2 CCE Encodes on my Dual Xeon

[Roveer]

I built a Dual Xeon 2.8ghz machine for doing CCE encodes and I wanted to share some of my results.

For the most part I was doing a single CCE Encode and getting 2.8 to 3.0 for the encode. Problem is, when CCE it only uses 40 - 50% of the CPU on a dual machine. I decided to try 2 instances of CCE on 2 movies and I'm getting 2.10 on each (90-98& CPU util in XP). I think I can bump that performance up a bit if I add a few HD's. Right now I have both VOB sets on one HD and the CCE's are both writing to another HD. If I have my VOB's on independant HD's (I have a 2 SCSI channel iwill MB) and have CCE write the information to 2 other HD's I think performance will increase even more. I can hear the drives thrashing quite a bit. All in all I'm pretty impressed. With my water cooled setup I'm barley getting 100 degrees on CPU0 and 78 degrees on CPU1.

I'll take a few pics and post them. I've got a matrox parhelia 3 headed video card driving 2 Samsung 191T's sitting side by side.

Oh, I'm getting these speeds while my SoundBlaster is taking digital input from my Audiotron and playing it on my Klipsch 5.1's All while I'm posting this message to doom9. I'm really happy with this setup.

Oh, I'm using VFAPI right now. Once I get a good script I'm going to give AVISYNTH a try. I've had trouble in the past, but I think my fastest encodes came from that combo.

Roveer

[I8UB4U81]

Oh, didn't you forget to tell us that all this is done while your marvelous system was doing the housework, programming the dishwasher, washing the car and carrying the tax declaration to the post office?
Also I would consider exchanging the two old 191T's by those new IBM panles with 9M pixels in order to increase the encoding speed by 0.1 Can't await for the pics. Have you ever calculated how many original DVD you can buy when just using a "normal" PC? But I'm pretty sure, the CCE stuff is only done during leasuring time of your system, while all the other time it does things a normal PC wouldn't be able to do. And who should do the housework then?


;)

[Roveer]

Naa, I've got other computers for that stuff. I really just like to built them, and when I got the bug for a Dual Xeon, then along came the water cooled part. Was probably the most expensive computer I had ever built, but I hadn't bought any toys in a while so I could justify it.

AVISynth gave me 3.75 on CCE encodes. I'm pretty pleased. Two instances of CCE using AVISynth gives me 2.1 on each, so that's even better. Going to add a few more 18.2 15K scsi's that I have lying around and see if I can get that speed just a little higher.

Just found out I can overclock the Xeon's. Can't wait to see what happens when I get them both running > 3ghz. Should be fun.

Roveer.

[lilhobo]

surely with all the speed and raw power, you would now have plenty of time to post some results.....

if you have a spare server can you host my neverwinter nights module ?? :D

[jmartina]

What version of CCE you are using?

I am asking because only applications that not supporting multiple CPUs
are giving 50% of CPU load on systems like yours (eg. CCE 2.5).

Newer versions of CCE should recognize two CPUs and and CPU load SHOULD BE 100%.

[Roveer]

CCE 2.66

I haven't found a version of CCE that will use 100% of CPU. I'm running 2 instances of CCE right now and getting 2.5 on each on my 2.8ghz dual xeon.

[Kedirekin]

Actually, in my experience, it's not CCE that's the bottleneck. CCE will use both processors if you can feed it data fast enough. The bottleneck is AviSynth, which is not MP enabled.

To test this yourself, create a short HuffyYUV avi file and let CCE encode that. I think you'll see CCE speeds and CPU utilization go up significantly.

Also FYI: with CCE 2.5, I couldn't get two instance of CCE to run right. The speed was impressive, but the results were often scrambled or just black (when I could even get both instances to run to completion - one of the two would often fail at 99%). I think that CCE 2.5 used a shared data memory space, so the two instance would completely trash each other's processing, but that's really just a guess on my part. The might have fixed this after 2.5, but you should check your output rather carefully.

