First Phenom II review + New review 26th

[WaitingForNehalem]

Originally posted by: Nemesis 1

Originally posted by: Lonyo

Originally posted by: Nemesis 1
I read this early this morning. Could you link to were PII kills IC7 in games. I MISSED that!

Also your aware as most are that IC7 getts faster with more GPU power . The PII is a nice little cpu that O/C well. But its not even close to a match for IC7. PII will have to play against Penryn . IC7 is years ahead.

However, they are also in completely different price ranges.
Assuming the £190 price for the Phenom II 920 is what we see retail, the i7 920 (2.66GHz) goes for £230, while the i7 940 is more like £450.

So while the Phenom II may perform worse than Core i7's, it's also cheaper both on a per-CPU basis, and on a per-platform basis (cheaper motherboards, cheaper RAM currently).

For comparison to Core 2 Quad, the Q6600 is about £150 and the Q9450 is about £250.

So it's:
Q6600 £150
Phenom II 920 £190 (approx)
Phenom II 940 £220 (approx)
Core i7 920 £230
Core 2 Quad Q9450 £250
Core i7 940 £450

Which means that overall the Phenom II may be a reasonable choice of product, since really it's competing with slower C2Q's than the 3.2GHz one used in this "test", assuming that the numbers are anywhere close to accurate.
PII will have to compete against Lynnfield (?) i.e. the lower cost dual channel version of Core i7.

Yep it is cheaper and ALL around alot slower. On the O/Cs A few guys are getting 4.4GHz on waterIC7. Until we see real reviews of PHII O/C its not a done deal. Also by your reasoning . O/Cs for both IC7 & PHII must be done at same voltage . The lowewr V. to do the samr GHz. Is winner. Does that sound fair?

My daughter just order 10 of each of these parts. + More.

IC7 920 $299
Memory DDR3 3x2mb. $239/
Motherboard $398
GPU $539.
$ 1475 The first meroms we built were = in cost.

You could save $500 if you get the Core i7 920 from MicroCenter for $250.

 

[ajaidevsingh]

My god i said Turbo was cheating because they already are using the i7 940 @ 2.93GHz which is almost equal to the clock rate of Phenom II 940 as it is. You get the clock to clock comparison of two different architectures.

If it was a i7 920 with Turbo enabled it would go a max of 2.66+0.20Ghz "2.86 Ghz". I think that Turbo can add upto 200 Mhz maybe i am wrong anyways that clock rate would be slower to the i7 940 without turbo but if turbo is enabled in an i7 940 you could reach speeds of 3.13 Ghz which would be faster than Phenoms II clock rate.

 

[Nemesis 1]

Ya So the IC7 940 self clocks higher. User doesn't do anything but benefit from small increase.

Man I wish real reviews would come out. I hate this. I normally have parts but NO amd cpus . I never ever count an O/C unless its Prime stable. Ananda will give us prime stable O/C numbers. I want to see were this chips are stable for 24/7

 

[dmens]

clock to clock is a load of crap *especially* when comparing different architectures.

 

[ajaidevsingh]

Originally posted by: Nemesis 1
Ya So the IC7 940 self clocks higher. User doesn't do anything but benefit from small increase.

Man I wish real reviews would come out. I hate this. I normally have parts but NO amd cpus . I never ever count an O/C unless its Prime stable. Ananda will give us prime stable O/C numbers. I want to see were this chips are stable for 24/7

Friend i7 920 ran Prime95 for a whole night at 3.6Ghz with Ultra 120 and temp spiked at mid 80's. But i think i7 on water cold breach 4Ghz and now regret telling him to go for the ultra 120.

Real review in Jan till then we can squat over these real/ non real results

dmens if we dont use clock to clock and try the sweet spot approach that would be the best review ever. I mean an i7 920 @ 3.6ghz vs a Phenom II 940 @ 3.6-4.0Ghz with both on air "Ultra 120 maybe" would be the best review ever.

