AMD Newark 3700+ @ 3105.1MHz Air-Cooled

[ozzimark]

All right, I just got this in today, and ran some tests. First are screens, the next post will be benches comparing my 3200+ clawhammer to my 3700+ newark with both at 2.0ghz, same settings for ram.

I tested the msi k8n neo platinum first. it booted with bios 1.6 with no problem. was at 2.4ghz, and 1.36v in the bios, which makes me think it defaulted to 1.4v instead of the 1.35v it should have been. One quirk to be aware of: I didn't see any option for Command Rate in the bios, and it defaults to 2T. once in windows, i switched to 1T without a problem and managed to run windows memtest for a while until i decided 1T wasn't about to kill anything

next up, then main rig's mobo, the dfi lanparty ut-nf3. it booted at 200x4 and .9v at first. i manually set it to 200x12 and 1.35v. curiously, i got a raid detect screen before getting into memtest. slightly confused, i went back to the bios and loaded optimized defaults, saved, then went back in and set everything i needed to set. it works!

below i have some screens attached of the core in operation, and in a few minutes i should have a post with the comparison between clawhammer and newark
(i especially like the "venice" label given by CBID!)


also, Vandi423 deserves recognition, as if it wasn't for him, this post wouldn't be here.



EDIT!
so that you guys don't have to dig through the thread, i'll keep this screen updated with my max suicide shot

[ozzimark]

all right, the benches. everything was done at 200x10 with the ram at 1:1, 2-2-2-10 timings (basically, the same exact settings in the bios) and with no programs open aside from what i was using to bench. you will see that the newark is consistantly faster, though a very small amount, which goes with all the other testing done on the revE cores

first up, we have the everest memory benches. i took the average of three runs. i can't quite figure out why read and latency are better, but write went down a bit.




then, 3dmark05. i ran the bench twice and recorded the score the second time.


a very old a ghetto benchmark called Final Reality.. i find it amusing. reminds me of doom2 somewhat



and lastly, something more real-world, converting a set from DJ Pi from mp3 to ogg vorbis. the song is 1:13:16 long

[tristancarton]

nice work ozzi. figured the stock settings would be about the same as a claw. now onward ho to the overclocking ones. strange the chip comes up in cbid as a venice core with a 1mb cache. also how are temps?

[ozzimark]

temps are curious actually, at stock, it's about 4-5ºc lower than my clawhammer at stock (2.4ghz, 1.35v vs. 2.0ghz, 1.5v)

[Vandi423]

Nice! Not worth getting a Clawhammer anymore even if you plan on keeping it at stock.

[computersmsa]

Newark = 512 Kb
1 Mb = Lancaster

[karelke]

Newark = 512 Kb
1 Mb = Lancaster

Oakville = 512kb (D0)
Newark & Lancaster = 1MB (E3)

[ozzimark]

yes, newark is basically a san diego in 754 clothing.

as an update, i'm having the same problem the guys with turion have with the ram.. anything that's not in slot 1 = suckage.
slot 2 goes up to ~210mhz
slot 3 won't even boot


DFI! GIVE ME A NEW BIOS

as a positive note, i can do 2.55ghz at 1.35v with no signs of slowing down

[iddqd]

yes, newark is basically a san diego in 754 clothing.

as an update, i'm having the same problem the guys with turion have with the ram.. anything that's not in slot 1 = suckage.
slot 2 goes up to ~210mhz
slot 3 won't even boot


DFI! GIVE ME A NEW BIOS

as a positive note, i can do 2.55ghz at 1.35v with no signs of slowing down

Hey, are you using official DFI bios, or Oscar Wu's (jess1313) beta bioses? Those work a lot better with mobiles. The clawhammer mobiles, at least.

[ozzimark]

right now, beta 4/15. i will use the modded version of it as soon as i can get it

somehow those ram issues fixed themselves thoughl. i think i missed something critical that was preventing me from getting my ram too much further

[computersmsa]

Oakville = 512kb (D0)
Newark & Lancaster = 1MB (E3)

No

Lancaster = 1 Mb (E4) -> San Diego on S754
Newark = 512Kb (E3) -> Venice on S754
Oakville = 512Ko (D0) -> Winchester on S754

[conrad.maranan]

Newark = 512Kb (E0) -> Venice on S754

Why then does ozzimark's chip have 1MB of L2 cache?

[frostedflakes]

Who cares? Let's see how it OC's.

[Gautam]

Who cares? Let's see how it OC's.

Agreed

[frostedflakes]

Looking good so far. Looks more like what the desktop 939s are getting, and not our results with Turion64.

[ozzimark]

yeah, it seems to top out at 2.64ghz with 1.35v. testing at 1.4v now

[WeStSiDePLaYa]

wow, getting some nice results there. im going to have to look at picking up a new chip soon. so are newcastles being phased out?

will new 2800+ and 3200+ chips be lancaster or newark???

[ozzimark]

lancaster = Turion

they should be replacing the entire mobile line with newarks except for the 2800+, it's just a matter of finding one from a store that has gone through all of it's clawhammers



edit: as a side note, i'm shocked by the transitions between stable and unstable.
2705mhz, 1.4v = prime95 stable for 20min, didn't test longer
2714mhz, 1.4v = bsod

i'll get some screens when i get to a speed i'm happy with

[WeStSiDePLaYa]

lancaster = Turion

they should be replacing the entire mobile line with newarks except for the 2800+, it's just a matter of finding one from a store that has gone through all of it's clawhammers



edit: as a side note, i'm shocked by the transitions between stable and unstable.
2705mhz, 1.4v = prime95 stable for 20min, didn't test longer
2714mhz, 1.4v = bsod

i'll get some screens when i get to a speed i'm happy with

so what cores will the regular PIB chips have?

[frostedflakes]

Nice. 2.8GHz at 1.4v would've been nice, but I'm probably asking too much from these poor chips. Hopefully I can get my mits on a 3400+ soon.

[ozzimark]

2.83ghz is good with 1.5v though.. still going well with voltage
though it appears that it needs to be backed down just a tad, or it's having issues with my ram. a new bios with better support for these micron chips would be reaaaaally nice, as running a divider gives really weird results

[tristancarton]

thanks for keeping us updated. seems this is my next chip. 2.83ghz 1.5v. drool.

2.83ghz is good with 1.5v though.. still going well with

[Vandi423]

2.83ghz is good with 1.5v though.. still going well with voltage

hehe

[ozzimark]

i think you should vandi. i know of a few people who are considering them over at [H].. but probably not this early in the game. i'll be back in a while with results with 1.55v

[Vandi423]

Damn, All I need are 100 posts and we could do it here. We'll figure something out this weekend.

Good choice ozzimark! I should have bought one of these and undervolted it.