[LB]

Now that we are on the topic of SPEED, maybe you all can answer a question for me. I recently also built a new dual xeon system @ 2 x 2.93ghz. Question is this: In CCE v2.5 I got what I expected, about 2.7-3.0 RT. But with the latest version of CCE v2.67 I get 4.3 RT. Yes, that's not a typo, I honestly get that speed. Does anyone else find the latest version of CCE also signifigantly faster than v2.5? I can't figure it out. However, I still use v2.5 because [1. v2.67 has a funky color issue on the output. Maybe it makes the final encode too dark? I haven't sat down to diagnose it]; [2. and I find that it will crash out more often than the stable v2.5].

[b0b0b0b]

On my p3-800 I get .67

:)

[auenf]

Originally posted by LB
Question is this: In CCE v2.5 I got what I expected, about 2.7-3.0 RT. But with the latest version of CCE v2.67 I get 4.3 RT. Yes, that's not a typo, I honestly get that speed. Does anyone else find the latest version of CCE also signifigantly faster than v2.5?

from memory 2.66+ included SSE2 optimizations which really helped the P4's and P4-XEON's.

didnt think it helped that much tho, but im pretty sure there are other internal optimizations somewhere in between 2.5 and 2.66 anyway ;)

Enf...

[shoarthing]

LB - Hi -

"[2. and I find that it will crash out more often than the stable v2.5]."

- without wishing to sound picky; what do you mean by 'more often'?

- in my [moderate] experience, CCE on a stable SMP platform simply doesn't crash unless, perhaps, you manage to find a bug. Otherwise, any CCE 'crash' can be assumed to be a symptom of an unstable system.

NuMenu4U is IMHO an pretty good stability-test if you're overclocking - gives the same kind of violently changing loads & load-patterns as Juonni Vuorrien's 'CPU Stability Test v4x' of yesteryear.

- as an aside, an optimised dual Barton at 2.25Ghz can be made to encode in the 2.7>3.0 RT range - stably, with all versions of CCE.

[lilhobo]

Originally posted by shoarthing
LB - Hi -

2.7>3.0 RT range - stably, with all versions of CCE.

just to qualify for newbie here....that means its encoding at 3 seconds of footage every seconds right??

[shoarthing]

- exactly as you said: RT [x] = source [time]/one-pass-of-an-encode [time]: modern versions of CCE give a running/cumulative average

[sehh]

Just wanted to add my bit on the SMP issue.

First of all, i'm running a dual with Athlon MP's. I'm also using
64bit RAID0 with 4 ibm ultrastar 10.000rpm U160 drives.

Here are my findings:

1) Avisynth is not multithreaded, no smp support. Thus you get
lower speeds with CCE than what CCE is capable of.

2) CCE 2.66 or newer versions are SMP optimized, thus give
a significant boost from earlier versions (we've seen up to 30%).

3) If you have the hard-disk bandwidth then don't encode one
movie at a time, do two movies at a time! It is more efficient
use of cpu power.

4) If you are using dvd2svcd, then you may run two dvd2svcd instances
at the same time. The only problem is that EclCCE does not support
this method because it does not allow you to run two instances. If
you are not using EclCCE then you won't have any problems.

5) If you got SMP then use SCSI only. IDE is... well bad.

6) RAID makes a difference. If you have it, USE IT. If you don't
have hardware raid, then try software, although don't expect
the same results.

[shoarthing]

sehh - Hi:

[also running dual Athlon/Barton w/ 3x [16x HDD] SCSI 10K rpm arrays on 6x U3W channels of 2x 64-bit/66MHz hosts]]

1 - agree: AVIsynth the limitimg issue for SMP systems
2: - disagree: 2.6x & later optimised very much for SSE2 [irrelevant for Athlons]: 2.5x IMHO every bit as fast using same version AVIsynth [have tested this].
3: - yes
4: - haven't tried this
5: - yes :)
6: - yes :)

BTW: have you tried Mike Lin's superb SMP Seesaw utility to isolate CCE/AVIsynth process on CPU#1; then after a few seconds reset to both CPU's - this gives CPU loads way in the 90%'s & increases speed per run]?

[sehh]

shoarthing, about the CCE speed of newer versions, i've done several
tests with 2.50 and 2.66.1.7, and i've come to the conclusion that
2.66 gives at least 30% increase in speed.

I did my tests on 2.5 hour dvd movies which require 3 cds (svcd).

When encoding a 3 cd movie with CCE 2.50 in just 2 cds, then the
quality drops considerably.

When encoding a 3 cd movie with CCE 2.66 in just 2 cds, then the
quality remains the same as if it was 3 cds. Great difference,
and of high importance to me since i no longer make 3 cd movies!

Now about the speed gain, with CCE 2.50 a 2 hour movie finished
in 6 hours, with CCE 2.66 the same movie finishes in just 5 hours.

PS:
I haven't tried the Seesaw utility, but now that you mention it
i will give it a try :)

[shoarthing]

sehh - Hi - aaah . . I was assuming we were talking DVD-level bitrates. I ran a long bunch of benches on 120-170 minute MPEG2 streams & found v2.5 to be if anything slightly faster [all other settings being equal].

Interesting if the newer versions are more effective at SVCD bitrates/settings.

Hope you find that SMPSeesaw works for you - it's something a bunch of us found out when tweaking for speed.

[sehh]

shoarthing, SMPSeesaw seemed to affect the system in a strange
way and it didn't make sense why it improves performance.

so i tested this method by hand. I changed the affinity to
one cpu, and after a few seconds i changed it back to both cpus.

before that trick CCE used 63% to 70% cpu, after i did the trick
CCE uses 77% to 81% cpu usage.

that is damn strange indeed!

[chipvideo]

I have a dual MP2800 on a K7D Master L. I am getting around 2.65 encode times using the dvd2dvd method with cce. Using avisynth 2 as I heard that 2.5 has bugs. How can I speed up my encode times.

Using 1 gig PC 3200 Registered ECC ram as well. Two WD HDD.

[sehh]

I would suggest you try avisynth 2.5x since that makes a big difference.

[LB]

Originally posted by shoarthing
LB - Hi -

"[2. and I find that it will crash out more often than the stable v2.5]."

- without wishing to sound picky; what do you mean by 'more often'?


I found that when my xeon's were oc'd to 2.93 I was crashing randomly in CCE only. Every other program I ever used at that speed was 100% okay. So CCE is somewhat picky. But, 2.67 crashes just as much. I lowered the speed to 2.8ghz and don't get any more crashes.

[chipvideo]

Ok I am now getting around 2.85 average and as high as 2.90

[TigerLord]

Something is wrong...

I have a 1.53 encode speed , and heres my comp:

PIV 3.0GHZ 800FSB
1GB DDR-400 Dual Channel
WD 120GB 7200RPM 8MB Cache

What am I doing wrong ? I do have other applications running, but theyre not very consuming ...

I am using the DVD2AVI/CCE Method

Please help

[I8UB4U81]

TigerLord: Talking about enconding speed makes ONLY sense if all are using EXACTLY THE SAME working set. In this case it seems that not
even the movie is known, at least I can't find it in the thread.
So don't worry, because so much things takes influence, that
the difference can be explained by ease (CBR/VBR, total amount of
pixel to be encoded, kind of frames/what movie, what software version for frameserver/encoding tool ...). Unless someone define a standardized benchmark here any speed comparison is totally nonsense.
If you've already tested the different software versions and are sure
you don't want to change the encoding parameters (CBR/VBR, cropping, resizing, etc.) than the job is done unless you are willing to change hardware.

LB: CCE is one piece of software that I would call a "real statbility checker". I've made this experince as well, and I recommend to test
an overclocked system with real stressing tools before I call it stable. My exprience is that you can substitute up to 10% from the top speed that seems to be stable under "normal" conditions, if you
plan to use your system for things like video conversion.
My recommendation:

1. Use memtest (without Windows/Linux impact). You MUST run all the
test for some loops without errors. Otherwise you are playing
"russian roulette", how we call it in Germany.
2. In Windows (can not speak for Linux) I use some apps in parallel
to test stability: SuperPi and Prim95/Torture Test is a good
combination. At least I combine this with a stability tester that
is doing additional I/Os on disk and network layer. I you plan to
use the system formultimedia purposes while encoding is going on,
use some sound and graphic card stressing thing in parallel to see
is some IRQ error like to pop up .

cu


:angry: :angry:

[peachjam]

Originally posted by Roveer
I built a Dual Xeon 2.8ghz machine for doing CCE encodes and I wanted to share some of my results.

[...]

I'll take a few pics and post them. I've got a matrox parhelia 3 headed video card driving 2 Samsung 191T's sitting side by side.

[...]

Oh, I'm using VFAPI right now. Once I get a good script I'm going to give AVISYNTH a try. I've had trouble in the past, but I think my fastest encodes came from that combo.

Roveer

Dude! I'm getting 0.8 on my single processor Athlon 900@1100, memory overclocked, too. And I must say I'm more than happy with it. With Avisynth 2.08 it didn't get faster than 0.68. Vfapi sucks, use avisynth 2.52 with a single 3 line script (loadplugin mpeg3decdg.dll, mpeg2source, and converttoyuy2) and please don't start such stupid threads. Personally, I'm not really interested how "badly" your powerplant performs. Read, read, read and learn from previous posts. Hope you won't take it amiss.

Just my 2 euro cents.

[LB]

Originally posted by TigerLord
Something is wrong...

I have a 1.53 encode speed , and heres my comp:

PIV 3.0GHZ 800FSB
1GB DDR-400 Dual Channel
WD 120GB 7200RPM 8MB Cache

What am I doing wrong ? I do have other applications running, but theyre not very consuming ...

I am using the DVD2AVI/CCE Method

Please help

Nothing is wrong. That is the speed you are supposed to be getting. If a dual 2.8 gets 2.8, then a single 3.0 should get around 1.5, which is what you are getting.

[chipvideo]

I am now getting constant 3.15RT now. Overclocked my MP2800's to 2400MHZ each. Using the bicubic resize for dvd's.

Got as high as 3.28RT yesterday with Terminator 3 on a 9 pass.

[chipvideo]

This is pretty off topic, but I just have to ask if anyone would want to sponser my 7 month old child. She has microvillus inclusion disease and is listed in the cota.org site. Her name is Rebekka Haltiner. She has a life span of 2 years and she needs a double transplant(small bowel/liver). She is at a stage 3 liver now. She digests no food. Hooked up to IV 24 Hours/day. Thank you.

[br408408]

LB,
Something IS wrong with TigerLord's system...I have an Abit AI7, P4 2.4c @ 3.0 MHz, and 2 WD Raptors 10,000 RPM HD's in RAID 0, and using DVD2AVI with AVIsynth 2.5 and CCE 2.66 encoding to DVD specs I see RT's of 3.04 to 3.1

I also have an AMD system. With an Abit NF7-S, AMD XP2500 @ 2156 (227 FSB x 9.5, memory in sync) and 2 WD Raptors in RAID 0 and using DVD2AVI with AVIsynth 2.5 and CCE 2.5 encoding to DVD specs I see RT's of 2.45 2.52.

