a 52X CDROM). While I was playing around
with his BIOS, it seemed like a logically thing to turn it on; since both CD drives were
PnP compatible. Sure enough, as soon as it was turned on and rebooted the system
recognized the CD drives and stopped disabling the dual FIFO drivers for the hard disk =)
Just my quick comment on that option.
Michael Fraser
psHardBall
Thanks for your observations on the PnP OS option, Michael!
I have never been affected in this way before but I think there will be quite a few
people out there who will be in the same situation as your friend. The account of your
experience will surely help them. Thanks again! ;)
Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/
Comment #23
Hi!
I've just read your guide and I find it great! But I'd like to make some comments:
About System&Video BIOS Shadowing: these BIOS are generally located in the
C0000-C7FFF and F0000-FFFFF areas in real mode. Actually, the memory modules are located
there, over the system RAM. If you choose to activate BIOS shadowing, the ROM BIOS will be
copied into the portion of RAM used for shadowing and then the ROM modules configured in
order for the RAM to be "on top". The RAM is then locked into read-only state so
writing there is no more permitted than on ROM BIOS, and the address ranges are the same.
I don't really know how the BIOS is configured in protected mode (Windows, Linux, OS/2...
well, every OS but DOS ;-) ) but anyway the processor would not allow a non-ring 0 process
to write on such an area. I don't think any game would try to write on BIOS, so why would
they try to write on write-protected RAM (which is at the same address range that the BIOS
they shadow)...
About System & Video BIOS Cacheable: the processor access to the Video RAM by the PCI
or AGP bus, which are 32 and 64bits bus. Unfortunately, BIOS chips use (uh, they were some
years ago, but I don't think things changed a lot about this) an 8 bits wide data bus, and
the access time is far over DRAM or SRAM's ones. If you disable shadow RAM, the
performance hit will be great when a program uses good old BIOS. If shadow RAM is
activated, the performance loss will of course be smaller. If none of your program use
your BIOSes, well it won't go into cache and you'll lost nothing, no cache nor speed...
About Video RAM Cacheable: I wonder if talking of the A0000-AFFFF range for the Video
RAM make sense in protected mode? There's other (and more protected from the point of view
of a multi- processing system) ways to access Video RAM. In real mode, you're quite right,
except that the bus used to access Video RAM is a 32bits 33MHz one (PCI) or a 64bits 66MHz
one (AGP 2x) so that make a bandwidth of 133Mb/s for the first one and 533Mb/s for the
second one. The bottleneck is not the bus between the graphic chip and video RAM for a
transfer between video RAM and the system!
Also, about a comment I read on your Comments page, about IDE drives: I think the way
the BIOSes manage HD geometry is a non-sense... but that's not the point :-) well the
504Mb (I'm talking of real megabytes, you know, the 1048576-bytes ones ;) ) limit applies
only for Int 13h accesses. Internally the ATA interface, starting from the very first
versions, accept CHS and, for many years now, LBA. Both gives access to about 120Gb of
storage space. If you're using Windows I don't think BIOS is used anyway, and Windows has
its own methods to access data. The 504MB is only the consequence of an absolutely stupid
design of the Int 13h interface that's not used by Windows. That's probably why you,
Kevin, are able to use your 6.4Gb HD in spite of the BIOS setting. I've not tested this,
but it sounds logical. By the way, that doesn't mean you can change your BIOS setting
anytime, because the geometry provided by the BIOS is still used in the partition table,
because Int 13h is used at boot time. If you change this you may have troubles accessing
your partition(s)...
I hope my english was not too ugly to read ;-)
Adrien Huvier
Hello there, Adrien,
I guess that statement about programs writing into the shadowed BIOS area is only true
for real mode DOS programs and improperly written ones at that. But it has happened before
which is why most manuals will include a warning of some sort regarding the shadowing of
the BIOS.
That's true. I believe that most motherboard manufacturers already hardwire the
shadowing of the motherboard BIOS. Thus, in most BIOS, only the shadowing of the Video
BIOS can be disabled or enabled at will. Since the Video BIOS is never used in most
circumstances, it would be pointless to shadow/cache it. But I see your point there and
it's a good one. I believe this goes to show that we need more research on this.
I have never advocated caching the Video RAM because I have previously tested it during
The
Phoenix Project case studies. They have gone to show that caching the Video RAM reduces
performance instead of improving it. As I have mentioned earlier, the L2 cache's bandwidth
would be better used to cache the slower system RAM instead of the already fast video RAM.
I understand that you are stating that the bottleneck is the bandwidth between the video
RAM and the system RAM.
However, you will note that the trend these days is to increase the size of the onboard
RAM. That's because AGP will never be able to provide enough bandwidth for texture
storage. So, by loading all the textures into the onboard RAM, AGP's deficiency can be
ignored. Thus, caching the video RAM will have minimal effect on the bandwidth between the
video RAM and the system, and a rather detrimental effect on the overall system's
performance.
Hey, thanks for that observation! I'm sure Kevin would be very happy to read about it.
:)
Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/
Hi,
I just read your Phoenix Project page! Benchmarks definitely have the last word ;-)
There's one point I can't get about your Video RAM caching explanations. You compare
the FSB bandwidth with the internal video bus bandwidth. But the bus between the chip and
the video memory is internal to the card, it never gets through the motherboard, right?
So, I can't see where the L2 cache could do any job there, since it doesn't
"see" the datas that use the video bus without leaving the video card...
Don't you think the cached access would be the CPU accesses to the video memory
instead? That sounds more logical since I don't see other kinds of video memory access the
L2 cache could be involved in... The bandwidths to compare would then be the CPU-to-L2
one, as before, and the AGP or PCI one...
However, I'm convinced by numbers, and I agree that enabling Video RAM caching could be
a waste of L2 cache resulting in a loss of performance...
Adrien Huvier
Hello Adrien,
If I'm not mistaken, the L2 cache is being used to cache the video RAM via the AGP/PCI
bus when Video RAM Caching is enabled. Thus, data in the video RAM is transferred to the
L2 cache via the AGP/PCI bus. Then, when that piece of data is needed, the L2 cache sends
it directly to the graphics processor, again via the AGP/PCI bus. As the AGP/PCI bus is
slower than the internal pathway of the graphics card, it's like sending the data through
a bottleneck twice!
When I compared the bandwidth between the onboard RAM and the graphics processor with
the bandwidth of the L2 cache, what I was trying to do is show that while caching the
video RAM may look great, it would be a folly to waste the L2 cache's bandwidth on the
video RAM which already has such a high bandwidth, rather than use it to improve the
comparatively pathetic bandwidth of the system RAM.
Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/
Comment #24
Hello,
I've just read your pages on the BIOS parameters explanation/optimization, and your
statement about disabling Video BIOS shadowing reminds me of an experiment I did some time
ago :
Seeing that this parameter was enabled by default in my AWARD BIOS, and seeing that
many people on the net were saying it should be disabled, I have run 3DMarks on my TNT PCI
card with both values of the parameter. The result is: I had a noticeably better score
with the parameter enabled.
So I always leave this parameter enabled, as it is by default. Thank you for your work,
Jean-Marc Berthaud
Hello Jean-Marc,
Now, that's certainly food for thought. While I have either noticed no difference or a
slight drop in performance when I enable Video BIOS Shadowing, I think this feature
warrants a closer look. Thanks for your contribution to this project. ;)
Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/
Comment #25
Thanks for the BIOS Optimization Guide... it should be included in every MB manual!
Anyway, I was wondering if Video RAM Caching should be enabled or not. I thought Video RAM
communitcates with outside by using either AGP or PCI bus, which is running at 66MHz or
33MHz. And I don't remember the bus bandwidth is 64bit or 32bits...but 128bits seems a bit
high. Even if running AGP in 2x mode, the Video RAM is still no comparison with L2 cache
since L2 cache is running either at the speed of CPU (as case for Celeron) or 1/2 speed of
CPU. This is generally much faster than the 66MHz. Also, CPU generally look for data in L2
first instead of RAM. Which will cause extra cycles to look for data at Video RAM because
CPU didn't find it in L2. I could be very wrong at this since I am just guessing by what I
know. Have you run some benchmark with this option enable/disabled?
Thanks.
Lee Chung-Ju
Hello Chung-Ju,
Nope, as I have mentioned in the Guide, Video RAM Caching should not
be enabled, irrespective of what the manual says. The Video RAM does communicate with the
system bus at 66MHz (AGP) or 33MHz (PCI) through a 32-bit bus (both AGP and PCI). The
128-bits bus width I mentioned was between the Video RAM and the graphics processor.
What Video RAM Caching does is cache the data of the Video RAM. This data is used by
the graphics processor, not the CPU. So, it would be rather pointless for the CPU's L2
cache to cache the Video RAM because the data will have to go through the slow AGP bus
twice to reach the graphics processor. The Video RAM-to-graphics processor's bandwidth is
incredibly high compared to the bandwidth of the AGP bus and also the system RAM. So, the
L2 cache would be better utilized to speed up the system RAM
You can check out the Phoenix
Project cases for benchmarks on this BIOS setting.
Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/
Comment #26
Aye fellow Malaysian...
I read ur article about tweaking the BIOS... Damn cool man but contradicts with wat I
have read in Tweak3D (the Video RAM BIOS cacheable and stuff like that). Ur guide iz
really more detailed and yupz, I am gonna follow wat u say in ur guide.
cheerio
Renlin Li
LOL! Well, it also contradicts what a lot of motherboard manuals say too but I based my
recommendations on what I observed during my tests.
Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/
Comment #27
Had a question about part 2 of your BIOS guide:
I do not have any ISA cards, so I was wondering if there is any option to speed up the
overall system performance with some setting. (ex. your 8-bit and 16-bit recovery times).
It seems if I don't use any ISA cards at all, the system shouldn't need to add the minimum
3.5 clock cycles you mentioned.
Any help would be greatly appreciated!
Derek Lee
Hello Derek,
If you don't have any ISA cards, then those BIOS settings (8-bit and 16-bit I/O
Recovery Time) would be meaningless. They are only applicable if you have ISA cards.
Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/
Comment #28
Hello and thanx for the great guide
I found some mysterious options in my BIOS' Chipset features section which u didn't
explain - yet:
1) Auto configuration
2) EDO DRAM speed selection
3) EDO CASx# Ma Wait State
4) EDO RASx# Wait State
5) SDRAM CAS Latency Time
6) Auto detect DIMM/PCI Clk
I will be greatful if u will include these in the guide.
Thanx in advance
Dima
Hello Dima,
The Auto Configuration setting just sets a few critical timings to values recommended
by the manufacturer. If you enable it, you may not be able to change some of the BIOS
settings.
I didn't include all the BIOS functions for EDO RAM because practically every new
system uses SDRAM and I didn't think the EDO RAM functions would be of any interest to you
all. But in any case, I'll just run through them quickly.
EDO DRAM speed selection usually has settings like 60ns and 70ns; and sometimes 50ns as
well. You set it to the speed of your RAM so that the correct timings will be applied.
However, it can be used to "overclock" the EDO RAM a little by using a lower
setting than what's rated for your EDO RAM. Both EDO CAS and RAS Wait States can be
reduced for added performance but that can cause instability if the values are too low.
SDRAM CAS Latency Time has been discussed in the Chipset
Features Setup part of the BIOS Optimization Guide.
The Auto Detect DIMM/PCI Clk function is exactly like the Smart Clock option in some
motherboard BIOSes. For more details, read about the Smart Clock option. It's under the
Spread Spectrum function in the Chipset
Features Setup part of the BIOS Optimization Guide.
Hope that helps you some! :)
Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/
Comment #29
Hi Adrian,
Enjoy reading your BIOS Optimization guide. Another good job done by you.
For all this while, I've been stuck with all the setting under BIOS (Doesn't really
know what are the setting for), but through your BIOS section, I gain a lot of information
on how to set my BIOS to optimal.
By the way, just some problem facing recently. I've upgrade my power supply to
ATX. I
am using Gigabyte Super 7 SIS chipset which support ATX power supply, but I just couldn't
shut my PC down properly. It always restart even though I am trying to shutdown. Just
wonder you could advise me on this problem.
Thanks and keep your good job.
Regards,
Melvin Loh
Hello Melvin,
I'm happy that the guide was of some help to you.
Unfortunately, I'm not really sure what's wrong with your Gigabyte motherboard. But I
think it has more to do with your OS rather than the BIOS. There are a lot of
possibilities why your system keeps rebooting when you want it to shut down.
You should scan your registry using Norton WinDoctor or a similar utility. Also, try
removing all the programs in the StartUp folder as one of them may be the cause. But the
best thing would be to contact Gigabyte about this problem. I wish I could help you here
but I have not touched a SiS Socket 7 motherboard before. Terribly sorry about that. :(
Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/
Comment #30
Hi !
Well I want to add something about the 'delayed transaction'. You said to enable
it.
But I've an ABIT BH6 with the HN BIOS, with Win98, and when I enabled this option, it
made the computer reboot indefinitely after the 'DPMI check' (maybe it was the Windows
Boot Phase but I don't know since DPMI Check was the last message displayed).
Flash
Hello Flash,
I'm unaware of any such problem with Delayed Transaction. But you could have an addon
card that does not support PCI 2.1 and thus Delayed Transaction. However, I have never
seen a card like that cause such problems. But in your case, I think it's better you
disable Delayed Transaction.
Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/ |
