i dont know tbh, i have always been under the impression that the chipset pretty much treats 1 512mb stick like 2 256mb sticks in the same slot (unless its a dual channel memory configuration :P)
Printable View
i dont know tbh, i have always been under the impression that the chipset pretty much treats 1 512mb stick like 2 256mb sticks in the same slot (unless its a dual channel memory configuration :P)
Wow, amazing results. I'm really looking forward to the water results as well.
Only 7 more days... plus shipping :)
Quote:
Originally Posted by saaya
Thats exactly what it does, is a per channel thing, slot 1 & 2 together become 1 channel that is why you can run dual channel with 2x256mb and 1x512mb (2x256mb on the first channel = 512mb, slot 1 & 2, and 1x512mb en chann2l 2, 3rd slot)
It depends how the traces are laid on the memory stick.Quote:
Originally Posted by saaya
2x256 DS can have the traces to each side so its actually still only "1 bank" per stick or could be done so its "2 banks" per stick... the number of chips on each side doesnt mean much.
With 16 chips per stick you've got no choice.
The 940pin chipset supports well doesnt suport, but will run 4x512 at 1t command at ddr400. All the 940 ads really is an extra pin for ground termination and an extra few wait states If I remember so I can't see it being all that hard to adapt it to 939.. But then again for some strange reason maybe its impossible, lol.
Maybe its the 940pin's thick arse PCB that can help with this as it uses a 6layer the 939 on a 4 layer. maybe its a combo of design and low noise that can allow the 940 to do it. "confused"
the extra wait states are through the reg memory afaik, right?Quote:
Originally Posted by sandybeach
it works pretty much like adding an extra cycle of latency, but with memory interleaving the performence impact is pretty small it seems, as the first 940 fx cpus with reg memory scores almost the same if not better than 939 unreg memory systems at the same settings.
and stevil, are there really ram makers that put 8 chips on a 256mb stick and wire it like its 2 banks? wth are they smoking? :lol:
i wonder what micron thougt as well though when they made this 256mb stick with 4 memory chips on one side i have here :stick:
bad news btw, pc ice hurt his back trying to lift his cascade (dont try that at home kids! :D) and he has probs going on with the testing it seems since his back hurts a lot.
and he said he has the water and air results on a raid array he has problems accessing now but will post the results asap.
lets all wish him to get well again soon :toast:
Ive got a few vicoden he can have if it helps testing? ;) lol
as for 940 yeah the wait states and so on don't seem to hurt performance at all. It was more the HT sped gain from 939 over 940 and that the newer chipset nf3-250 - nf4 where at the time all 939.
I would like to see an nf4 Tyan with a single 2.6ghz 252 opty at 2-2-2-5-1t and twekaed with a64 tweaker and then compared side by side an fx55+dfi nf4 at same settings ;)
ouch that gotta be heavy such a cascade lol :)
OPP wasn't there to spot him? :hehe:
Man that sucks. No one knows back pain more than me. I've just been recovering from recent fusion surgery which was the 6th surgery I've had on mine in the past 2 years. I'm feeling your pain Pc Ice. And I also have some very good narcotics if you need them. :toast: Feel better
lol, no drug trading on xs please :lol:
yeah, i also have probs with my back a lot, get well again soon dave! :toast:
Holy Mackrel I dont think I want a cascade if its that heavy. Oh well Patience is all I have anymore. Take care of that back ICE!
pc ice told me he could bench with 2.8ghz and 1.4v (stock) on watercooling, so sounds like good news to me! :D
cant wait for the results! :D
is that prime stable?
many 3500+ clawhammers are able to do 2.8 on water with a bit more voltage prime stable, some can even do it on air..
I hope your back starts to feel better soon pc ice. Take it easy as your health is more important than some oc'ing although some may beg to differ.
stock volts + water @ 2.8Ghz = :slobber: Once you get better keep up the great work.
i dont think pc ice wasted the few days he could play with this cpu running prime to see how stable an oc was ;)Quote:
Originally Posted by S0nic
and this is the first time i hear that many clawhammers do 2.8ghz stable on water... i have seen SOME few, but its def not normal afaik.
and again, it did 2.8ghz stable for all benches with STOCK vcore, so im pretty sure speeds above 2.8ghz can be obtained prime stable with more vcore.
this is great news for me, now all i need is a preorder button
no screenshots???