 

[aigomorla]

Originally posted by: ajaidevsingh

dmens if we dont use clock to clock and try the sweet spot approach that would be the best review ever. I mean an i7 920 @ 3.6ghz vs a Phenom II 940 @ 3.6-4.0Ghz with both on air "Ultra 120 maybe" would be the best review ever.

no becase all of us who come here arent hardcore gamers that only look at games when benching cpu's.



as demens says clock to clock is pointless unless we have a basis line to compare both too.

And im willing to put money a i7 920 with HT enabled, and Turbo enabled STOCK... can gobble about anything AMD throws STOCK in pure SMP and work numbers....

The price difference, is another story, but, once again... these chips arent on the same class.

So why is people throwing i7's with AMD PhII again? when AMD's market plan was yorkfield conquest and not i7 conquest?

You inflating a product without any real numbers.. this is what im trying to get across..


And im running a Gainestown machine with 2 x qpi enabled... however only 1 chip..
If systool overclocking is allowed like it was on harpers and yorkies....

Then AMD only has a deep and dark pit it can go throw, because i find it very hard for any AMD machine on 2 processor 8 thread bed to eat a Neha-EP package with 2 cpu's and 16 threads @ greater then 3ghz...

Originally posted by: ajaidevsingh
My god i said Turbo was cheating

Oh sorry, lemme go spank IDC for leading me to the wrong water hole..

 

[dmens]

Originally posted by: ajaidevsingh
dmens if we dont use clock to clock and try the sweet spot approach that would be the best review ever. I mean an i7 920 @ 3.6ghz vs a Phenom II 940 @ 3.6-4.0Ghz with both on air "Ultra 120 maybe" would be the best review ever.

or you can just review them stock because that is what is being sold on the market. overclocked reviews are bullshit because both companies send cherry picked parts, which are completely deviant from typical parts.

i know a "nice" 920 i7's can easily hit 4ghz on air around 1.3V but most of them won't, so all the 920 reviews with 4ghz oc's gave me chuckles.

im also convinced the whole super 4ghz oc capability of phenom II is nothing more than an idiotic publicity stunt and you and many others are buying it hook, line and sinker. you might luck out and get a fast part, but more than likely you'll get an average or below average part.

why'd you figure some people can do ridiculous overclocks with c2d's, but some people can't? the only way to beat the bell curve is to have inside info.

 

[SunnyD]

Originally posted by: dmens
clock to clock is a load of crap *especially* when comparing different architectures.

Correct, which is why clock to clock comparison between a Phenom II and a Core i7 is perfectly valid. They're both x86 "architectures". The end result is the same - they run the same exact software. So if you're benchmarking a title with, say FPS, then a clock to clock comparison shows the strengths and weaknesses of each platform, assuming everything else is apples/apples (video card, OS, etc).

 

[dmens]

Originally posted by: SunnyD

Originally posted by: dmens
clock to clock is a load of crap *especially* when comparing different architectures.

Correct, which is why clock to clock comparison between a Phenom II and a Core i7 is perfectly valid. They're both x86 "architectures". The end result is the same - they run the same exact software. So if you're benchmarking a title with, say FPS, then a clock to clock comparison shows the strengths and weaknesses of each platform, assuming everything else is apples/apples (video card, OS, etc).

they are different x86 architectures

it is still a garbage comparison. a clock to clock comparison merely gives you a rough IPC measurement (which is also misleading). designs sacrifice width for higher clocks, or they can sacrifice clock speed for fat buffers and pipelines. either way frequency is a fundamental design metric that cannot be ignored at any time, which is what you do when you equalize clocks.

to summarize, would you downclock a P4 to make a comparison with a K8? would you be able to draw any meaningful performance comparisons that way?

 

[Nemesis 1]

Originally posted by: SunnyD

Originally posted by: dmens
clock to clock is a load of crap *especially* when comparing different architectures.

Correct, which is why clock to clock comparison between a Phenom II and a Core i7 is perfectly valid. They're both x86 "architectures". The end result is the same - they run the same exact software. So if you're benchmarking a title with, say FPS, then a clock to clock comparison shows the strengths and weaknesses of each platform, assuming everything else is apples/apples (video card, OS, etc).

