Phenom II 6 GHz+ OVERCLOCK

[Macadamia]

Based on these benches, Intel can use the Q9550 to compete with PII. The Ci7 looks to be a great workstation part where better multithreaded software is available.

The problem with relying on multithreading is that AMD can retaliate easily with 6 cores.

Yes, Intel will probably do the same (6 on 32nm), with AMD having 2 Shanghai/Istanbul (8/12c) dies on one chip as a strikeback.


So server/workstation side, it's rather futile. There is not going to be a Conroe like 2 year lead. It's 6 months - 1 year at best. Sandy and Bulldozer should shake things up a little, but they're both emphasizing on special instructions like AVX/SSE, I wonder how much more core improvements they can have.


Oh, and about the benchmarks. Remember- this is the "oh-so" inferior and narrow K10 architecture that only issues 3 instructions per clock, unlike Core 2. It getting even/beating the C2 is something quite remarkable.

[informal]

Where does it indicate TURBO mode was on for Ci7 940? Are you assuming so? And why is it when PII doesn't fare well, you try to downplay it? I thought there were many core optimizations done over PI?

Despite your effort to explain this away, it is quite clear to any discerning person that a C2Q @ 3Ghz, would win 5 out of those 6 tests, the 6th test could possibly swing either way. THIS IS VERY GOOD FOR PII, NONETHELESS, I don't know why you want to look past the data though.

I'm telling you that games just don't scale well with CPU clock,especially in 16x10 .You can say 3Ghz C2Q will perform better than 2.83Ghz one and that's just fine.But it will still be C2Q with only 6% higher clock in a game that is GPU limited(in 16x10).It will hardly be 6% faster in those settings than a 2.83Ghz C2Q..I think you can understand this?
PhII over PhI is not just a clock uplift but a core and cache one.Total cache amount is doubled and some core tweaks can help too.So PhII is more than a C2Q@ 3Ghz compared to 2.83Ghz one.That's the point i was trying to make.

edit: Turbo is by default always ON...

[qurious63ss]

The problem with relying on multithreading is that AMD can retaliate easily with 6 cores.

Yes, Intel will probably do the same (6 on 32nm), with AMD having 2 Shanghai/Istanbul (8/12c) dies on one chip as a strikeback.


So server/workstation side, it's rather futile. There is not going to be a Conroe like 2 year lead. It's 6 months - 1 year at best. Sandy and Bulldozer should shake things up a little, but they're both emphasizing on special instructions like AVX/SSE, I wonder how much more core improvements they can have.


Oh, and about the benchmarks. Remember- this is the "oh-so" inferior and narrow K10 architecture that only issues 3 instructions per clock, unlike Core 2. It getting even/beating the C2 is something quite remarkable.

6 cores on 45nm will make the die 50% larger so I don't see this as a wise choice for the desktop.

[informal]

6 cores on 45nm will make the die 50% larger so I don't see this as a wide choice for the desktop.

Acually one Shanghai/Deneb core is 15.3mm2.The 6 core part(as shown by Hans De Vries) would be a sub 300mm2 chip.Nehalem single core size(one core on the monolithic die) takes up ~24.4mm2 or 1.6x more than in Shanghai(SMT among other things).

"It's probably not a bad idea for AMD to take advantage of their relatively
small core size. At 45nm it can cram two cores (with 256 kB L1 total) into
30mm2. A single Nehalem core with 256 kB L2 also occupies ~30 mm2.

So in terms of IPC/mm2 you get two cores against one core with 2 threads."

[qurious63ss]

I see. It makes good sense then.

[Zucker2k]

I'm telling you that games just don't scale well with CPU clock,especially in 16x10 .You can say 3Ghz C2Q will perform better than 2.83Ghz one and that's just fine.But it will still be C2Q with only 6% higher clock in a game that is GPU limited(in 16x10).It will hardly be 6% faster in those settings than a 2.83Ghz C2Q..I think you can understand this?
PhII over PhI is not just a clock uplift but a core and cache one.Total cache amount is doubled and some core tweaks can help too.So PhII is more than a C2Q@ 3Ghz compared to 2.83Ghz one.That's the point i was trying to make.

edit: Turbo is by default always ON...