Very nice
:woot: still on stock voltage
keep going, dave....
Dave... Did you have a chance to see what was the max 3D or SuperPI benchable for 3500+ with max voltage neccesary? Did upping voltage give any gains? Were the 4x512MB sticks all same brand-speed rating TCCD?
odd so it didnt respond to higher volts under water cooling. Bummer, so its either stock volts or nothing.
Now that is what I wanted to hear. :DQuote:
Originally Posted by pc ice
woah..let me get this straight...no gains at all from voltage?
that's exactly what I was affraid to hear :(Quote:
Originally Posted by pc ice
2.8GHz is not too bad but unfortunately, not great either :( there are many 3000+ Winnies that are benchable on good air/water around 2.8GHz...
also, SSE3 extensions are not really that usefull at this very moment, besides the fact that SuperPi could be run with a patch now :D
it might seem that if one has 2.8GHz 3000+ Winnie, then be very carefull and do not sell it before you actually get Venice and test it...
~$300 for 3500+ Venice vs ~$150-220 for 3000+/3200+ Winnie..... hmmm
EDIT: I know, I know... latest Winnies are crap... :D but I had those that already have great clocking Winnies...
thanks for testing, dave :toast:
Will there be more "new's on this" , in this thread ? Any Ideal what a 3200 will go for ?? ThanksQuote:
Originally Posted by bachus_anonym
:)
well i have orderd a 3500+ Venice and i'll get mine at 050426 so it's just waiting time
Where did you order one from ? Do they have preorder for san diegos? anyone know where i could order one?Quote:
Originally Posted by chefnr1
on this webshop in sweden, 4000+ is san diego but here is two links one for 3500+venice and one for 4000+
http://www.sbs.o.se/default.asp?artnr=ADA3500BPBOX
http://www.sbs.o.se/default.asp?artnr=ADA4000BNBOX
they have from 3000+ ->4000+
sweden? I think Ill wai ttill they get here I hoope it isnt too much longer for them to get here than there.Quote:
Originally Posted by chefnr1
i think it's the same date in US as in swe:)
thing is they dont seems liek they scale well with more voltage (:
i hate to say it but i knew that everyone would be disapointed, they arnt that better than what we got atm.
Got any air results?
I sure hope they respond to voltage while 2.8ghz on water is good but not great, i wanna see 3ghz on water.
A little disappointed now but hope its just afew that behave this way
I like you're attitude. Stay positive. Under the same circumstances I'm sure some members would be wondering why God hates them so much and talking to their local suicide prevention hotline workers about how they think they'll never reach 3ghz. :hehe:Quote:
Originally Posted by pc ice
Great stuff BTW. Keep it coming. I'm going to wait at least a month to let the dust settle from the general retail channel release, and then decide if I'm buying in. 2gig (4x512mb) 1T for EQ2 would be too tempting to pass on.
So the 3500+ doesn't scale and the 3800+ does according to what pc_ice and s7 have said.Quote:
Originally Posted by s7e9h3n
I think PC_Ice is also saying we shouldn't be drawing any conclusions froma couple CPUs. This is just a preview. Who knows how things will really play out. A good picture should start to emerge when the early adopters (and the guys who shown a lot of discipline passing on the on the Winnies and Clawcastles/NewHammers over the last few weeks) start posting their results.Quote:
Originally Posted by JDizzle
aMEN.Quote:
Originally Posted by gundamit
what seems to be the temp threshold here? what are the chances OPP and pc_ice can swap cpus? I am planning on running one of these on my single stage, so scaling is really the issue for me personally. My winnie was only about 100mhz better on phase. it would be nice to just crank her up and get a 1ghz oc but I think those days are over. :(
I, for one, am not disappointed with these results at all. As PC Ice said, 2.8GHz AT STOCK VOLTAGE is nothing to complain about.