Sonny its been shown that IC7 on todays GPus is weaker with single card . Add cards and the numbers change. IC7 is ahead of the GPUs its about time.

 

[dmens]

delete

 

[SunnyD]

Originally posted by: dmens

Originally posted by: SunnyD

Originally posted by: dmens
clock to clock is a load of crap *especially* when comparing different architectures.

Correct, which is why clock to clock comparison between a Phenom II and a Core i7 is perfectly valid. They're both x86 "architectures". The end result is the same - they run the same exact software. So if you're benchmarking a title with, say FPS, then a clock to clock comparison shows the strengths and weaknesses of each platform, assuming everything else is apples/apples (video card, OS, etc).

no, it is still a garbage comparison. a clock to clock comparison merely gives you a rough IPC measurement (which is also misleading). designs sacrifice width for higher clocks, or they can sacrifice clock speed for fat buffers and pipelines. either way frequency is a fundamental design metric that cannot be ignored at any time, which is what you do when you equalize clocks.

to summarize, would you downclock a P4 to make a comparison with a K8? would you be able to draw any meaningful performance comparisons that way?

Yes.

You imply clock speed is meaningless, then you say "...either way frequency is a fundamental design metric that cannot be ignored..."

You say designs sacrifice width for speed, etc. You're absolutely correct. A lower clocked part can be more efficient than a higher clocked part. Your P4 example is perfect. Once you equalize the clock speeds you really begin to see where a given part shines. If you know a given part is going to take insanely high clocks to match a competitor, you can look at the longevity of the part (physically and market-wise), and you'll be able to tell that part won't be competitive in the long haul without drastic changes. If the P4 were able to scale to 8GHz without much in terms of added cost (thermals/power/etc), do you think it would have been supplanted by Core quite as soon as it was (given the position of the Athlon 64 not changing substantially)?

Simply put, with the same ISA you can get a rough indicator of where a platform's performance will put it with the competition clock for clock. Now when you talk about thermal efficiency, price, etc... That's another story.

 

[ajaidevsingh]

Originally posted by: dmens

Originally posted by: ajaidevsingh
dmens if we dont use clock to clock and try the sweet spot approach that would be the best review ever. I mean an i7 920 @ 3.6ghz vs a Phenom II 940 @ 3.6-4.0Ghz with both on air "Ultra 120 maybe" would be the best review ever.

or you can just review them stock because that is what is being sold on the market. overclocked reviews are bullshit because both companies send cherry picked parts, which are completely deviant from typical parts.

i know a "nice" 920 i7's can easily hit 4ghz on air around 1.3V but most of them won't, so all the 920 reviews with 4ghz oc's gave me chuckles.

im also convinced the whole super 4ghz oc capability of phenom II is nothing more than an idiotic publicity stunt and you and many others are buying it hook, line and sinker. you might luck out and get a fast part, but more than likely you'll get an average or below average part.

why'd you figure some people can do ridiculous overclocks with c2d's, but some people can't? the only way to beat the bell curve is to have inside info.

At stock being the keyword what i said was that they used an i7 940 even if HT and Turbo are disabled it would still be better than a 920 with HT and Turbo enabled. But then again as aigomorla said this is from the point of view of a gamer.

The review it self shows the i7 beats the crap out of the PhII in crunch's but not in games. The fact that its better than a QX9770 over all in games and below i7 means it can sell in the land of gamers at least. qx9770<Ph940<i7 "No one said i7 is not better but a half priced cpu comes close its performance is a good thing.

I dont know what people like but i for one would mind giving a premium for intel products when intel has no competition.

 

[JackyP]

Both tests have their merits, I think it's as simple as that.

 

[ajaidevsingh]

Originally posted by: dmens

Originally posted by: SunnyD

Originally posted by: dmens
clock to clock is a load of crap *especially* when comparing different architectures.

Correct, which is why clock to clock comparison between a Phenom II and a Core i7 is perfectly valid. They're both x86 "architectures". The end result is the same - they run the same exact software. So if you're benchmarking a title with, say FPS, then a clock to clock comparison shows the strengths and weaknesses of each platform, assuming everything else is apples/apples (video card, OS, etc).

no, it is still a garbage comparison. a clock to clock comparison merely gives you a rough IPC measurement (which is also misleading). designs sacrifice width for higher clocks, or they can sacrifice clock speed for fat buffers and pipelines. either way frequency is a fundamental design metric that cannot be ignored at any time, which is what you do when you equalize clocks.

to summarize, would you downclock a P4 to make a comparison with a K8? would you be able to draw any meaningful performance comparisons that way?

Amm C2D when compared to Phenoms are done on bases of clock and Athlon 64 compared to P4's was also done on bases of clock that is why Athlon 64 was hailed as the king back then.

Sweet spot comparisons can be made to show how good a system can become with a simple upgrade to cooling systems.

 

[dmens]

Originally posted by: SunnyD

Originally posted by: dmens

Originally posted by: SunnyD

Originally posted by: dmens
clock to clock is a load of crap *especially* when comparing different architectures.

Correct, which is why clock to clock comparison between a Phenom II and a Core i7 is perfectly valid. They're both x86 "architectures". The end result is the same - they run the same exact software. So if you're benchmarking a title with, say FPS, then a clock to clock comparison shows the strengths and weaknesses of each platform, assuming everything else is apples/apples (video card, OS, etc).

no, it is still a garbage comparison. a clock to clock comparison merely gives you a rough IPC measurement (which is also misleading). designs sacrifice width for higher clocks, or they can sacrifice clock speed for fat buffers and pipelines. either way frequency is a fundamental design metric that cannot be ignored at any time, which is what you do when you equalize clocks.

to summarize, would you downclock a P4 to make a comparison with a K8? would you be able to draw any meaningful performance comparisons that way?

Yes.

You imply clock speed is meaningless, then you say "...either way frequency is a fundamental design metric that cannot be ignored..."

You say designs sacrifice width for speed, etc. You're absolutely correct. A lower clocked part can be more efficient than a higher clocked part. Your P4 example is perfect. Once you equalize the clock speeds you really begin to see where a given part shines. If you know a given part is going to take insanely high clocks to match a competitor, you can look at the longevity of the part (physically and market-wise), and you'll be able to tell that part won't be competitive in the long haul without drastic changes. If the P4 were able to scale to 8GHz without much in terms of added cost (thermals/power/etc), do you think it would have been supplanted by Core quite as soon as it was (given the position of the Athlon 64 not changing substantially)?

Simply put, with the same ISA you can get a rough indicator of where a platform's performance will put it with the competition clock for clock. Now when you talk about thermal efficiency, price, etc... That's another story.

a part "shines" when it is running at its design frequency, not when equalized to the arbitrary frequency of some other x86 architecture. if the design target calls for insanely high clocks, then it should be run at whatever frequency it was sold at.

by your logic, intel and amd should outrace each other to outdo each other and come out with a part that is just wide enough and lower frequency than the other. too bad that approach will produce designs with lower and lower performance.

and since when did the longevity of a part matter to the consumer? that's a long-term engineering problem. the only thing that matters is the performance, here and now.

 

[Lonyo]

Best comparisons would be price to price, both CPU price to CPU price and platform price to platform price.
Comparing clock for clock is useful to gauge clock for clock performance (obviously), but when the clock speeds aren't equal at a given price, it becomes less meaningful.
But you also can't compare just based on CPU price, since i7 currently has more expensive components (mobo/RAM), so really you want to do a clock for clock comparison alongside a comparison against chips in the same market segment.
The i7 940 isn't in the same market segment as the P2 940.

 

[Cogman]

When comparing parts, Generally the metric that the public uses and cares about is Price/performance. not performance/clock speed. If Intel releases a CPU that can only run from 100-200 mhz but creams everything available today, I would be one of the first in line to get one. Even if they release one that is 1000Ghz and creams everything that is available (so long as the price is reasonable) then you can bet I would get that as well.

I know it is pretty arbitrary, and intel/amd both have the freedom to change their prices at will, but thats the only measurement that I or most everyone else cares about, How well will this $200 CPU compare to that $200 CPU. Not, how well will this 2.4GHz CPU compare to that 2.4GHz CPU.

Heres the kicker, even with that in mind, often another consideration that enters my mind (not everyones), is something like this "Ok, this architecture has a 1.0 GHz chip and a 5.0GHz chip, They have the same level of cache, the only difference is the clock speed. So, Why not get the 1.0Ghz and overclock it as far as it can go? It should do pretty well". That is probably the only time where clock speed really enters my mind when buying a CPU. The rest is purely Price/Performance based.

Edit: Wow Lonyo that was weird, we said about the same thing at about the same time.

 

[SunnyD]

Originally posted by: dmens
a part "shines" when it is running at its design frequency, not when equalized to the arbitrary frequency of some other x86 architecture. if the design target calls for insanely high clocks, then it should be run at whatever frequency it was sold at.

by your logic, intel and amd should outrace each other to outdo each other and come out with a part that is just wide enough and lower frequency than the other. too bad that approach will produce designs with lower and lower performance.

and since when did the longevity of a part matter to the consumer? that's a long-term engineering problem. the only thing that matters is the performance, here and now.

A "part" does indeed shine at it's design frequency. That's what it was designed for. When you're comparing the gamut of "parts" out there - that is what we're talking about. You have to take the least common denominator. Hence - x86 instruction set, they run the same thing, the only thing that needs to be equalized is the clock in order to compare equivalent performance of the architectures.

Simply put analogy: It's like comparing a Prius to a Mustang. The Mustang will kick it's ass left in right in terms of speed, but the Prius will get there in fewer tanks of gas. So how do you compare the cars? They're both cars, they both run on gas. One has 8 cylinders, one has 4, but they're both cars. Gallon for gallon at 60 miles per hour, the Prius is going to kick the Mustang's ass left and right.

And I didn't say anything about longevity of a "part". I said longevity of an "architecture". Huge difference there.

 

[Tempered81]

If they are equal clock-for-clock, and both cost the same, then your choice between i7 or Deneb should come down overclocking potential. If one reaches 3.6ghz steady, and the other 4.4ghz... well it's a no-brainer.

If they are equal clock-for-clock, and both overclock regularly to 4.2ghz, then decide based on price.

If they oc the same, cost the same, and perform the same, umm...

 

[taltamir]

If they oc the same, cost the same, and perform the same, umm...

That is highly improbable.

 

[cusideabelincoln]

Eh, enough arguing. Reviewers should just bench every chip and we, the informed readers, can judge however we like - whether it be performance per clock per price per watt per overclock per application per etc...

 

[Arkaign]

This part had me laughing for some reason :

"In fact, the QX9770 compared is not widely available and cost a bomb."

 

[aigomorla]

Originally posted by: ajaidevsingh
The review it self shows the i7 beats the crap out of the PhII in crunch's but not in games. The fact that its better than a QX9770 over all in games and below i7 means it can sell in the land of gamers at least. qx9770<Ph940<i7 "No one said i7 is not better but a half priced cpu comes close its performance is a good thing.

i totally agree with ya on this statement...

as i said, the i7 is not everyone's cup of tea...

but for the people looking for its niche... nothing can catch up to it..

i still have a 4ghz yorkie for gaming when needed...


i got uber burned on spider, so i highly doubt i'll go AMD this route until i see some solid numbers..

 

[piesquared]

Well it's telling when the Intel surrogates are up in arms about clock speeds. My how this has come full circle. It looks as though all those people comparing X2's at 3.0 ghz to C2D's at 2.6 ghz have been vindicated, thanks to the borg. And Phenom's at 2.6 ghz to C2Q's at 2.4 ghz as well. To me, it's clear evidence that Intel feels they are going to lose the OCing enthusiast influence, auto overclocking or not.

Speaking of auto overclocking on i7, I suppose it's as valid as auto overclocking through BIOS, so reviewers should enable that option on PhII as well when comparing the 2 architectures.