AMD Phenom II 920 & 940 Full Review [UPDATED with more tests]

[Mechromancer]

While the Phenom IIs don't suck, they still are not equal to Penryn clock per clock. Performance/Price is a different story though.

[Spacemaster]

BTW can you do a cinebench x64 quick run with Q9450 and PhII 920 to complete the summary?
Thanks!

OverClocker_gr will do test at same clock with all CPUs so patiente please

[TL1000S]

I try again (since the first thread was closed):

OK. Back to results & facts.

Next relevant question about Cinebench10 (32-bits btw).

Why did "Overcklock gr" get so much better numbers for i920/i940 than for instance Xbitlabs?

[Smalltimer]

While the Phenom IIs don't suck, they still are not equal to Penryn clock per clock. Performance/Price is a different story though.

And clock too!
Now if we can get away with raising the clocks to the point where we are getting a bargain... the we get right into that magic moment territory.

So I think price and O/C'ability could very well be the deciding factors on the PII's success. But even so... as it stands, I'd try one out especially since it works on existing platforms.

[OverClocker_gr]

I try again (since the first thread was closed):

OK. Back to results & facts.

Next relevant question about Cinebench10 (32-bits btw).

Why did "Overcklock gr" get so much better numbers for i920/i940 than for instance Xbitlabs?

i will check out if there is any flip between the 64bit and 32bit results.

this is your only problem?

[roofsniper]

snip

the phenom II holds its own at its price and is on par with yorkfield(its compeition) on benchmarks that matter and close to i7 for gaming. to me its not nearly a budget system but i guess you are entitled to your own opinion. don't see how this will bother you much since you are "Intel & nVidia GeForce user, exclusively." id like to remind you of this: http://www.xtremesystems.org/forums/...&postcount=480

seems like a pretty good cpu but it looks like how it overclocks/scales will make it or break it.

[OverClocker_gr]

excuse me,but where did he said that???????????????

snip

plz man.stay on topic

[Hornet331]

Thanks man for the update!

Updated (and extended) my previous summary too :P

hmm that chart is nearly identical (performance gap wise) to the Kentsfield - Phenom 1 chart i made.

Seems P2 was/is a good shrink with minor performance enhancements, with the performance right where i thought it would be (between Kents and Yorkfield)

[informal]

10percent?
where?

Here :
Phenom 940 should score 18.6*1.07=~19.02 which is 9% lower than score in the test(19.02/18.21).
Also i940 scores the same as 920 with and without Turbo mode,in second image/chart which is kinda odd.Turbo scores are different than the ones without it so we know CPU is the bottleneck here(as the scores change with the CPU clock when Turbo kicks in).One explanation could be that Turbo adds more speed bins on 920 model but the scores are practically identical,so i guess it could be an error with data input(when you created the charts).Also the OFF and ON bars are misplaced as "Turbo moded" 940 is slower than a i7 920 with Turbo off ,so it must be a typo when you created the charts.

[Hornet331]

I try again (since the first thread was closed):

OK. Back to results & facts.

Next relevant question about Cinebench10 (32-bits btw).

Why did "Overcklock gr" get so much better numbers for i920/i940 than for instance Xbitlabs?

64bit makes a huge difference on Cinebench for all cpus not just intel cpus.

For Ci7 as example its ~20-25%.

[qurious63ss]

the phenom II holds its own at its price and is on par with yorkfield(its compeition) on benchmarks that matter and close to i7 for gaming. to me its not nearly a budget system but i guess you are entitled to your own opinion. don't see how this will bother you much since you are "Intel & nVidia GeForce user, exclusively." id like to remind you of this: http://www.xtremesystems.org/forums/...&postcount=480

seems like a pretty good cpu but it looks like how it overclocks/scales will make it or break it.

PII holds it's own with Kentsfield, but on par with Yorkfield in clock for clock...well that I don't see especially when SSE benches come out. Overall it looks like around 6% better IPC then PI as expected.

[roofsniper]

PII holds it's own with Kentsfield if priced lower, but on par as clock for clock well that I don't see especially when SSE benches come out. Overall it looks like around 6% better IPC then PI as expected.

i want to get some pricing from amd but i doubt we will know that until the day it is released. or the night before as amd likes to do. in the stock benches deneb overtakes the q6600 and hangs with yorkfield. it seems to perform well when the fps is below 60 which is the refresh rate of most lcds out there. it also looks good on 64 bit but too bad there isn't many 64 bit applications out there.

[OverClocker_gr]

Here :
Phenom 940 should score 18.6*1.07=~19.02 which is 9% lower than score in the test(19.02/18.21).
Also i940 scores the same as 920 with and without Turbo mode,in second image/chart which is kinda odd.Turbo scores are different than the ones without it so we know CPU is the bottleneck here(as the scores change with the CPU clock when Turbo kicks in).One explanation could be that Turbo adds more speed bins on 920 model but the scores are practically identical,so i guess it could be an error with data input(when you created the charts).Also the OFF and ON bars are misplaced as "Turbo moded" 940 is slower than a i7 920 with Turbo off ,so it must be a typo when you created the charts.

first of all,lets forget about "should"

no the results are correct
its just the delta that i said before or these are gpu bottlenecked

[OverClocker_gr]

roofsniper + qurious63ss

plz.tell your opinions and argue over pm

else
http://www.xtremesystems.org/forums/...&postcount=480

[roofsniper]

first of all,lets forget about "should"

no the results are correct
its just the delta that i said before or these are gpu bottlenecked

its just odd that turbo lowers the performance in some benchs. but the thread isn't even about i7's performance so i wouldn't dwell on it.

[informal]

first of all,lets forget about "should"

no the results are correct
its just the delta that i said before or these are gpu bottlenecked

Well the 2.8Ghz Phenom II model is scoring 2% better than a 3Ghz Phenom II,so with 7% lower clock.There's your 9% delta,a double of 4-5% margin of error

But i don't doubt your tests if you were implying that.I just see inconsistency in some phenom and core i7 scores,which could be an error not related to test but to results presentation/creation.

[roofsniper]

roofsniper + qurious63ss

plz.tell your opinions and argue over pm

else
http://www.xtremesystems.org/forums/...&postcount=480

didn't think we were arguing.

[OverClocker_gr]

I try again (since the first thread was closed):

OK. Back to results & facts.

Next relevant question about Cinebench10 (32-bits btw).

Why did "Overcklock gr" get so much better numbers for i920/i940 than for instance Xbitlabs?

damn,you were totally right

these numbers are from Cinebench R10 x64.
with so many screenshots might be a flip with the results.

I ll correct them tomorrow.

Sorry again and thanx for insisting on that


for the history

940 i7
single 3335
multi 13962

[OverClocker_gr]

Well the 2.8Ghz Phenom II model is scoring 2% better than a 3Ghz Phenom II,so with 7% lower clock.There's your 9% delta,a double of 4-5% margin of error

But i don't doubt your tests if you were implying that.I just see inconsistency in some phenom and core i7 scores,which could be an error not related to test but to results presentation/creation.

just saw the screenshots from these tests and they are correct.maybe the problem is on the progy.

But a friend that saw the results said that its normal

[informal]

damn,you totally were right

these numbers are from Cinebench R10 x64.
with so many screenshots might be a flip with the results.

I ll correct them tomorrow.

Sorry again and thanx for insisting on that


for the history

940 i7
single 3335
multi 13962

Dang i completely missed the wrong Cinebench scores .
Good to see you will correct them

just saw the screenshots from these tests and they are correct.

I totally believe you,but there is no way a 2.8Ghz Phenom II to score 2% better than a 3Ghz Phenom II.It is 9% delta per clock,quite a big one.As with Cinebench/Core i7, there is some error with the testing method or somewhere in the process.

[roofsniper]

overclocker gr think you might be able to find power consumption?

[qurious63ss]

didn't think we were arguing.

Me neither. Lost in translation I guess. Anyways, keep the numbers coming Overcloker much appreciated.

[OverClocker_gr]

I totally believe you,but there is no way a 2.8Ghz Phenom II to score 2% better than a 3Ghz Phenom II.It is 9% delta per clock,quite a big one.As with Cinebench/Core i7, there is some error with the testing method or somewhere in the process.

i ll make a rerun tomorrow for the 3d render and i ll see.


whats wrong about cinebench?
What could be wrong?
its a click n run

[informal]

i ll make a rerun tomorrow for the 3d render and i ll see.


whats wrong about cinebench?
What could be wrong?
its a click n run

I meant the core i7 cinebench scores compared to xbit labs results(32 bit vs 64bit,a typo).

[OverClocker_gr]

OverClocker_gr just gave a big one finger salute to AMD! Thanks for not signing (or ignoring) NDA.

nda?
Here in greece denebs are on the market about a week now

I meant the core i7 cinebench scores compared to xbit labs results(32 bit vs 64bit,a typo).

xbitlabs have R10 32bit

their 940 scored 14674 but they are not mentioning if this is with turbo boost or w/o.

cause see my results

turbo off - turbo on
s 3335 3471
m 13962 14387

so maybe they are with turbo on