If I can get a 3000+ and run it at 2.8GHz under water cooling I will be happy, and it looks ilke I might be able to do just that.
Thanks for taking the time to test this stuff, PC Ice! :toast:
True different cpus will react differently to volts. I remembered the old 3.0c SL6WU didnt respond to volts at all, while the 3.0c SL6WK responded nicely to more volts. So we will just have to wait and see.
Really need some air results chaps so the rest of us can work out what kinda oc we can expect.
Same here. But what concerns me is the 4x512 ram stuff 'cos right now i'm facing 4x512 of TCCD (just landed) that i expect to play nice with a lil' Venice, so really hope this will be sorted one way or another.Quote:
Originally Posted by mcnbns
i agree, these are the most important seeing as there are more air coolers out there than water and phase put togetherQuote:
Originally Posted by s0ck
If you have an XP90 equivalent or better water cooling won't help with an overclock much at all, so these results are just about as good as it gets.Quote:
Originally Posted by s0ck
Water cooling doesn't give you that much better temps than air, but it has a higher capacity for cooling (it can deal with more heat better).
I'd disagree there...Quote:
Originally Posted by mcnbns
Exactly my point... it makes it difficult to gauge the results, especially when water has such a massive cooling capacity in comparison with air, despite the temps being very similar.Quote:
Water cooling doesn't give you that much better temps than air, but it has a higher capacity for cooling (it can deal with more heat better).
What I mean is, in this case, since a good air cooled setup will be able to deal with this amount of heat that water doesn't have an unfair advantage.
In either case, these results are a lot more informative than we often see for CPUs that aren't even released yet.
xp90 and xp120 and other heatsinks with a high cfm fan get very very close to watercooling, and only top notch good watercooling setups can get notably better temps in my expirience.Quote:
Originally Posted by s0ck
i received a zip with more screenshots i will post here in 30mins or so :D
:D im so glad i just checked this thread.......gotta love dumb luck.
ok, thats all so far, as soon as i get more results from pc ice or oppainter i will post them :)
what caused the error in sandra just pushed it to high and what you said in the first post about the volts is that still there with water?
does anybody have any reservations about buying one of these venice chips and it might not work on phase change? i know i do. i been looking for a solution incase my winchester doesnt agree with the chilly1 phase change cooler that hes making soon, and well i know it wont reach -90c and -40c might be more reasonable but who says all venice chips work like this one. i guess we have to wait till they are released for some solid feed back on subzero compatibility, besides one chip.
Well.....so far it's two outta two that work with single stage. Both the 3800+ and the 3500+ didn't have a problem at sub zero. The coldest I could get the 3800+ was ~ -35C. Also, the 3800+ was able to run benches @ ~2.9G on a stock hsf:Quote:
Originally Posted by odb
http://img57.exs.cx/img57/5968/10x2909pv.jpg
Personally, I think that these cpu's do scale with voltage. It's just that the range of vcore is smaller that what we're used to. The 3800+ stopped responding to bumps in Vcore at around 1.65V.
check the first post of this thread, i sorted all results regarding water and phase change cooling :)Quote:
Originally Posted by Fzero
and yes, the error was caused by a too high oc.
you mean you couldnt get the temps to go lower than -35°C because the mach2 didnt get the cpu cooler, or do you mean the cpu crashed when you reached temps below -35°C? evap temps or cpu temps read by the internal diode?Quote:
Originally Posted by s7e9h3n
hmmm did you play with the 3500+ on air or phase change steven?
did it respond to vcore? dave said the 3500+ didnt respond to vcore bumps AT ALL and he tried up to 1.6v :confused: maybe its a problem with his board? :confused:
Hey saaya,
All temps I noted were read on a uei dt200 from a thermocouple attached to the side of the cpu. The -35 was just the limit of the cooling, not the 3800+. I don't think my unmodded mach2 could have gone any lower with this cpu.
