Any votes on which way to go? Strickly RC5, OGR, or the homebaked goodness that is a happy medium? Should I care? Did I leave the iron on?
Hoss
Hoss
To RC5 or to OGR, that is the question . . .
- Thread starter Hossenfeffer
- Start date
I ran OGR for a day or two and switched my herd to strictly RC5(manually fixed 20 machines with mapping and srvmgr, doh..). Didn't have any real reason to switch 'em, RC5 just felt better 
. Besides I will catch Gleep within a week now
. Besides I will catch Gleep within a week now Well with 2000+ participants in the Anandtech RC5 team, we seem to be doing ok, so maybe the same for OGR will keep things sweet.
JonB
Platinum Member
Of the two projects, I will choose OGR. While the purpose of OGR is an obscure mathematical concept about spacing things optimally along a line, it has real world uses that could affect us in years to come. RC5 is great fun because of the competition with /. and DPC, but it has no greater purpose.
I think OGR is definetly more useful, but the fact that we are so far into RC5 just makes ya wish they only did one at a time, or you knew the exact dates ahead of time for all these projects
There is a compelling reason to do OGR over RC5 if you are running an AMD K6 CPU. K6s do 10% more OGR work than an MMX Intel, where K6s do ~20% less RC5 than an MMX.
Ken_G6 where are you? I need my opportunity costs!
viz
Eug-
Cel IIs rock at OGR you say? What's yours benchmark?
Ken_G6 where are you? I need my opportunity costs!
viz
Eug-
Cel IIs rock at OGR you say? What's yours benchmark?
BoberFett
Lifer
- Oct 9, 1999
- 37,562
- 9
- 81
Eug
OT, bug I've gotta ask you how you've got that 533 at 897?!?! I notice you've got a very similar setup to mine. I'm running:
Asus P2B rev. 1.02
Cel 533 @ 824 w/G.Orb
V3 3000, stock speed.
What rev. P2B are you using? What slocket? What hard drive? I was planning on trying 896, but I'm a little nervous about corrupting data. I want to make a full system backup before I try to bump it to the 112 FSB.
OT, bug I've gotta ask you how you've got that 533 at 897?!?! I notice you've got a very similar setup to mine. I'm running:
Asus P2B rev. 1.02
Cel 533 @ 824 w/G.Orb
V3 3000, stock speed.
What rev. P2B are you using? What slocket? What hard drive? I was planning on trying 896, but I'm a little nervous about corrupting data. I want to make a full system backup before I try to bump it to the 112 FSB.
Eug
Lifer
- Mar 11, 2000
- 24,048
- 1,679
- 126
Maybe I just have a good chip... it actually runs fine at 920 MHz 1.9 V, but I don't run that high because I don't like the voltage. (And it violates the Eug rule of running at least one step below max speed. ) However, I have run Windows at 960 MHz (or was that 976, on my Soyo?) for testing purposes. Not too stable though.
1.02 eh? Too bad you don't have the 105 and 110 MHz speeds, which I have on my 1.10 P2B mobo. They would come in useful if you don't make it to 112 FSB. I was running 840 MHz for the longest time since I preferred using 1.7 V, but I got the overclocking bug again.
I am using:
Asus P2B 1.10 BIOS 1012
Abit Slotket !!! jumpered to 1.8 V.
Arctic Silver thermal compound
Alpha PEP66 with Sunon fan
3 case fans in an Aopen HX08 case
IBM 34GXP 7200 rpm drive and Quantum Fireball 5400 rpm drive.
Promise Ultra66 controller
V3 2000/180 MHz with two heatsink/fans and Arctic Silver and memory sinks.
Bunch of other stuff
Actually the drives aren't such a big issue, since at 112 FSB, the PCI bus is running at 37+ MHz. (Most, but not all, drives will handle this fine.)
I test with one or two burn-in programs and something else that uses video as well, simultaneously overnight. I also test with 3DMark2000 looped, more for the video though.
Out of interest's sake, what does your setup give you in OGR?
) However, I have run Windows at 960 MHz (or was that 976, on my Soyo?) for testing purposes. Not too stable though. 1.02 eh? Too bad you don't have the 105 and 110 MHz speeds, which I have on my 1.10 P2B mobo. They would come in useful if you don't make it to 112 FSB. I was running 840 MHz for the longest time since I preferred using 1.7 V, but I got the overclocking bug again.
I am using:
Asus P2B 1.10 BIOS 1012
Abit Slotket !!! jumpered to 1.8 V.
Arctic Silver thermal compound
Alpha PEP66 with Sunon fan
3 case fans in an Aopen HX08 case
IBM 34GXP 7200 rpm drive and Quantum Fireball 5400 rpm drive.
Promise Ultra66 controller
V3 2000/180 MHz with two heatsink/fans and Arctic Silver and memory sinks.
Bunch of other stuff
Actually the drives aren't such a big issue, since at 112 FSB, the PCI bus is running at 37+ MHz. (Most, but not all, drives will handle this fine.)
I test with one or two burn-in programs and something else that uses video as well, simultaneously overnight. I also test with 3DMark2000 looped, more for the video though.
Out of interest's sake, what does your setup give you in OGR?
Hez Viz, You should really emphasize that that is a K6 not a K6-2 or K6-III, although it's true for them as well. In fact, preliminary results show that the same CPUs good at Gamma Flux (i.e. more efficient at GF than at RC5) are good at OGR (i.e. more efficient at OGR than RC5).
B.K., Your Cyrix almost works better at RC5 than at OGR.
2288.972 keys/sec/MHz vs 4983.837 nodes/sec/MHz giving a ratio of 4.59
Here are the other ratios that I have so far.
Intel non- MMX from 3.23 to 3.49
AMD 5x86 3.5
AMD K6-2 and K6-III from 3.41 to 3.43
Celeron from 5.04 to 5.10
PIII from 5.10 to 5.16
Intel MMX from 5.12 to 5.18
AMD Athlon 5.26
This is a summary of 26 CPU's that had the long benchmarks run 6 times each and averaged.
B.K., Your Cyrix almost works better at RC5 than at OGR.
2288.972 keys/sec/MHz vs 4983.837 nodes/sec/MHz giving a ratio of 4.59
Here are the other ratios that I have so far.
Intel non- MMX from 3.23 to 3.49
AMD 5x86 3.5
AMD K6-2 and K6-III from 3.41 to 3.43
Celeron from 5.04 to 5.10
PIII from 5.10 to 5.16
Intel MMX from 5.12 to 5.18
AMD Athlon 5.26
This is a summary of 26 CPU's that had the long benchmarks run 6 times each and averaged.
- Dec 11, 1999
- 16,633
- 4,562
- 75
Hey, Viz! Nice to know I'm wanted!
I've been waiting for some good numbers before doing OGR Opportunity Costs. I don't think I'll get around to them anytime soon, since I just (finally) got a summer job. But I think Joe O. may have just done the RC5/OGR opportunity costs; am I right?
BTW, I recently discovered I relatively urgently need to complete a range of primes I had signed up to check with Proth (a semi-distributed project that unfortunately has no teams.) So for the next couple of weeks I'm on a prime sabbadical!
I've been waiting for some good numbers before doing OGR Opportunity Costs. I don't think I'll get around to them anytime soon, since I just (finally) got a summer job. But I think Joe O. may have just done the RC5/OGR opportunity costs; am I right?
BTW, I recently discovered I relatively urgently need to complete a range of primes I had signed up to check with Proth (a semi-distributed project that unfortunately has no teams.) So for the next couple of weeks I'm on a prime sabbadical!
Ken_G6,
You are right, these are "opportunity costs". The lower the number, the fewer RC5 WU's are not done when you are doing OGR stubs.
Eug,I had tried to give a hint by my comment to B.K.
I don't want to tell anyone what to do, and I am not telling anyone what to do. In my opinion CPU's with ratios less than 4.5 ....
No, I'll not finish that sentence. Everyone has to decide for themselves. All I will say is:
The lower the number the better the CPU is at OGR than at RC5.
Conversely, the higher the number the better the CPU is at RC5 than OGR.
You are right, these are "opportunity costs". The lower the number, the fewer RC5 WU's are not done when you are doing OGR stubs.
Eug,I had tried to give a hint by my comment to B.K.
I don't want to tell anyone what to do, and I am not telling anyone what to do. In my opinion CPU's with ratios less than 4.5 ....
No, I'll not finish that sentence. Everyone has to decide for themselves. All I will say is:
The lower the number the better the CPU is at OGR than at RC5.
Conversely, the higher the number the better the CPU is at RC5 than OGR.
Eug
Lifer
- Mar 11, 2000
- 24,048
- 1,679
- 126
Well I now understand, but your numbers seem to have a misplaced decimal point. Plus they don't seem to jive with my Celeron 897, so mileage may vary.
RC5/OGR = x
Therefore, the higher the number, the higher the number the better it is at RC5 as you said.
But on mine (Celeron 897):
2824/6911 = 0.41
If I take my best RC5 and worst OGR benches it's still 0.41. So, on my Intel system it would seem it just happens it's better suited to OGR rather than RC5. The numbers are close on my Celly 457 too. (But, as you say, people can do whatever they damn well please.)
By the way, for the people doing the calculation themselves, you don't have to come up with a nodes/s/MHz number since (obviously) your MHz is gonna be unchanged on your machine. Just take your RC5 long bench score and divide it by your OGR long bench score.