I found the CCE 2.5 runs faster on my AMD (CCE 2.66 ran at about 2.1 0n the AMD) and CCE 2.66 runs faster on my Intel (CCE 2.5 ran at about 2.6 on the Intel)

TigerLord, You and I have about the same systems....something seams way off with your speeds...what bit rate are you encoding to?....what versions of the software are you using?....how about some more info on your system...what mobo, memory brand, are you using?

Bill

[peachjam]

Originally posted by chipvideo
I am now getting constant 3.15RT now. Overclocked my MP2800's to 2400MHZ each. Using the bicubic resize for dvd's.

Got as high as 3.28RT yesterday with Terminator 3 on a 9 pass.

Why do 9 passes? The more passes will not give you better quality only more precise prediction of filesize. Use Robshot do distribute bitrate manually for best quality.

Try encoding The Blue Planet in 9 passes :) 185 mins of water/DVD, 2 audio tracks, menu, etc. and we'll see how well 9 passes serve :)))

[chipvideo]

I figured what the heck. Start when I go to sleep and wake up and it is finished. I have the power why not use it.

[LB]

Originally posted by br408408
LB,
Something IS wrong with TigerLord's system...I have an Abit AI7, P4 2.4c @ 3.0 MHz, and 2 WD Raptors 10,000 RPM HD's in RAID 0, and using DVD2AVI with AVIsynth 2.5 and CCE 2.66 encoding to DVD specs I see RT's of 3.04 to 3.1

Bill

Hrm. I'm sorry but something smells fishy. I have a XEON system at essentially 6GHZ ( Dual 3ghz ). I also use a Cheetah 15k U320 for my boot while my source material is located on a second Cheetah.

In CCE 2.67 I get around 3-RT.
Source is HUFFY YUV2 720x480 w/4pixels L/R black
Output is same - i.e.., 720x480

So with a single CPU you are saying you get the same speed as me, and I just don't see how that can be possible. Also your single AMD speed shouldn't be that high either. So, tell us a few things:

1. Your source material format
- and if it is AVS, what functions you using
2. Your output resolution

[br408408]

My sources are 720x480 VOB's. I will post my AVIsynth script below...it is simple...as no filters are needed. Output size is also 720x480. Everything is Progressive and NTSC. I am not making this up...I have no reason to lie....I started making SVCD's before Doom9 even had a site...it took me 28 hours to encode 2 hours of video using Flask and TMPGEnc 12a on my old Celeron 500...lol :) If you still think something is "fishy" I will send you a screen shot...btw, I am 46 years old...not some kid trying to BS someone just to stir things up...don't have time for that nonsense.

Script:
LoadPlugin("C:\Program Files\AviSynth 2.5\plugins\MPEG2Dec3.dll")
MPEG2Source("C:\TwoTowers1.d2v")
YV12toYUY2(interlaced=false)
AddAudio()

Bill

[LB]

br408408: so you're simply feeding the avisynth script straight into CCE? I use Huffy AVI's, but have found them to be on average at least 40% faster than any .avs I feed into CCE. But yeah, so .avs into cce?

I'll go home and try that script but I'm almost sure that the speed will be even lower because CCE encoding with a Huffy source is the only way to the fastest speeds in CCE.

And yeah, I'm not saying you're lying but I know for a fact something is not correct here. I've been encoding for ~5yrs now, and ~3yr in CCE, and the speeds of 3.0 cannot be attained if you are (encoding TO 720x480) + ( have a single CPU system). With those two factors I know that 3.0 is impossible to attain.

[br408408]

Originally posted by asifanwar
Hi,

System:
Abit IS7-E mobo (Intel 865 Springdale with 800MHz bus)
P4 2.4 o/c to 3.3GHz
2 x 256 Corsair DDR
WinXP SP1

DVD2AVI 1.76
AVIsynth 2.08
CCE 2.66.01.07

Encode Speed: 3.54x

I used to have a 2.4 P4 o/c to 2.62 with a Gigabyte Intel 850 mobo and RAMBUS and was running at 2.8x encode speed. Much happier now of course!