Nope, I never even considered testing the 3500+. I needed a break after the 3800+ and put my fx55 back. I actually had spent more time testing the 3800+ with the mach2 than I had with the fx55.
In no way am I questioning PC-Ice's results with the 3500+, but there probably is a problem with this board or something of that nature. I have yet to see a cpu which didn't respond at all to bumps in vcore. The Venice scaled as well as, if not better, a lot of cpu's I've seen. It's just that the way it scaled was a little different. It seemed as if bumping the 3800+'s Vcore didn't have a large affect on the cpu's temperature. I don't think I recall the cpu temp ever idling in the 40's, even @ 1.70 Vcore. Basically, the cpu just stopped responding OC-wise beyond ~1.65V.
Quote:
Originally Posted by saaya
Damn, that sux pc ice. Lets just hope it's just your Venice that doesn't like volts.
Could be that they're now using the "crappy" cores for s939 512kb cpu's, and the good ones for the 1mb cache mobiles (turion) on s754 that have a lower wattage rating (around 35w?). this is however just a theory...
hi. first I want to say HI :D o/c potencial of Venice is realy big! there is some test os San Diego too?
Welcome to XS Raul...If you look at the stickies in this forum and "The Best of AMD" you'll find a peek at the 4200+ San Diego....Quote:
Originally Posted by Raul
i dont think so, even average winchester cpus are able to work at the frequencies and voltages of a turion, i dont think that the turion cpus are the best a64 silicon amd can make, i think the best is used for fx chips, then high end a64 chips and opterons and then they test wich of the remaining cpus works good with low volts and make them turions or mid to low end a64s...Quote:
Originally Posted by _Eduard_
hmmm im going to ask amd about this, it shouldnt be a big secret they wont tell us :D
hi raul, welcome to XtremeSystems :toast:
Could it be possible that Venice and San diego has some kind of overvoltage protection ?
Quote:
Originally Posted by s7e9h3n
As I said before, you can not just assume that CPUs scale with voltage. My Clawhammer gets to 2.6 with near stock voltage, but it is entirely unachievable to get to 2.7 at ANY voltage with air (XP-120). I also do not believe this is an "overvoltage protection" issue since on the 3500+ Clawhammers this phenomena has been seen on most CPUs. Since the Clawhammer is much older than the Venice, I think this implies that this is just a result of the CPU itself and not a new "feature".Quote:
more than stock vcore barely seems to do anything to this chip
Like mine winnie 3000+ i could run on stock and up to 1.8v vcore and when it was going at the best it run 2840mhz, but my 3500+ winnie just like stock and up to 1,6 over that it doesnt happen anything it run as best on water 2870mhz cold water..3500+ did on stock 1.4v 2800mhz right out of the box :)Quote:
Originally Posted by saaya
possible the San Diago somehere got to use was Stable to 2850 Mhz then just dies of doesnt scale well at all to more voltage.Quote:
Originally Posted by WOLF_OF_DK
Um......I'm not sure where I stated that there was an "overvoltage protection" "feature" on these cpu's. :confused: And, unless your clawhammer is able to get to 2.6 AT default vcore, then it would be considered scaling -albeit minor- if you acheive a higher clock with increased voltage.Quote:
Originally Posted by Ackbar
Sorry, I wasn't answering you about the "overvoltage" but I quoted you kinda as to show that there are two sides to this story. Sorry for the confusion! :DQuote:
Originally Posted by s7e9h3n
I'm able to get 2.6 at ~1.55V (default is 1.5V). So you're right, that 0.05 V helped. But no, nothing after that really did anything. Don't get me wrong, voltage helps, by no means am I able to use anything below 1.55V to get to 2.6 GHz. That's obviously not the point though, the point is that after a certain point, the the voltage to clock curve reaches a point where it is no longer well-behaved (ie. it blows up and becomes infinite for all intents).
here is a nice peek at the short preview/review oppainter wrote with the 3500+ to get you all excited :D
because what we had not mentioned so far is that venice is faster clock for clock than a winchester ;) :D
Specs:
DFI 939 Ultra
MSI X800XT 520/560
OCZ 3700 2.5-4-4-8 @ 300
All CPUs ran at 300x8 2400MHz
this is an average result!
there was no bad/buggy run :D
http://img232.exs.cx/img232/9610/a3d034nr.jpg
i wont post more results in this thread.
opps review/preview should appear on the front page very soon, dont miss it! :D
Hehe! SSE3 at work. A promising result.
I wonder how a 3500+ clawhammer would do in that graph? worse than winnie?
From anandtech's talk with AMD...Quote:
Originally Posted by saaya
Top story there today... the talk with AMD. There is more to it than Turion.Quote:
Much as we suspected, all of the power optimizations that went "into" Turion 64 are all transistor level optimizations. Basically, selecting transistors that provide better thermal and power characteristics at the expense of lower switching frequencies.
Hehe, this answers all the "Why should i need a Venice" threads and posts.
True, all these folks thinking they are going to overclock Turion to 3GHz + are sadly mistaken. AMD has traded transistor speed for lower power.
The parts to overclock are the Venice/San Diego parts.
That's a 1% improvement over Winchester in that benchmark... not exactly thrilling, but better than no improvement at all.
terrace, I think it was also done on fewer volts too.
Exactly.Quote:
Originally Posted by terrace215
I don't like graphs like that 'zoomed in' they are misleading and at first glance, performance looks to be over 100% increase.
I think that what OPP wanted to highlight is the fact that the Venice is spectacularly closing the gap with the FX53.
must have venice. it sucks my winnie doesn't like the cold
just wanted to say that... to measure cpu performance please don't use 3dmark03 or 05, use 01 or pifast etc...Quote:
Originally Posted by µnrealneo²
anyway this shows that even with 512kb cache less the venice makes up the speed in some way...
:slap: the cpus were all running at the same clockspeed!Quote:
Originally Posted by S0nic
yeah yeah, i know, amd told me the same when i talked to them at cebit, but i dont really buy it... i dont think they have an independant line of cpu cores for their turion cpus...Quote:
Originally Posted by texuspete00
its 3dmark 2003!!!Quote:
Originally Posted by terrace215
lol....Quote:
Originally Posted by s0ck
at least one of the guys here got it :P :toast: :DQuote:
Originally Posted by LowRun
I'd rather see the graph start from zero; is it too much to ask? :rolleyes:
3Dmark 2003 doesn't interest me as much as 3Dmark2001 or SuperPI.
How does the performance differ in those benchmarks. I'm interested in performance gain with the same clocks and voltages on venice over winchester. 3Dmark2003 always seems to be rather CPU limited.
EDIT: I meant to add also, a 1% improvement is also very promising when run with the same paramaters. What else do you expect from AMD basically respinning the same chips with a few upgrades. 3Dmark03 obviously doesn't support SSE3, so the only thing that can attribute to the gain is the upgraded mem controller and the few misc bug fixes. That to me says a lot.
Damn this venice makes me go :banana:
I sure can use some more CPU power to fill the memory busses
I know that, but since the venice is faster than the winnie clock for clock I was wondering where the clawhammer 3500+ would fit in, is it slower than the winnie clock for clock?Quote:
Originally Posted by saaya
No big deal!Quote:
Originally Posted by alexio
Here you se 3Ghz at 1.3V
http://www.xtremesystems.org/forums/...chmentid=22352
benchstable..
Cascade -125 C
yeah I too have seen 3.2ghz at 1.35v on an fx55 so what we are seeing isnt out of this world ;)
if you would ask politely we might even do it, but you are demanding something LOL who do you think you are demanding something from opp or me? :stick:Quote:
Originally Posted by s0ck
and if you really care so much about seeing a graph starting from zero why dont you just make one then? :stick:
go ahead, you have the numbers...
hmmm sorry, i dont want to spill the beans on opps results. his article will be up later today anyways, so if you have a little more patience you will know :)Quote:
Originally Posted by Orthogonal
you give me a headache! :lol:Quote:
Originally Posted by S0nic
a clawhammer is a clawhammer is a clawhammer. ohhh now i get it, could it be you didnt know that an fx53 is also a clawhammer? :D
go look at the charts, and you will see that a venice is almost as fast as a clwahmmer clock for clock even though it onlyhas half the l2 cache. at least in 2003... :D
braggart :P :D hehehe j/kQuote:
Originally Posted by Tom Holck
this was with a cascade at -125°C though, the venice was only at -40°C :D
and if amd gets them to work at -125°C im sure we will see similar results... and then again dont forget that this is a mid to low end cpu, while the fx55 costs 4 times as much if not more! :D
Are we harsh today , he did ask ?Quote:
Originally Posted by saaya
Is that what you call asking?Quote:
Originally Posted by s0ck
Edit: Glad to see things worked out unrealneo :toast:
it is asking cuz at the end of his question he asked if it's too much to ask, apperently it is, however i wouldn't wanna take the time to make a graph either, too many numbers
That's nice @ -125C, but how about @ -25C? :DQuote:
Originally Posted by Tom Holck
http://img63.exs.cx/img63/4923/cpuz30000xe.jpg
dono my temp can't go so high :DQuote:
Originally Posted by s7e9h3n
IMHO those are both kick butt speeds. Like arguing over a s2000 or an rx8 :)
What is getting my goat now is that the FX can boot under a multi-cascade situation but the a64s are having problems. Quite puzzling. It has to be a BIOS glitch. It makes no since otherwise why a claw hammer and a newcsl core can go under heavy cascade, but the new 90nm stuff can't. Even the pres'hots' can go under cascade and the are SOI and SS. hmmmm. ARGGG. Would AMD even answer us if we asked? Like sayya said, us super overclockers are about .001% of the computer world. Does AMD or Intel really care if a chip doesn't boot when exposed to temps that would kill any living thing it touched? :? I just dont know
I didn't mean to be arsey, sorry guys :)
I,m just reading this thread, and the results look nice on the extreme cooling, but when are you guys going to test them on Air. (or did i missed this somewhere)
i just cant stand when people demand more in an unfriendly tone instead of beeing thankfull for the hard work others have put into something.Quote:
Originally Posted by Little-Iron
asking would look like this:Quote:
Originally Posted by i found nemo
he said:Quote:
can you please make a graph that starts from 0? id really like to see a graph like that :)
Quote:
Originally Posted by s0ck
ok, sorry for my :stick: ing :D :toast:Quote:
Originally Posted by s0ck
it seems that there wont be any air tests from opp or pc ice... sorryQuote:
Originally Posted by Hugo van Dijk
i dont know why pc ice didnt do them, but opp clearly said "i dont do air :D"
hehe :D, so no venice air results... but the reviews should be out VERY soon and they will have air results :)
We should all take note,some people may required being ask with afew pretty please first.?? Why are U making a big deal out of this ? Yes he could of asked a little nicer> " YOU " need to re-read the post,the man DIDNTQuote:
Originally Posted by saaya
DEMAND anything ,nor was he unfriendly. The problem is on your door step.
I'll not say anymore on this ,not worth it. May you have a good day. :)
to me his post sounded and still sounds unfriendly and demanding, and how am i making a big deal out of it? he explained he didnt mean to sound unfiriendly and i said i apologize for :stick: ing him.Quote:
Originally Posted by Little-Iron
the problem was that his post could easily be understood as beeing demanding and unfriendly, and i dont think theres a "problem" with me perceiving his post as unfriendly and demanding as im not the only way who understood it that way :)
and again, the only one who is making a big deal out of this is you by bringing it up again after we cleared the misunderstanding and apologized and it was already a thing of the past :P
btw, the AMD Venice Preview from OPP should appear on the front page tonight! (i was too busy to post it yesterday, my bad :slap: )
it is demanding when someone says, id like to see the graph start from 0, thats a demand, and then when they say is that too much to ask? well, from my personal expierience most ppl say that last part harshly when actually speaking anyways, when is opp's post gonna be up?!?!?!