AMD Phenom II 920 & 940 Full Review

[annton]

I start to think Phenom II is like the HD3800 gpu family. Competitive enough but cant beat the best. Still hoping for an amd cpu like the HD4800.

[roofsniper]

i think its safe to say that the phenom II won't beat out i7. but it will have its niche and should be a successful cpu. just not the performance leader.

[Leeghoofd]

Just await final OC'ed numbers instead posting the same over and over again... first you want clock for clock, now we need to max out, then OC compared to stock Intels...get a grip man... it's due to these posts that the admin closed the thread...as some of you are repeating the same over and over again. Luckily he reopened it, so the guys from HWbox get a chance to give us their real numbers you all have been asking for... just give the greeks guru's some time...

And for your info : a good Q6600 can also do 4ghz on water (24/7), a bad one hardly reaches 3.4Ghz...

JUST AWAIT THE NUMBERS !!

EDIT:

if you are taking ocing the q6600 into consideration then why not oc the phenom II? sure maybe an oced q6600 can be better than a stock phenom II but who says a oced phenom II can't be better than a stock q9xxx or an i7? if you are gonna oc one then you oc them all. the clock to clock at 3.7 will help some but you still have to remember the fact that if the q6600 maxes at 3.7ghz then you should max out the phenom II as well.......

So you aren't wondering if an Oc'd PII can beat a stock I7 ? 16 mins later you mention that PII can't beat out I7... I'm lost here

I mentioned a few times this will be a nice CPU for all AMD users as an upgrade as it closes the gap. The thing is many people react to PII as horrible as it has been hyped big time... and till now it doesn't deliver ( according the hype created ) the goods as clock for clock the older kentsfields are right on it's heels... Maybe the OC'ed numbers will shine a different light on this CPU... and with "you" I'm talking to everyone that feels adressed (otherwise I would have quoted your post in the first place, which I added now)

[roofsniper]

Just await final OC'ed numbers instead posting the same over and over again... first you want clock for clock, now we need to max out, then OC compared to stock Intels...get a grip man... it's due to these posts that the admin closed the thread...as some of you are repeating the same over and over again. Luckily he reopened it, so the guys from HWbox get a chance to give us their real numbers you all have been asking for... just give the greeks guru's some time...

And for your info : a good Q6600 can also do 4ghz on water (24/7), a bad one hardly reaches 3.4Ghz...

wait what are you talking about? i don't remember ever asking for clock for clock and im saying yes max it out and i don't see myself looking for oc compared to stock intels im just saying how ridiculous does that sound when you say oced intels to stock phenom IIs. maybe if you would read my posts correctly you would understand what i am saying. i don't recall ever saying this was a bad review either im just saying its not a full review and there are other things i want to see. but that doesn't have to be done in this review there can be many others to do that as well. i just think its crazy that so many people are coming to conclusions on just one review that is not even full. this thread would of never been locked if people would have just taken the info and absorbed it in instead of saying phenom II sucks without having any proper info behind that.

[BenchZowner]

roofsniper
Most if not all Q6600 G0's do 3.6GHz on air.
Will every or nearly every Phenom II 920 do 4GHz on air to be equal to the Q6600's @ 3.6GHz on the 45% of the apps and a bit faster than them on the rest ?

If you ask me, the answer is no.
Most Phenom II 920's won't be able to hit 4GHz 24/7 stable on air cooling.
The CPUs are out now and people can buy them, and we shall have a decent sample to calculate the average OC of the Phenom II's soon.

[knissel]

AMD has not enough money to compete with the mighty Intel!

[roofsniper]

roofsniper
Most if not all Q6600 G0's do 3.6GHz on air.
Will every or nearly every Phenom II 920 do 4GHz on air to be equal to the Q6600's @ 3.6GHz on the 45% of the apps and a bit faster than them on the rest ?

If you ask me, the answer is no.
Most Phenom II 920's won't be able to hit 4GHz 24/7 stable on air cooling.
The CPUs are out now and people can buy them, and we shall have a decent sample to calculate the average OC of the Phenom II's soon.

we will have to see how the 920 and 940 overclock but for right now its wayyyyyyyyy to early to make conclusions about this. especially when only knowing the performance results from one review. wait a little while and we will see what it can do. and it will evolve over time with boards made for it instead of beta bios. its mostly posts like the one above me now that make no sense at all.

[BenchZowner]

we will have to see how the 920 and 940 overclock but for right now its wayyyyyyyyy to early to make conclusions about this. especially when only knowing the performance results from one review. wait a little while and we will see what it can do. and it will evolve over time with boards made for it instead of beta bios. its mostly posts like the one above me now that make no sense at all.

We have 3 Phenom II's here.
Another friend of mine also bought one yesterday.
And I've asked 2 friends of mine who also happen to have one ( each one ) to review for their sites.

You can hope... but I wouldn't count on it.