This is from another thread in this forum...I am not the only one getting these speeds on a single CPU system. Like I said before, I am not making this up.

Bill

[br408408]

LB,
I sent you a PM about contacting me for a screen shot

Bill

[DDogg]

This test (http://forum.doom9.org/showthread.php?s=&threadid=66134) should yield valid benchmark comparisons. It would be great to see some of the more powerful machines try it.

[br408408]

Hi, DDogg,

I am going to run the test this weekend. The only thing is, I use 2.66 not 2.67, but that shouldn't be a problem, should it? I also have 2.5...it is not loaded now, but I will install it and try it too. Too bad my AMD rig (Abit NF7-S, XP2500 @ 9.5 X 227)is down now, as 2.5 runs fast on it.

Bill

[DDogg]

Thanks, Bill. Please reply in the test thread if you don't mind. I hope you can load the ecl, but I just don't know. I'll be interested to see your tests and I hope you get the 2500 back up. What happened to it...heh, ya melt it? :)

[Diable]

TigerLord, something is wrong with your system. The specs for my system are pretty close to yours:

3GHz P4
1GB of PC3500 memory(2x512MB)
Albatron PX865PE PRO motherboard(it uses the i865 chipset)
100GB Western Digital hard drive(8MB of cache)

Using this avs file:

LoadPlugin("C:\Program Files\Gordian Knot\mpeg2dec3.dll")
mpeg2source("E:\Work Files\matrix.d2v",idct=5)
ConvertToYUY2()

I got 2.84RT when I encoded The Matrix using CCE 2.67.00.11. Check to see if your memory is running at DDR400 and uses a newer version of CCE and AVIsynth.

[br408408]

Found a place to host my screen shot.

http://www.theforumisdown.com/uploadfiles/1203/CCE.JPG

Edit: Won't let you just click on the link...right click it and "Copy Shortcut" it into your address bar.

[kazam]

OK, some more results. I will try 2.67 and see if I can get the promised 30% increase in speed.

CCE Version: 2.50.01.00
CCE Speed: 2.65

DVD2AVI using Forced Film and Avisynth 2.5.

Terminator 2 UE R1 Chapter 20 with the following scripts:

LoadPlugin("F:\Program\AviSynth 2.5\plugins\MPEG2Dec3dg.dll")
mpeg2source("F:\Rips\test2\test2.d2v")
ConvertToYUY2(interlaced=false)
AddAudio()


; Cinema Craft Encoder SP -- Encoder Control List

[item]
title=Standard.avs
vid_out=1
vaf_out=1
aud_out=0
vid_file0=F:\Rips\test2\Standard.m2v
vid_file1=F:\Rips\test2\Standard.m2v
vaf_file=F:\Rips\test2\Standard.vaf
aud_file=F:\Rips\test2\Standard.mpa
file_focused=0
encode_mode=0
packet_size=4096
timecode=0x0
width=720
height=480
vmode=2
frame_rate_idx=1
cbr_brate=5000
opv_q_factor=40
opv_brate_min=0
opv_brate_max=9800
vbr_brate_avg=5000
vbr_brate_min=0
vbr_brate_max=9000
vbr_bias=30
vbr_pass=9
quality_prec=15
use_filter=0
filter_val=6
seq_endcode=1
dvd=0
half_width=0
half_height=0
fast_mode=0
progressive=1
alternate_scan=0
non_linear=1
top_first=1
lum_level=1
intra_dc_prec=-1
aspect_ratio=3
gop_m=3
gop_nm=4
gop_hdr=12
seq_hdr=1
all_closed_gop=0
fix_gop_length=0
mpeg1=0
mpeg1_cps=0
samples_per_sec=44100
stereo=2
brate_idx=7
crc=1

[file]
name=F:\Rips\test2\Standard.avs
type=0
frame_first=0
frame_last=1695
encode_first=0
encode_last=1695

[kazam]

CCE Version: 2.67.00.10
CCE Speed: 2.76

So, no 30% increase for me.