RC5/OGR = x
Therefore, the higher the number, the higher the number the better it is at RC5 as you said.
But on mine (Celeron 897):
2824/6911 = 0.41
If I take my best RC5 and worst OGR benches it's still 0.41. So, on my Intel system it would seem it just happens it's better suited to OGR rather than RC5. The numbers are close on my Celly 457 too. (But, as you say, people can do whatever they damn well please.)
By the way, for the people doing the calculation themselves, you don't have to come up with a nodes/s/MHz number since (obviously) your MHz is gonna be unchanged on your machine. Just take your RC5 long bench score and divide it by your OGR long bench score.
Eug
Lifer
- Mar 11, 2000
- 24,048
- 1,679
- 126
By the way, I wonder what the big discrepancy is.
My Celly 897 at 6911 nodes/s/MHz is on a BX board and has SSE (which I'm assuming is not used at all anyway). FSB is 112 MHz, and memory is Cas 2 112 MHz.
My Celly 457 at 6827 nodes/s/MHz is on a VIA board and has no SSE. FSB is at 83 MHz, and memory is Cas 3 124 MHz.
Weird that PIIIs are running slower. Anybody have an explanation?
My Celly 897 at 6911 nodes/s/MHz is on a BX board and has SSE (which I'm assuming is not used at all anyway). FSB is 112 MHz, and memory is Cas 2 112 MHz.
My Celly 457 at 6827 nodes/s/MHz is on a VIA board and has no SSE. FSB is at 83 MHz, and memory is Cas 3 124 MHz.
Weird that PIIIs are running slower. Anybody have an explanation?
Would be great if we really were "almost" done with the RC5 project. With around 28% of the keyspace checked, there's going to be quite the wait. Overtaking /. is inevitable. ("Resistence is futile" Might be quite the slick move to go like gangbusters on OGR. Chicks dig that.
Is it sick and wrong that I lose an inordinate amount of sleep thinking about this?
Might be quite the slick move to go like gangbusters on OGR. Chicks dig that.Is it sick and wrong that I lose an inordinate amount of sleep thinking about this?
Eug,
Is your Celeron @897 a Celeron II? You make reference to SSE instructions.
You are correct that the MHz will factor out. However, It is a good 'sanity check' on the numbers.
Here is a table with the two Celeron's I have access to and your two:
INTEL CELERON 450 RC5 1,289,532.99 2865.629 0.504
INTEL CELERON 450 OGR 2,557,661.16 5683.691
Celeron 500.. 500 RC5 1,374,192.42 2748.385 0.509
Celeron 500.. 500 OGR 2,698,406.06 5396.812
Celly 457.... 457 OGR 6,827,000.00 14938.731
Eug Celery... 897 RC5 2,824,000.00 3148.272 0.409
Eug Celery... 897 OGR 6,911,000.00 7704.571
Your Celeron 457 does Twice as many OGR node/sec/MHz than your Celeron! What is it's RC5 benchmark?
From another thread here are Celeron benchmarks reported:
Description.......... MHz Nodes/sec Nodes/sec/Mhz
Celeron 300a/450,.... 450 2557661.16 5683.69
Celeron 366@550...... 550 3046955.00 5539.92
Celeron 433@468 Avg : 468 2575154.08 5502.47
Celeron 533A@897..... 897 6162521.22 6870.15
Celeron 300a/450,.... 450 2520000.00 5600.00
Celly 533A@897....... 897 6199089.91 6910.91
Celerons at 552MHz... 552 3,108,542.03 5631.42
Oh yes, I shifted the numbers over 1 position. Why is a long story, for another time.
Could you run long benchmarks 6 times on each machine and e-mail me the log as osieckis@nji.com? I'd like to add them to my table. Thanks.
Is your Celeron @897 a Celeron II? You make reference to SSE instructions.
You are correct that the MHz will factor out. However, It is a good 'sanity check' on the numbers.
Here is a table with the two Celeron's I have access to and your two:
INTEL CELERON 450 RC5 1,289,532.99 2865.629 0.504
INTEL CELERON 450 OGR 2,557,661.16 5683.691
Celeron 500.. 500 RC5 1,374,192.42 2748.385 0.509
Celeron 500.. 500 OGR 2,698,406.06 5396.812
Celly 457.... 457 OGR 6,827,000.00 14938.731
Eug Celery... 897 RC5 2,824,000.00 3148.272 0.409
Eug Celery... 897 OGR 6,911,000.00 7704.571
Your Celeron 457 does Twice as many OGR node/sec/MHz than your Celeron! What is it's RC5 benchmark?
From another thread here are Celeron benchmarks reported:
Description.......... MHz Nodes/sec Nodes/sec/Mhz
Celeron 300a/450,.... 450 2557661.16 5683.69
Celeron 366@550...... 550 3046955.00 5539.92
Celeron 433@468 Avg : 468 2575154.08 5502.47
Celeron 533A@897..... 897 6162521.22 6870.15
Celeron 300a/450,.... 450 2520000.00 5600.00
Celly 533A@897....... 897 6199089.91 6910.91
Celerons at 552MHz... 552 3,108,542.03 5631.42
Oh yes, I shifted the numbers over 1 position. Why is a long story, for another time.
Could you run long benchmarks 6 times on each machine and e-mail me the log as osieckis@nji.com? I'd like to add them to my table. Thanks.
Eug
Lifer
- Mar 11, 2000
- 24,048
- 1,679
- 126
Celly 457.... 457 OGR 6,827,000.00 14938.731
Eug Celery... 897 RC5 2,824,000.00 3148.272 0.409
Eug Celery... 897 OGR 6,911,000.00 7704.571
The 457 gets about 3 Mnodes/s and the 897 gets about 6.2 Mnodes/s.
Those numbers are incorrect. The numbers I had given you were keys/s/MHz for RC5, and nodes/s/MHz for OGR, not the raw scores.
The 457 and 897 are the same ones as in that other thread.
The 457 is a 366 plain Celeron (with L2) and the 897 is a 533A Celeron II (with L2 and SSE).
You will be getting an email soon.
Eug Celery... 897 RC5 2,824,000.00 3148.272 0.409
Eug Celery... 897 OGR 6,911,000.00 7704.571
The 457 gets about 3 Mnodes/s and the 897 gets about 6.2 Mnodes/s.
Those numbers are incorrect. The numbers I had given you were keys/s/MHz for RC5, and nodes/s/MHz for OGR, not the raw scores.
The 457 and 897 are the same ones as in that other thread.
The 457 is a 366 plain Celeron (with L2) and the 897 is a 533A Celeron II (with L2 and SSE).
You will be getting an email soon.
I will be doing both throughout the OGR cycle. I like the thought of kicking some /. butt in OGR and it will be several weeks before I have my Win2k image ready to go on all my lab machines since I will be on vacation, so those 130 (90+ in my 2 labs, and 40 in a small pair of labs that I set up for another department recently) computers will mostly be doing OGR...
and it will be several weeks before I have my Win2k image ready to go on all my lab machines since I will be on vacation, so those 130 (90+ in my 2 labs, and 40 in a small pair of labs that I set up for another department recently) computers will mostly be doing OGR...Eug,
I received your e-mail. Thank You. I think that the key phrase may be "latest client". I don't have results for a Celeron yet, but when I installed 463b on my AMD K6-III, the first results were 30% faster . I'll be able to get one of the Celerons benchmarked tomorrow, as well as a P200MMX, and the K6-III. If these results hold across the board....
Can you benchmark either 460 or 462 on your machines? You can do it from a floppy if you still have the client.
I saw consistent results from 458 thru 462 on all the machines. There has been no announcement of a speedup so I am suprised.
Viz, Which version of the client did you use for the benchmarks that you submitted for the other thread? If 460 or 462 can you redo them with 463?
Remember if you change client versions it will restart from the beginning. So either benchmark in a different directory or on a floppy. When you do change the client try to do it when the % complete is very low so as not to waste cycles.
I received your e-mail. Thank You. I think that the key phrase may be "latest client". I don't have results for a Celeron yet, but when I installed 463b on my AMD K6-III, the first results were 30% faster . I'll be able to get one of the Celerons benchmarked tomorrow, as well as a P200MMX, and the K6-III. If these results hold across the board....
Can you benchmark either 460 or 462 on your machines? You can do it from a floppy if you still have the client.
I saw consistent results from 458 thru 462 on all the machines. There has been no announcement of a speedup so I am suprised.
Viz, Which version of the client did you use for the benchmarks that you submitted for the other thread? If 460 or 462 can you redo them with 463?
Remember if you change client versions it will restart from the beginning. So either benchmark in a different directory or on a floppy. When you do change the client try to do it when the % complete is very low so as not to waste cycles.
- Dec 11, 1999
- 16,633
- 4,562
- 75
I tried this "getting in on the ground floor" thing with DCypher Gamma Flux. I was one of the first participants, with the very first version of the client.
At the beginning it was great being on top. But pretty soon more, faster people (like The Magicman) got involved. Eventually I couldn't even make the top 100 daily anymore.
So I'm not going to do OGR for awhile, until it gets going well. It's fun climbing the stats, but no fun being in a free-fall.
At the beginning it was great being on top. But pretty soon more, faster people (like The Magicman) got involved. Eventually I couldn't even make the top 100 daily anymore.
So I'm not going to do OGR for awhile, until it gets going well. It's fun climbing the stats, but no fun being in a free-fall.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 21K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter