Printable View
Well ran out of time so I didn't get to run everything I wanted to, did the best i could with available time....
No changes made in Bios, or OS tweaks, same memory, mobo, vid card default clocks
K8NPro No Vmods
9800P default clocks
boxed cooling (air)
512 mb Samsung PC2700
first 1m at default speeds New Castle Core
1m @ Default speeds Clawhammer
1m Clawhammer @ 2200
1m Clawhammer max clocks for this proc on air..
4m clawhammer
4m newcastle
3DM01 Clawhammer Default, 9800P Default..
3DM01 default newcastle, and 9800P
3DM01 clawhammer @2200, default 9800P
Wow... good testing man! That honestly surprised me... I would have figured the CH would've been a lot faster than the NC shows what I know:D
So what do you think now though? You going to keep the CH or the NC?
`s
last one, 3DM01 Newcastle @ 10x200, 9800P Default
i splitted the results from the original thread here:
http://www.xtremesystems.org/forums/...5&pagenumber=1
and made it a sticky here.
looks like its worth to get the 1mb chips now before the 512kb ones replace them.
great job, looks like the 1mb cache is the way to go.
I Would like to see the NewCastle at 10x235 in super pi 4m also.. that would be a good compare.
BUT what I want to know is about the 11x multi. Say you have ram that can handle 250MHz and a chip that'll do 2750MHz. What would be the fastest:
3200+ 1024kB 10x275 5:4
or
3200+ 512kB 11x250 1:1
?
Hmm, I wonder if dual channel ram will make up for the 676 lower 3Dmarks when they're clocked @ the same speeds...
I think it should tho.
DC only made about 100 points difference in 3DMark2001 on my XP-M platform @ 2.0GHz with a 9800 Pro. Hypertransport may provide a difference story, but I'm skeptical.
CPU = 2.0GHz (167*12)
GPU = Sapphire Radeon 9800 Pro 128MB (stock, 1024*768, high performance settings, Catalyst 4.2, Dx9.0b, AGP Aperature 128MB)
RAM = Kreton T-Rex PC3500 (CH-5) 2*256MB [Single Channel]
FSB = 167 (DDR 333)
Timings = 2,3,3,11
3DMark2001SE = 12,957 / 13,002 (12,981 Avg)
CPU = 2.0GHz (167*12)
GPU = Sapphire Radeon 9800 Pro 128MB (stock, 1024*768, high performance settings, Catalyst 4.2, Dx9.0b, AGP Aperature 128MB)
RAM = Kreton T-Rex PC3500 (CH-5) 2*256MB [Dual Channel]
FSB = 167 (DDR 333)
Timings = 2,3,3,11
3DMark2001SE = 13,045 / 13,108 (13,076 Avg)
Interesting, what's the last 1mb chip gonna be a 3400+ or higher?
Ture, but DC will have a much bigger effect on A64 performance than on Athlon XP.. (in a DC athlon XP setup the limiting factor is the processors FSB, but with onboard contorler this bottleneck is removed.) I'd say it should make up for it easilyQuote:
Originally posted by WiCKeD
DC only made about 100 points difference in 3DMark2001 on my XP-M platform @ 2.0GHz with a 9800 Pro. Hypertransport may provide a difference story, but I'm skeptical.
CPU = 2.0GHz (167*12)
GPU = Sapphire Radeon 9800 Pro 128MB (stock, 1024*768, high performance settings, Catalyst 4.2, Dx9.0b, AGP Aperature 128MB)
RAM = Kreton T-Rex PC3500 (CH-5) 2*256MB [Single Channel]
FSB = 167 (DDR 333)
Timings = 2,3,3,11
3DMark2001SE = 12,957 / 13,002 (12,981 Avg)
CPU = 2.0GHz (167*12)
GPU = Sapphire Radeon 9800 Pro 128MB (stock, 1024*768, high performance settings, Catalyst 4.2, Dx9.0b, AGP Aperature 128MB)
RAM = Kreton T-Rex PC3500 (CH-5) 2*256MB [Dual Channel]
FSB = 167 (DDR 333)
Timings = 2,3,3,11
3DMark2001SE = 13,045 / 13,108 (13,076 Avg)
This is very helpfull for alot of people i believe thx for that m8
But i mis one thing ;)
Wht where the max clocks of the 3200+ newcastle ? ;)
No screenshots of it as I no longer have either CPU:eek:
The Newcastle did max stable 2526 @ 1.7V Air cooled, boxed cooling, didn't get a chance to put it under the Prom.
The Clawhammer maxed out 2500 under the prom @ 1.7 :(
How bout some 3dmark01 benchie with both processor's max oc?
it doesnt seem the extra 50 bucks is worth it for the CW, and it looks like most NC's OC more anyways so it makes up for the lost cache
As I said in the post immediately above yours I quoted, I no longer have either processor .Quote:
Originally posted by bluestar88
How bout some 3dmark01 benchie with both processor's max oc?
Peen, the 3200+ was a below average clocker, most aren't that bad, I still prefer, and feel the extra cache is a good thing. Although since it takes an A64 with 1 mb of l2, clocked over 2.5 to do better then my P4 the A64 really isn't worth it for me.
And probably higher considering you have your 3.0C running at 4.5G. You really got unlucky with your clawhammer.
LOL, yep was definately dissapointed with the processor:(Quote:
Originally posted by pkrew
And probably higher considering you have your 3.0C running at 4.5G. You really got unlucky with your clawhammer.
When you say that, what speed is your p4 running? I am realy torn between getting a 3.0c or a refurb 3200 for a friend right now. I want to oc, but can't mod mobos, and a64 mobos seem to realy need modding before they shine.Quote:
Originally posted by Karnivore
As I said in the post immediately above yours I quoted, I no longer have either processor .
Peen, the 3200+ was a below average clocker, most aren't that bad, I still prefer, and feel the extra cache is a good thing. Although since it takes an A64 with 1 mb of l2, clocked over 2.5 to do better then my P4 the A64 really isn't worth it for me.
Quote:
Originally posted by gouda96
When you say that, what speed is your p4 running? I am realy torn between getting a 3.0c or a refurb 3200 for a friend right now. I want to oc, but can't mod mobos, and a64 mobos seem to realy need modding before they shine.
My 3.0C will run 4.5
So you are saying you needed to get the 3200 clawhammer to 2.5 to beat your p4 @ 4.5?
hm...dunno whether to go with a 3200 or a 3.0c. Trying to decide what to put in a machine I am building for a friend. On just air, and without vmods I can't figure out which would be better.
Quote:
Originally posted by gouda96
So you are saying you needed to get the 3200 clawhammer to 2.5 to beat your p4 @ 4.5?
hm...dunno whether to go with a 3200 or a 3.0c. Trying to decide what to put in a machine I am building for a friend. On just air, and without vmods I can't figure out which would be better.
No, thats not it, the 3200+ ran @2.5, it needs to run somewhere over 2.6, probably closer to 2.7 to beat the P4.
So pick up a clawhammer, seems to be the best thing
Hm...what if the nc oc's higher. Will it cost less? Seems like it performs pretty close, hell, comparing max oc's it beat the clawhammer in superpi 4m, and if it is cheaper...
Is that on AIR, if so, then what air?Quote:
Originally posted by Karnivore
My 3.0C will run 4.5
otherwise, it's silly to compare them.
edit, nvm, thought you were comparing the newcastle and the P4. My bad.
Quote:
Originally posted by boshi
Is that on AIR, if so, then what air?
otherwise, it's silly to compare them.
LOL, maybe you should make sure you know ALL the facts, and know EXACTLY WHAT I compared before making such a irrelevant, and ignorant statement..:rolleyes:
Do you even know what I compared, what cooling, and why?
Karnivore, I'm also interested in the difference between an overclocked P4(w/HT) and an overclocked A64. Considering they're similarly priced, with moderate watercooling and unmodded motherboards, whichis going to be better?
An overclocked 3GHz Northwood or An overclocked 3200+ Clawhammer?
Karnivore, your results seem to stress the importance of a large L2 cache. I am impressed that you seem to be getting a +5% increase in 3dmark performance by just an increase in L2 cache size.
I was wondering: at what speeds did you run the memory? Was it always at PC2700?
If so then the CPU will be more memory-starved than in a PC3200 config and will therefore very much like the extra cache memory.
Also, the CPU bus speeds differ, but I don't know if that is such a great difference. RAM however I am not so sure. Perhaps, once you get new CPU's, it's possible to check both CPU's at both the same hypertransport bus and clock speeds?
Does anyone know when the newcastles will show up in america? More speciffically newegg?
Quote:
Originally posted by [-Stash-]
Karnivore, I'm also interested in the difference between an overclocked P4(w/HT) and an overclocked A64.
Going to try and keep this thread on track, if you start a new thread and ask the same question I'm sure you can get some answers.
Memory used was exceptional Samsung PC2700, the single 512 stick easily did 1:1 @2.5,2,2,5 at all HTT speeds tested. So memory was not an issue.Quote:
Originally posted by mr_a
I was wondering: at what speeds did you run the memory? Was it always at PC2700?
Also, the CPU bus speeds differ, but I don't know if that is such a great difference. RAM however I am not so sure. Perhaps, once you get new CPU's, it's possible to check both CPU's at both the same hypertransport bus and clock speeds?
If you looks at my results a bit closer you will notice the processors were tested @ same HTT.
This was just an informal, and VERY time limited set of benchies I ran, mainly to settle my own curiousity, and posted results here because I figured others may have the same questions.
Gouda, I'm in the US, and so was the newcastle I triedQuote:
Originally posted by gouda96
Does anyone know when the newcastles will show up in america? More speciffically newegg?
:eek: More specifically, I would imagine Newegg has some Newcastles now, but probably wish to move the older stock. None of my local stores have any 1 meg L2 3200+ any more.
Realy? Newegg needs to get with the program. What kind of price difference do you see locally?
Quote:
Originally posted by gouda96
Realy? Newegg needs to get with the program. What kind of price difference do you see locally?
3200+ was $299.00, since the Newcastle hit the shelves the price went up to $319.00
Hmm... Did you do any memory benches with the newcastle? Since it's a CG stepping, it should default to 2T command rate compared to the C0's 1T.
Just what you see here, anything else i had was gone with my last format...Quote:
Originally posted by HKPolice
Hmm... Did you do any memory benches with the newcastle? Since it's a CG stepping, it should default to 2T command rate compared to the C0's 1T.
Hmm, do you still have the newcastle now? Do a sandra bench @ stock, Command rate 2T drops it by about 200-300MBQuote:
Originally posted by Karnivore
Just what you see here, anything else i had was gone with my last format...
Quote:
Originally posted by HKPolice
Hmm, do you still have the newcastle now? Do a sandra bench @ stock, Command rate 2T drops it by about 200-300MB
Nope sorry don't have it anymore.
NewCastles are finally at newegg...but they are the exact same price as the cw's. That stinks. I think I will be going for a 3200 cw and hope for a cg.
so the 1mb cache is better? i hope i can get it to at least 2400mhz when i get mine on a thermalright heatsink ;o
Quote:
Originally posted by tylerhskate
so the 1mb cache is better? i hope i can get it to at least 2400mhz when i get mine on a thermalright heatsink ;o
Clock for clock yes, however I would expect as the Newcastle core matures the benefit of 1mb cache will be no more, If the CW does 2.4 and the NC does 2.7, NC is going to be the winner, and since it seems the 1mb cache is being done away with so won't have much choice in the matter....
Well at least in 754...
Those score are way low for those systems, a 3000+(512k cache)/9800pro gets 19.5k in 3dmark01
Quote:
Originally posted by yuri
Those score are way low for those systems, a 3000+(512k cache)/9800pro gets 19.5k in 3dmark01
Did you miss the part regarding DEFAULT clocks on the 9800P, no tweaks or optimizations, no mention of RAM timing, Not a word on DirectX version, or Drivers? I wasn't trying for the biggest 3dM score, rather a simple head to head compare of exactly what the difference between the processors was...
Obviously the system could have been optimized and WOULD have scored much higher....
Did a Newcastle vs. Clawhammer test som weeks ago myself (original thread). Used the Hexus PiFast test:
Code:Oppsett:
Athlon 64 3200+ "Newcastle" (2,2GHz, 512kB L2 cache)
Athlon 64 3400+ "Clawhammer" (2,2GHz 1024kB L2 cache)
Gigabyte GA-K8NS (bios: F3)
OCZ EB PC3200 512MB (2,5-3-2-8)
ATi Radeon 9700 Pro (324MHz/310MHz)
Windows 2000 Pro SP4
DirectX 9.0b
Catalyst 4.6
PiFast:
| Tid
3200+ |------
10*200 | 62,66
11*200 | 59,33
10*220 | 57,09
10*230 | 54,50
3400+ |
10*200 | 61,27
11*200 | 56,94
10*220 | 55,94
3DMark2001:
| Car Car Drago Drago Lobby Lobby Nature
3200+ | Sum low high low high low high
1T |--------------------------------------------------------
10*200 | 18082 261,0 91,5 322,6 170,4 246,5 112,1 115,1
11*200 | 18560 269,2 100,1 321,9 170,3 258,4 117,7 115,2
10*220 | 18698 257,0 101,5 323,0 170,3 269,3 123,1 115,4
2T |
11*200 | 18171 263,7 94,8 322,3 169,8 248,1 111,9 115,0
3400+ |
10*200 | 18456 256,8 94,7 322,4 170,0 262,6 121,5 115,7
11*200 | 19044 275,8 101,8 322,5 170,3 274,2 128,0 115,8
10*220 | 19238 268,1 105,8 323,6 170,0 282,1 133,3 116,0
One problem if you test with PC2700 RAM is that the RAM will run at 157MHz (/14 of CPU) on the 3200+ NC 2,2GHz and at 167MHz (/12 of CPU) on the 3200+ CH 2,0GHz.
Quote:
Originally posted by Kamerat
One problem if you test with PC2700 RAM is that the RAM will run at 157MHz (/14 of CPU) on the 3200+ NC 2,2GHz and at 167MHz (/12 of CPU) on the 3200+ CH 2,0GHz.
Mem was 1:1 all tests...
OK then... :)
Karnivore, are the fastest speeds you tested the max oc's for the chips?
Quote:
Originally posted by gouda96
Karnivore, are the fastest speeds you tested the max oc's for the chips?
Quote:
Originally posted by Karnivore The Newcastle did max stable 2526 @ 1.7V Air cooled, boxed cooling, didn't get a chance to put it under the Prom.
The Clawhammer maxed out 2500 under the prom @ 1.7V
The clawhammer was pretty much a DUD of a clocker, maxed about 2250-2300 on boxed cooling. Would have really liked to have more time with the processors, including putting the NC under the Prommie..
Oh god...I forgot who I was talking too here because you changed your avatar lol. I think I asked that same question or something like it before.
Quote:
Originally posted by gouda96
Oh god...I forgot who I was talking too here because you changed your avatar lol. I think I asked that same question or something like it before.
No matter, only dumb question is the one unasked, or some other Confucious saying:D
So if you were going to do gaming, and not oc either of them which one would you go with?
You wouldn't notice any difference in gaming....Quote:
Originally posted by Riddic
So if you were going to do gaming, and not oc either of them which one would you go with?
Which one runs hotter? I've been playing with my clawhammer and newcastle both 3200+. At default speeds the clawhammer runs lots cooler than the newcastle.
Which one performs better at max oc, and what are the respective max oc's?
The newcastle beats the clawhammer by a little bit. The newcastle 0416 rpmw I have maxs out with watercooling at 2600mhz. The clawhammer maxes out 2380 around there. Performance is about the same. However the clawhammer is cooler. I had a delta of 6c full load and idle with default speed and volts fo clawhammer. I have a delta of 8-10c for newcastle default speeds and volts. One was 2000mhz clawhammer and the other 2200mhz newcastle. Both 3200+Quote:
Originally posted by gouda96
Which one performs better at max oc, and what are the respective max oc's?
I tested clawhammer against newcastle too, the biggest advantage of the NC is that they normally clock higher. From here to there, I needed 100-150MHz more with the nc to get the same frames in games as with the ch.
But: The nc clocked ~ 250MHz higher than the ch, so I kept that one.
I wonder how high they would max out under a prommy. I wonder if the difference in clock would stay the same, increase, or decrease...If only there was a fry's here in Georgia for me to be able to test and return processors.
do you think that EPOX 8KDA3J + Newcastle 3000+ & TT VENUS 12 would give me like 9x295 ? :)
The mem is 512MB Hynix d43 ran fine @ 295Mhz on friend p4
So Both of the 3200+'s are good for gaming on stock? Or should i spend an extra $100 and get the 3400+? And what Ram would be good with it? 1gb(512x2)
Probably possible, but that would be a very good oc'er. I would think more like 2.5 would be a realistic goal depending on temps.Quote:
Originally posted by pots4t4j4
do you think that EPOX 8KDA3J + Newcastle 3000+ & TT VENUS 12 would give me like 9x295 ? :)
The mem is 512MB Hynix d43 ran fine @ 295Mhz on friend p4
Jdeazy, my friend's machine plays D3 quite nicely stock with a 9700np, but this isn't a forum that usually discusses stock speeds, infact saying you are running at stock speeds is usually regarded as blasphemy and shouldn't be done :D If you are going to run stock go for the 1mb cache nc...but don't run it stock.
Hmm~ so would the smart thing to do be to buy a 3200+ newcastle and oc it as far as i can? Im waiting on the DFI board so i can upgrade
Based on what coldflare and radical have said, and some oc'ing #'s from others, it seems that the nc's can often oc high enough to outperform the cw's on air. I have based that on a fairly limited number of comparisons, and I have no idea how they would compare with extreme cooling. Seems like a good nc beats a bad cw, and a good cw beats a bad nc, and good v good seems about equal. Basically the difference is very small and both chips are quite good. For the budget minded a 3000 nc would be a good choice.
It may seem that way, but things are rarely as they seem.
My 3200+ ClawHammer runs 2663MHz 3D bench stable. Overclock ANY Newcastle to 2900MHz and try to beat my 3D scores. You'd have a tough time, even with the 3800+ which represents the fastest NC to date.:D
There are some things that a DC Newcastle can compete with the S754 ClawHammer in, but 3D isn't one of those things.;)
Realy, so the ch's performance advantage increases the higher you go? So that would mean a prommy I would be better off with a ch than an nc?
What kind of cooling and volts are you using to get 2663 with your proc, and what is the batch code?
ive had two claw hammers so far and they both maxed at around 23/2400 on air and one of them didnt gain me one single mhz under my mach1!!! then again it was with the old an50r 150 series but still.... i just bought and sold my last 3200 ch within 3days flat never again! either a new 939 chip or anf FX but the 754 platforms dont do nothing for me im not impressed at all. i dont know what it is with those chips maybe its the chipsets who knows but if your buying these things to oc then i hope you will be happy with 300/500mhz oc cause thats about it unless you get a lucky one or have a -100c cascade and live next door to oppainter :D
peace perc,
If only they would open up a damn fry's in Atl I could do my own ch vs. nw testing and decide which I want for no cost.
Is it more likely to get a fairly better CPU that oc well with either 3400 or 3700? I guess bought are CG but the 3700 always comes with 1mb cache. Witch is to prefere?
Thanks!
Hi guys,
I don't quite understand the comparison here...
What was compared here was not only the diference between having 1mb or 512kB of cache, but also the 1T, 2T timming...
So, the only conclusion that I took from all those comparisons is that a Clawhammer @ 2T is almost the same as a Newcastle @ 1T.
Now, what I would REALLY like to see would be the comparisons regarding ONLY the cache.
Clawhammer @ 1T compared with Newcastle @ 1T
Hasta!
Quote:
Originally posted by Sir Lancelot
Hi guys,
I don't quite understand the comparison here...
What was compared here was not only the diference between having 1mb or 512kB of cache, but also the 1T, 2T timming...
So, the only conclusion that I took from all those comparisons is that a Clawhammer @ 2T is almost the same as a Newcastle @ 1T.
Now, what I would REALLY like to see would be the comparisons regarding ONLY the cache.
Clawhammer @ 1T compared with Newcastle @ 1T
Hasta!
Offered what I could at the time, this was when the NC core was first released @ 3200+ speeds, check the date of the first post... the 2T/1T issue was not common knowledge..
FEEL FREE TO RETEST.........
ok... someone lend me 3.4 Newcastle :P
Have CG clawhammer here :)
Quote:
Originally posted by Sir Lancelot
Hi guys,
I don't quite understand the comparison here...
What was compared here was not only the diference between having 1mb or 512kB of cache, but also the 1T, 2T timming...
So, the only conclusion that I took from all those comparisons is that a Clawhammer @ 2T is almost the same as a Newcastle @ 1T.
Now, what I would REALLY like to see would be the comparisons regarding ONLY the cache.
Clawhammer @ 1T compared with Newcastle @ 1T
Hasta!
Hi karnivore,
Nice work:)
Take a look at this which seems to verify all your work. Not only does the 3200NC beat the 3200CH, the 3400NC beats both 3500 939 and 3400CH.
http://hardware.fr/articles/496/page4.html
Zebo, did you just reverse ch and nc? I don't speak that language (french?)?
yeah did you mean to say that it was the 3200 and 3400 CH that meat out the others? doesnt make sense that you say the article confirms his findings when your saying the 3200and 3400NC beats them out??
perc,
:D
3200 newcastle beats a 3200 clawhammer in every benchmark.
3400 newcastle beats a 3400 clawhammer and a 3500 939 in every benchmark
just like karnivore showed the 3200 newcaste beat the 3200 clawhammer in superpi. 3dmark trash no good, realworld test only for consumer.
In any case, it'll be interesting to see where I can go with my DFI Lanparty UT and NC 3200+ that I'm getting tomorrow together with my current Corsair (bh5) RAM.
edit: the cpu in my sig goes past 2.6ghz.. replacing stock cooling to get better performance...
i dont believe that. i need to go back to page one of this thread and reread cause i just dont believe that a 3200NC at lets say 2400mhz is going to beat the 3200CH at 2400mhz.... you got any links to the benchmarks you speak of? would deff like to see how a NC is beating the CH at the same speeds with less cache? i may be wrong but i just dont see how....Quote:
Originally Posted by Zebo
peace perc,
The "W" will beat THEM ALL - you KNOW what Im talking about :D
Quote:
Originally Posted by Perc
No need to go back Perc...
Clock for Clock the Newcastle NEVER HAS, Nor NEVER WILL, beat its big brother... The best it can hope for is to come close, and beat in SOME benches with its additional clock speed.
The only real advantage to the NC core is it typically clocks higher. The additional clock speed may or may not make up for the added cache of the hammer. Depends how much better it is, and what your using to judge the difference. Some applications 200mhz is enough, others its not.
Perc,
You are correct, claw@2ghz will beat a NC@2Ghz everytime, but your assumptions do not match what I wrote. Please take time to re-read.
We got ourselves a battle to see who is right... :D :toast: