Slower speed after new ver!! plz help!

[gigabot]

As i posted in the "increased speed" thread, my speed has acctually droped and i am posting a new thread as per markb's instructions.
My speed is currently .30 or so, and am using Multipass VBR 4, and other defaults from the guide.

Here are my system specs

AMD Athlon 1.0Ghz
Asus A7V MOBO
512MB PC133 Ram
WD 40GB HDD <--- O/S on this drive
Maxtor 30GB HDD <--- where encoding takes place
Running WinXP Pro

Thanks in advance.

[xrv1138]

gigabot
have you gone thru your bios and set up optimized settings ?
and have u ever run sisoft sandra on your machine
ie do you have any system benchmarks that show your pc is running ok outside dvd2svcd ?

[ev]

Have you selected Temporal smoother by any chance? That will drop the speed dramatically.

Ev

[M1ckey M0use]

Does your mobo has via chipset and did you install via's 4-in-1 drivers?

M.

[xrv1138]

pretty sure that board has the via kt133 chipset.

[gigabot]

Bios tweaks: Where can I find these? Anything Specific?

SiSoft: Will test later today but I am quite sure its running fine.

VIA4in1: Latest ver from www.viahardware.com

Temporal Smoothing: When off, gets a little faster, but nothing to write home about.

Thanks, keep em coming :)

[markrb]

I looked up your motherboard. You have the KT133 chipset. With the XP OS I do not recomend updating the 4 in 1's. It's not needed unless you have problems that you can comfirm those will help with.

From the few posts I have seen your motherboard is very slow compared to others. You would be well served to upgrade to a 266a based motherboard with PC2100 DDR, but if you can't try these things.
Understand that any of these tweaks can cause corruption and you use them at your own risk. I doubt they will hurt your hardware, but you may be forced to re-install the OS if you have data corruption.

Get the PCI latency patch at www.viahardware.com
Read and try these tweaks http://www.viahardware.com/wpcreditkt133.shtm

Read the Via Tweaks forum at www.viahardware.com

Also understand that if you are doing PAL movies your speed will always be slower then those of us doing NTSC.

Make sure you have as little running as possible in XP.
There are many OS tweaks to try at www.tweakxp.com

Good luck,
Mark

[dvd2svcd]

And just for fun try setting the Priority to High in the Misc. Tab

[gigabot]

well.... i tried all the tweaks, even the memory one which i already had maxed out.

still only .3

anything else??

[Bullseye]

Also having the same problem as the guy above. Used default settings also.
Rips quickly but CCE speed varies between .28 and .31
Tried 2hr DVD and video encoding took 11 hours and 'saving' was estimated at another 19 hours before I aborted.
Previously used Smartripper, DVD2AVI and TMPGEnc202 and this was quick. Have run Sandra and found no probs. Not installed Via 4in1's because of XP. HDD is alone on the RAID controller so shouldn't be slowed. Any more ideas? Thanks.

Athlon 1200
512MB Ram
KT7A Raid
Win XP Pro
30GB IBM Deskstar75GXP with 15GB free

[jasond]

that seems very slow... try killing every other process.. I get about .400 -.550 on a 1 gig duron..

good luck

[markrb]

Your RAID controller might be part of the problem. Via chipsets have a bug in the PCI bus. This effects RAID controllers very much. Look around the forums at www.viahardware.com
You might try putting the drive on the onboard IDE controller, but in the chipset you have this is effected as well.

I really think the motherboard is at fault here. You are trying to compare speeds with people that are using DDR and newer chipset boards. The CPU is fine, but the bus is your limiting factor here. CCE as well as being CPU intensive is memory intensive as well and PC133 memory is 1/2 as fast as PC2100 memory, theoretically of course.

One thing to try is that if you are using NTFS change it over to FAT32. NTFS is faster for smaller or randomly accessed files, but for large continuos files Fat32 has less overhead.

Mark

[gerti67]

Hi Mark,

i really think that it must be the onboard RAID controller or some settings he uses, because the memory bandwith shouldn't affect it so much as i have nearly a comparable system:

- 1.2 GHz Duron
- 256 MB PC133 SDRAM CL2 Infineon
- EPOX EP-8KTA2 (VIA KT133/686B Southbridge)
- Win2k SP2
- 20 GB IBM + 40 GB Maxtor (both NTFS)

And i get a CCE RT of about 0.955-1.075 when doing PAL (i only encode PAL) with SimpleResize which is not bad for this not "up to scratch" system, i think.

Bullseye,

i suggest you attach your Harddisk to the Primary IDE of your Mainboard and see if this changes something, and tell us what you try to encode, and what are your settings, are you using temporal smoother, do you deinterlace, do you use BicubicResize, if you encode NTFS is IVTC used, ... this all effects more or less your overall encoding speed. (Try to post a complete log of an slow encode here)

HTH,
gerti67

[markrb]

gerti67 I based my assumption on the fact that both are using the same motherboard based on the KT133 chipset. I have seen this elsewhere as well. I was blaming the chipset, but it very well could be something else. It is odd that this particular motherboard keeps showing up as a problem. Maybe it's that this particular motherboard and CCE somehow don't like each other.

I would suggest taking a stroll over to the Asus forum at www.amdmb.com. They might be able to help you tweak some more performance out of the board. At least over there you would get more people that have actually used your particular board.

Another place to look for help making the board a little faster is the forums at www.viahardware.com.
I would stongly suggest you both grab the PCI latency patch by George Breese that you can get there.

If you can, disable the RAID controller and move your HD over to the regular IDE controller. These RAID controllers do take a little CPU horsepower away when they are used. Give it a shot. I might be wrong, but it's worth a try.

I really don't know what else to have you guys try. If you haven't done so already uninstall CCE and re-install it.

Mark

[gerti67]

Me again,

i was just curious as i'm always interested in learning more on AMD troubles (though i never had some) so i just had a look at the forums over at www.amdmb.com as you mentioned Mark.

And i must say, you are right, these mobos seem to make a lot of trouble when using WinXP on them. Although they are not the same, it's an Asus A7V and an Abit KT7A i assume that they are rather old with some low revision numbers. And as stated in the Abit forum revisions from 1.0-1.2 are really buggy (not to say the hell) when using Windows XP on them. This seems to be nearly the same with the Asus though i don't know which of the dozens of different A7Vs it is.

So i would suggest that you Bullseye and gigabot are trying to run Win2k on it and in the case of Bullseye (and if gigabot has a RAID variant of the A7V too) to run your Harddrive on the default onboard IDE as trouble seems to double when using the onboard RAID with a not up to date BIOS.

Greetings,
gerti67

[markrb]

I have been doing some reading and I found some interesting stuff. Not all of it is good news however. The original runs of that board used the 686A southbrige and not the 686B that gerti67 has. There wouldn't have been a B version if there were not problems with the A version. You may have the B version it's just that these boards came originally with the A version. Also many people are reporting poor hard drive performance as well. Especially where the RAID controller is concerned. There seems to have been 1 or 2 bios upgrades for the Promise controller since the board was released. Have you upgraded to the latest Bios for both the motherboard and the RAID controller?

One very interesting post over at www.amdb.com suggests to use the bios from the AV7133 on the AV7. A few of the posters got quite a jump in speed. Here is a link to the actual post.
Edit: afetr further reading into the post it seems the latest official bios for the A7V is just as fast as using the A7V133 Bios.
This new bios came after the post was started, but read it anyway it has some good info in it.
http://www.amdmb.com/vb/showthread.php?s=&threadid=24563&highlight=A7V

From reading everything that I did I am more convinced it is your motherboard. It is one of the very first boards made with the KT133 chipset.

You maybe able to tweak a little more performance out of the board, but I believe if you can afford it your best bet would be to upgrade. For the fastest performance if you wish to stay with Via a 266a board can be had for under $100. If you just want a temporary solution a newer KT133 or KT266 used board can be had for under $50 on the forums.

gerti67 I believe your board to be a newer board based on the fact you have the 686B and not the 686A these boards had.

Mark

[gerti67]

Ups Mark,

seems we posted nearly at the same time but isn't that bad after all as i was mainly reading the ABit forum.

Edit:
And yes, you seem to be right with the 686A and 686B issue, too.

Greetings,
gerti67

[DDogg]

I would sure be interested to see the SiSoft memory score on those machines. Also I would be very curious if the memory shows as interleaved.

[markrb]

DDogg I don't believe any KT133 motherboards had the interleaving option available in the bios although Via chipsets had the capability.
That's why I suggested the tweaks with Wpcredit. I believe one of them was to activate it.

There is also a seperate patch on viahardware to set it as well.


Mark

[gerti67]

Just for Bullseye and gigabot to compare,

i get the following results in SiSoft Sandra 2002 Pro when doing the Memory Bandwith Benchmark, CPU Arithmetic Benchmark and the File System Benchmark:

- RAM Int Buffered aEMMX/aSSE Bandwith: 1034 MB/s
- RAM Float Buffered aEMMX/aSSE Bandwith: 1018 MB/s

- Dhrystone ALU: 3281 MIPS
- Whetstone FPU: 1657 MFLOPS

- Drive Index IBM (5400rpm NTFS): 16113
- Drive Index Maxtor (5400rpm NTFS): 17398

And i'm using the fastest BIOS settings my BIOS allows me to. (4-way Interleave (@Mark: EPOX always has many BIOS tweaking options other manufacturers don't enable like the memory interleave, that's why i always choose EPOX), Fast Read-Write Turnaround, ...) and never had problems with my RAM and from what i can see from the Sandra reference list my memory is doing quite well compared to other KT133 systems although i was really astonished when seeing those KT266A bandwiths. I think it's time to upgrade in the near future ;)

HTH,
gerti67

[DDogg]

I fairly convinced that the best and cheapest way to speed up CCE is by speeding up memory bandwidth. As Markrb mentioned, for those of you that do not have a way to interleave your memory in bios, check the AMD specific sites and find the patch he mentioned. That will make a small but noticable difference. If you can run CAS 2 instead of 2.5 do so.

Gert, I didn't understand where you got those numbers until I got the 1/4/02 SiSoft update. They have changed stuff quite a bit, especially in the memory area. As you mentioned, the 266A is doing a good job increasing the ram bandwidth, especially when you can run the FSB fast while remaining rock solid. As you all know, CCE is *so* picky about the slightest error induced by overclocking.

My numbers are on my 266A Epox (8kHA+)with FSB@150:

- RAM Int Buffered aEMMX/aSSE Bandwith: 2241 MB/s
- RAM Float Buffered aEMMX/aSSE Bandwith: 2167 MB/s

- Dhrystone ALU: 4374 MIPS
- Whetstone FPU: 2175 MFLOPS

Epox 266A TB 1.4 FSB@150 512 DDR Cas2

My encode speed runs about 1.8 on a 16:9 and about 1.70 on a 4:3 for reference.

[Bullseye]

RAM Int Buffered 935Mb/s
RAM Float Buffered 938MB/s (512MB Crucial CL2 SDR)

Dhrystone ALU 3296 MIPS
Whetstone FPU 1638 MFLOPS

Drive Index 18388 (IBM 7200RPM 30GB)

The memory type seems to make a big difference then although I've got 4 way interleave enabled

[markrb]

gerti67 most 266a based boards have that option in the bios now.
I was about to get an Epox board, but at the time there were real problems with the early 8kHA+ boards. Many DOA's and a couple of real bad bios' in a row. The rush to be one of the first ones out of the gate really hurt Epox, but they have seemed to have fixed everything now. For awhile with the Epox it seemed that either you had the fastest most stable board or it was crazy bad. That scared me away. I ended up buying a Soltek and I am happy enough with it, especially since it was over $40 cheaper then the Asus. There is one problem that drives me nuts with this board. With 2 sticks of memory you can't go above 138Mhz FSB. With 1, the skies the limit. It's not that bad though because I unlocked my CPU and just run the multiplyer higher, but I will never get the kind of memory scores that DDogg gets.

Bullseye Yep that's the difference between DDR and SDR, but remember to that DDogg is running 17/34Mhz over spec giving him much nicer numbers.

I might get 1.7 tops on 16:9 material on my overclocked 1700XP(1644 actual) CPU. So here you can see that even though I have a faster Cpu and the same amount of RAM as DDogg I am a little slower because he runs at a faster bus/memory speed. I believe both DDogg and I only do NTSC video. When I finish my current encode I will run Sandra. Just for reference I have seen many benchmarks comparing the Epox with my Soltek and they tend to be around the same.

Mark

[xrv1138]

just checked my main machine
dry alu 3866
whet fpu 1918
int alu/ram 1761
float fpu 1630

ath 1.4
512 ddr ram cas2
using a kt266 chipset

i get pal conversions of about 1.1
and ntsc of about 1.3

im so annoyed that the 266a came out about 2 days after i bought the mobo (msi kt266)
machine is not overclocked for the best stability ive had in a long time
cant wait to get a faster chipset/cpu
still not fussed about overclocking tho
used to be but now the % gain for each new step vs possible stability issues is not worth it

my old piii 800 @ 1000 was fun overclocking
25% cpu gain, and idles at 1 degree C with peltier cooling
just doesnt seem to be worth it now
that machine just sits in the corner cranking out svcds thanx to dvd2svcd :D

[gigabot]

wow, i didn't check the thread for 1 day and look at all the posts!!
thanx for all your help and suggestions, unfortunately it looks like the memory bw is the problem in sisoft im getting 400mb/s bw. hardly close to ddr.

looks like if im gonna solve this i need to upgrade
lucky i work at a computer store :)

thanx guys!!

[MadMaxx]

@Gigabot

Did you really use Sandra 2002 Pro for the Benchmark?
I use DDR Memory, and I got about 600 mb/s in Sandra 2001 trial and about 1600 in Sandra 2002 pro.

@xrv1138

Could you please give me a more detailed description of your system?
(Operating System, Raid, Temporal Smoother,Resizing Method,Anti noise Filter, VBR/CBR)
I have Exactly the same configuration but get way less frames.

[xrv1138]

madmaxx sure:
winxp
i do use raid on my board 1 drive to each raid channel
(so not real raid but it is faster due dedicated channels)
ts off
simple resize
anti noise 2 (tho im tempted to raise this after listening to gerti67)
4 pass vbr
im happy to give u as much nfo as u want if u have the same mobo and chip
might be a bios setting

[MadMaxx]

Thanx for the fast reply!

I think i used the wrong settings with encoding. Until today i used to have TS activated. Also i used Bicubic resize. That are obviously the worst settings you can make. I will try your settings for the next rip and i am shure that will help.

[xrv1138]

of course you have to judge how the results look in your eyes
overall waiting a few extra hours to produce a result which pleases you is more important then raw speed
although, dvd2svcd is becoming a pretty good benchmarking tool ;)

[gerti67]

Hi MadMaxx,

That are obviously the worst settings you can make

your settings are not really worse, using temporal smoother helps with compression when doing a CBR or 1-pass-VBR encode (though using BicubicResize with it nearly eliminates this positive effect of temporal smoother). But it is not necessary to use temporal smoother when doing a 3/4-pass-VBR encode (Perhaps, just if you want to fit longer movies to 2 CDs only).

And the BicubicResize isn't worse at all, it is just about 20% slower than the new SimpleResize (but renders slightly better visual quality when using 0.00/0.60 for b and c value with it, i think)

But after all, if you are in speeding up the encoding then you're right that these are the "worst" settings (though it could have been worser by using "safe mode" and "IEEE-1180 Reference" and "Deinterlacing" and so on ;))

Greetings,
gerti67

[MadMaxx]

I am doing the vaf file at the moment and the speed looks pretty well :p 1.056 RT but with other programs running.

I don't think i would see a difference in qulality on my TV becouse it is about 7 years old. Maybe I will see the difference on my CRT.

What about you, have you realized a difference in using TS vs no TS?

There is a big thread about encoding speeds and it seems there is no one out there using TS becouse most of them reach speeds from 0.8 to 1.5.

My memory is currently running at 140/280 Mhz, maybe i am able to improve my framecount by going up to 150 Mhz :D .

[gerti67]

Hi MadMaxx,

What about you, have you realized a difference in using TS vs no TS?

as TS is a filter that works on a time axis (detects noise between frames) you can't tell a visual difference in quality directly as it just leaves over a few more bits for CCE to encode (compress) the MPEGs so that for "heavy action scenes" CCE can just increase the bitrate a little bit (what it wouldn't have done before without using TS, i think).

And the speed impact (when using TS) lies in the extra work TS uses to detect this noise. The "Anti noise filter" from CCE is completely different as it softens only some (outstandig) pixels during the encoding of a frame which has nearly no impact on encoding speed.

Hope that helps,
gerti67

[MadMaxx]

Hi Gerti!

Thanx for the help, i think i now have understood the major differences in the encoding methods. I will have a try making some 3/4 pass movies without TS and look if the quality is right for me.

I recently encoded movies always 2/3/4 Pass with TS on and using Bicubic Resize. When I viewed the partly encoded Files, I could hardly realize a difference between the passes 2,3 and 4. Maybe it is ok to make 2 Passes with TS instead of making 4 Passes without.
I sure will find out soon.

[DDogg]

Just a reminder, if you try to push up your FSB speeds you should do a hard test with Prime95 (Like 12-18 hours)as well as the memory tester Markrb has mentioned. If either get an error you should not try using CCE at that FSB speed.

High FSB can cause computational and memory errors although the machine may seem to be doing fine and not crash. Those type of errors give CCE a major case of heartburn.

[markrb]

Promised my Sisoft scores. So here they are run on Sisoft 2002 Pro.

CPU
ALU 4573
FPU 2290

RAM
Int 2089
Float 1982

XP 1700+ overclocked to 1644 (standard is 1463)
Bus at 138Mhz.

Mark

[MadMaxx]

Yep, Dogg you are right, 150 is a bit too fast for the old kt266 chipset. The system seems stable but cce crashes after about an half hour of encoding. I think i will leave FSB at 140 Mhz, this setting seemed to work.

Memory testing is not a bad idea, becouse I am running my CL2,5 memory sticks with CL2 at 140 MHz. I really was suprised that they handled the 150 MHz for a short time.

My encoding speed in CCE is now 1.18 RT (4:3 PAL). The quality looks ok for me. I am not able to tell a movie with TS from one without TS when watching them on my TV Screen.

Here are my Sandra 2002 Values:
CPU
ALU 4065
FPU 2016

RAM
Int 1685
Float 1623

Athlon 1400 @ 1500
FSB @ 140Mhz