[informal]

roofsniper
Most if not all Q6600 G0's do 3.6GHz on air.
Will every or nearly every Phenom II 920 do 4GHz on air to be equal to the Q6600's @ 3.6GHz on the 45% of the apps and a bit faster than them on the rest ?

If you ask me, the answer is no.
Most Phenom II 920's won't be able to hit 4GHz 24/7 stable on air cooling.
The CPUs are out now and people can buy them, and we shall have a decent sample to calculate the average OC of the Phenom II's soon.

Clock for clock,Agena core was ~5-6% slower than Kentsfield core(in stuff people actually use on their computer-no superpi lovin' and smilar).This would mean that you would need ~3.78-3.8Ghz Agena chip to equal a 3.6Ghz Kentsfield(which will never happen since there are no 3.8Ghz Agena chips that can run like that 24/7).Deneb on average should be from 7 to 10% faster than Agena,clock for clock, in real world apps and with equal settings(ram,L3 etc).This means Deneb at 3.6Ghz is more than a match for Kentsfield(I'd say it would handily beat it in majority of applications-excluding SPi,Sisoftsandra and the likes).Also,at 3.6Ghz it will draw less power and dissipate less heat.At 4Ghz it would be no contest between the two since Kentsfield would never reach that clock for 24/7 and would be a good deal slower than Deneb...

[Chad Boga]

Clock for clock,Agena core was ~5-6% slower than Kentsfield core(in stuff people actually use on their computer-no superpi lovin' and smilar).

On just what applications are you basing this on?

[informal]

On just what applications are you basing this on?

Search Hornet's posts,he made a table from 3 or 4 reviews sites.Various applications,mostly non synthetic but there are a couple of those as well .Two charts,one with Sisoftsandra bs and one without;adobe photoshop results are skewed/low with Phenom due to CnQ issues with this app,but it wouldn't change much in Phenom's results.

[qurious63ss]

yes i understand that but it is not proven. and if you are taking ocing the q6600 into consideration then why not oc the phenom II? sure maybe an oced q6600 can be better than a stock phenom II but who says a oced phenom II can't be better than a stock q9xxx or an i7? if you are gonna oc one then you oc them all. the clock to clock at 3.7 will help some but you still have to remember the fact that if the q6600 maxes at 3.7ghz then you should max out the phenom II as well.......

No I'm more interested in clock for clock to see where IPC is at with these chips. I have heard anywhere from 5-25% better then Phenom I so it would be interesting to see what it really is. IPC is pretty much standard and can be easily proven, while overclocking would be all over the board with some cherry chips hitting 4ghz while others might struggle to get to 3.5ghz, this goes for both Q6600 and possibly PIIs.

[OverClocker_gr]

Hey guys.

Due to time limitations maybe the update will be on air tomorrow.

What do you think about that test?

http://www.spec.org/gwpg/gpc.static/vp10info.html

[qurious63ss]

Clock for clock,Agena core was ~5-6% slower than Kentsfield core(in stuff people actually use on their computer-no superpi lovin' and smilar).This would mean that you would need ~3.78-3.8Ghz Agena chip to equal a 3.6Ghz Kentsfield(which will never happen since there are no 3.8Ghz Agena chips that can run like that 24/7).Deneb on average should be from 7 to 10% faster than Agena,clock for clock, in real world apps and with equal settings(ram,L3 etc).This means Deneb at 3.6Ghz is more than a match for Kentsfield(I'd say it would handily beat it in majority of applications-excluding SPi,Sisoftsandra and the likes).Also,at 3.6Ghz it will draw less power and dissipate less heat.At 4Ghz it would be no contest between the two since Kentsfield would never reach that clock for 24/7 and would be a good deal slower than Deneb...

Wasn't there an "AMD slide" that gave deneb an 8% IPC advantage over agena. So that would make deneb 2% faster then kentsfield?, which is within the margin of error to consider them equal for IPC.

[Shintai]

Clock for clock,Agena core was ~5-6% slower than Kentsfield core(in stuff people actually use on their computer-no superpi lovin' and smilar).This would mean that you would need ~3.78-3.8Ghz Agena chip to equal a 3.6Ghz Kentsfield(which will never happen since there are no 3.8Ghz Agena chips that can run like that 24/7).Deneb on average should be from 7 to 10% faster than Agena,clock for clock, in real world apps and with equal settings(ram,L3 etc).This means Deneb at 3.6Ghz is more than a match for Kentsfield(I'd say it would handily beat it in majority of applications-excluding SPi,Sisoftsandra and the likes).Also,at 3.6Ghz it will draw less power and dissipate less heat.At 4Ghz it would be no contest between the two since Kentsfield would never reach that clock for 24/7 and would be a good deal slower than Deneb...

Just 2 years late to the party. Now you got Yorkfields instead thats faster and uses even less power than PH2 and OCs even more. Oh, even those are old....

[informal]

Just 2 years late to the party. Now you got Yorkfields instead thats faster and uses even less power than PH2 and OCs even more. Oh, even those are old....

Yeah,let's just wait for the official launch and results on power draw,speed,OCing etc.

Wasn't there an "AMD slide" that gave deneb an 8% IPC advantage over agena. So that would make deneb 2% faster then kentsfield?, which is within the margin of error to consider them equal for IPC.

Yeah,and Penryn is on average ~5% faster than Kenstfield,so it's 3-4% from Deneb. Deneb will receive a minor boost(~3%) from DDR3 ram due to dram prefetch in IMC + higher bw/lower latency to memory.

[TL1000S]

Clock for clock,Agena core was ~5-6% slower than Kentsfield core(in stuff people actually use on their computer-no superpi lovin' and smilar).This would mean that you would need ~3.78-3.8Ghz Agena chip to equal a 3.6Ghz Kentsfield(which will never happen since there are no 3.8Ghz Agena chips that can run like that 24/7).Deneb on average should be from 7 to 10% faster than Agena,clock for clock, in real world apps and with equal settings(ram,L3 etc).This means Deneb at 3.6Ghz is more than a match for Kentsfield(I'd say it would handily beat it in majority of applications-excluding SPi,Sisoftsandra and the likes).Also,at 3.6Ghz it will draw less power and dissipate less heat.At 4Ghz it would be no contest between the two since Kentsfield would never reach that clock for 24/7 and would be a good deal slower than Deneb...

I think the thread was closed last night, partly due to "flamewars" between a certain type of fans and dissidents (I would not use the words of Nasgul).
And partly due to "endless-loop-posts" like the one I am quoting regarding "normal apps everybody is using".
This is xs, not your regular "chatter-forum".
But yes, there will be little difference in "performance" to open files, surf the net using MSN, Facebook and running torrents for all the music/films that is "must have" on the average 2TB disks.
We that bench and like to read about wr at xs.org really, really do not care one bit about these "apps".
If you get my point.

Kudos to the greek gurus for this review/comparisons.

[Rammsteiner]

Just 2 years late to the party. Now you got Yorkfields instead thats faster and uses even less power than PH2 and OCs even more. Oh, even those are old....

Errr, Yorkfield doesnt OC all that much better than PhII thus far, and then we're on one of the first steppings. And faster? I dont know, Agena wasnt too far off, main problem was clocking which is solved now pretty much.

Get seom DDR1200 with a NB of 2.4Ghz (or even higher) and you get some damn nice scores already

[informal]

I think the thread was closed last night, partly due to "flamewars" between a certain type of fans and dissidents (I would not use the words of Nasgul).
And partly due to "endless-loop-posts" like the one I am quoting regarding "normal apps everybody is using".
This is xs, not your regular "chatter-forum".
But yes, there will be little difference in "performance" to open files, surf the net using MSN, Facebook and running torrents for all the music/films that is "must have" on the average 2TB disks.
We that bench and like to read about wr at xs.org really, really do not care one bit about these "apps".
If you get my point.

Kudos to the greek gurus for this review/comparisons.

Nobody can stop you from running SuperPi all day long.But hey,I guess you never heard of a/v compression,3D rendering using Maya,archiving of huge files,encryption,compiling,gaming etc. For you "xtreme user" they may be lame,but some people use them a lot and may even earn money using them .

As for Greek review,as you can see thsoe guys are slowly adding some proper tests apart from the initial ones,and I can say kudos to them . If someone lieks to see SuperPi perf. since he likes to compete for fun in this particular piece of software,than fine.Same goes for other synthetic apps. But being xtreme doesn't mean only "benching" superPi all day long,it means using your extremely tuned/OCed system for something useful,be it 3D rendering or crunching for common good of the People .

[qurious63ss]

Yeah,and Penryn is on average ~5% faster than Kenstfield,so it's 3-4% from Deneb. Deneb will receive a minor boost(~3%) from DDR3 ram due to dram prefetch in IMC + higher bw/lower latency to memory.

So then for IPC all the chips are equal and we should all be happy. Let's turn our attention to overclocking and power.

[informal]

So then for IPC all the chips are equal and we should all be happy. Let's turn our attention to overclocking and power.

I agree,but so far we only have a few forum threads from end users,no "official" info at all.All the big hw websites are waiting for the 8th to publish their articles.Until then OCing ,thermals and power draw will be a matter of speculation(unless the Greek site or someone else measures the power draw).

[OverClocker_gr]

guy's.

Please don't make mods to lock it again

[qurious63ss]

guy's.

Please don't make mods to lock it again

Yup, cease fire boys. Anyways, any idea why the cinebench score is so low for deneb? If you look at Anand's numbers for X4 9950, at stock (2.6ghz) they got a score of 2599 for single cpu and a score of 10108 for multi-cpu. Any ideas to what's going on?

http://www.anandtech.com/cpuchipsets...px?i=3448&p=17

[OverClocker_gr]

Windows X64

[qurious63ss]

Windows X64

I see thanks. So for cinebench, X64 makes a difference for deneb but absolutely no difference for nehalem? The scores for nehalem are identical between yours and Anands.