I guess with some overclocking it would be possible to break the 3.0 barrier quite easily seeing as I'm so close already.


Some suggestions if you're seeing lower than expected speeds:

1. Make sure Dual-Channel is enabled.
2. Enable PAT if your MB supports it.
3. If there's a Turbo mode, turn it on.
4. Try different memory timings. Be careful with stability.
5. Run a RAID configuration if possible since it offloads IO processing from the CPU.
6. Make sure DMA is enabled and functioning if you're not using RAID.

Run Prime95 for a couple of hours in Test mode to make sure your system is stable.

[br408408]

Your speed seems about right....I found that with P4c's clock speed in GHz's is about = to CCE speed in RT. Your clock speed 2.8 ~ CCE speed 2.76. My clock speed 3.0 ~ CCE speed about 3.0. Your P4 2.8c should o/c to 3.0 or more without any loss of stability.

Bill

[kazam]

Some more results:

CCE Version: 2.67.00.10
CCE Speed: 3.04

P4 2.8C @ 3080MHz (14x220MHz).

It seems to me that CCE is a memory bandwidth limited application. I have HT enabled and I can see that I still have cycles to spare. CPU utilization is around 86%.

Bill, are you running your FSB at 1.25 core clock? That would help I guess.

You're right, the rule of thumb seems to be:

If you have a P4 with 800MHz FSB and are encoding progressive NTSC material the CCE 2.67 RT speed should be about equal to your CPU clock in GHz.

Substantially lower than this and you have a problem.

/kazam.

[kazam]

Originally posted by LB
Hrm. I'm sorry but something smells fishy. I have a XEON system at essentially 6GHZ ( Dual 3ghz ). I also use a Cheetah 15k U320 for my boot while my source material is located on a second Cheetah.

In CCE 2.67 I get around 3-RT.
Source is HUFFY YUV2 720x480 w/4pixels L/R black
Output is same - i.e.., 720x480

So with a single CPU you are saying you get the same speed as me, and I just don't see how that can be possible. Also your single AMD speed shouldn't be that high either. So, tell us a few things:

1. Your source material format
- and if it is AVS, what functions you using
2. Your output resolution


LB,

If CCE is memory bandwidth limited, like I think it is, your nice SMP SCSI system brings absolutely nothing to the table. Your two CPUs still have to share that memory bandwidth.

I have seen this behaviour many times in my work. I deal with large computational codes which often are mem bw limited and upgrading your CPU simply does nothing for your throughput. It's just waiting faster for data to process.


/kazam

[br408408]

I am using the 5:4 divider so my FSB is at 250 but my memory is at 200. I have a 2.4c, so I have a mulitplier of 12. My memory will only go to 225-230 so that would only give me a core speed of 2.7 GHz. So the only way I can get the core speed to 3.0 GHz is to use the 5:4 divider. I think you a right about memory bandwidth...I also have an AMD machine (Abit NF7-S with an unlocked XP2500) and my encode speeds went up as I lowered the multiplier and increased the FSB but keeping the core speed about the same. I also lost speed when I ran my memory in single channel. When we all had CPU of 1 GHz or less, it seemed like core speed was all that mattered...now that CPU are running at 3.0 GHz or more, I think that is no longer true. My XP2500 at 227 x 9.5 liked CCE 2.5 and ran it about 2.45-2.5 RT

Bill

EDIT: I found a good AMD system with a high FSB (220 MHz or more) will run CCE 2.5 at CPU speed in GHz x 1.15 ...better than our Intel systems. Now, if we could only get an AMD to run at 2.6-2.7 GHz!!....I know, I know, with water cooling or better.....

[br408408]

kazam,

I think LB thought I was making this stuff up, that's why I posted the screen shot.

Bill

[chipvideo]

I got up to 3.39RT with Titanic.