But you don't know IF TURBO MODE was turned OFF for this shootout (for a clearer comparison). Usually, if these performace enhancers are turned on, there is usually a disclaimer. Even if it wasn't, it is a known fact that Nehalem generally offers NO improvements over C2Q in gaming.

Well, accrding to you, the benchmarks are GPU limited, so I guess:


CPU Effect = 0

Therefore:

PII 920 = Q9550 = PII 940 = Ci7 940

In case it's hard for you to understand, my argument is that that measly 6% you talk of, is enough to make up for the marginal difference (yes marginal difference) between Q9550 and PII 940. Mind you, not LINEARLY, if it was, it'll beat it by a relatively bigger margin.


This one needs quoting:

PhII over PhI is not just a clock uplift but a core and cache one.Total cache amount is doubled and some core tweaks can help too.So PhII is more than a C2Q@ 3Ghz compared to 2.83Ghz one.That's the point i was trying to make.

What you're saying will be true if the margins were bigger than is represented. Don't try throw dust into this; I'm looking at fractions of fps difference with compared to a relatively healthy 6%. I think it's enough to make up for those fractions, regardless of uarch, don't you? About uarch improvements to PII 2, my point was if that is the case, why isn't PII fairing any better in certain games than PI?

Finally, let's appeal to logic here:

IN ALL the AMD SLIDES WITH SEEN SO FAR, the Q9650 has not been featured. WHY? Its STOCK CLOCK IS 3GHZ TOO? WHY IN YOUR OPINION, INFORMAL, HASN'T ANYONE TESTED DENEB ALONGSIDE Q9650? YOUR GUESS IS AS GOOD AS MINE.

[informal]

But you don't know IF TURBO MODE was turned OFF for this shootout (for a clearer comparison). Usually, if these performace enhancers are turned on, there is usually a disclaimer. Even if it wasn't, it is a known fact that Nehalem generally offers NO improvements over C2Q in gaming.

Well, accrding to you, the benchmarks are GPU limited, so I guess:


CPU Effect = 0

Therefore:

PII 920 = Q9550 = PII 940 = Ci7 940

In case it's hard for you to understand, my argument is that that measly 6% you talk of, is enough to make up for the marginal difference (yes marginal difference) between Q9550 and PII 940. Mind you, not LINEARLY, if it was, it'll beat it by a relatively bigger margin.


This one needs quoting:

What you're saying will be true if the margins were bigger than is represented. Don't try throw dust into this; I'm looking at fractions of fps difference with compared to a relatively healthy 6%. I think it's enough to make up for those fractions, regardless of uarch, don't you? About uarch improvements to PII 2, my point was if that is the case, why isn't PII fairing any better in certain games than PI?

Finally, let's appeal to logic here:

IN ALL the AMD SLIDES WITH SEEN SO FAR, the Q9650 has not been featured. WHY? Its STOCK CLOCK IS 3GHZ TOO? WHY IN YOUR OPINION, INFORMAL, HASN'T ANYONE TESTED DENEB ALONGSIDE Q9650? YOUR GUESS IS AS GOOD AS MINE.

First of all,we have obscure slides with no configurations at all.Turbo is out of the box feature and is always on.It is just as SMT ,a built in feature.

Second,your premise is false.CPU effect is not = 0.Therefore your conclusion is invalid.CPU effect is not as important in 16x10 as it is in 800x600(in which btw no1 games).In low res. you could probably see some differences that come from game engine optimizations(compilers,datasets used etc.).

In which games PhII doesn't fare any better than Phenom I?In the GPU limited ones where all CPUs have closely grouped scores?

As for the last question about slides and pricing,did it occur to you Q9650 is not priced the same as PhII 940?? AMD done-if the slide is from AMD in the first place- the price per. comparison by comparing systems that will cost similarly.PhII 940 will not cost the same as Q9650 and is logical not to compare it with that model. Have you heard of Radeon 4850/70 card? You know that 4870 model is at 80-85% the perf. of GT200(GTX280),somewhere even higher ,while costing A LOT less. That's called great value.

[Zucker2k]

First of all,we have obscure slides with no configurations at all.Turbo is out of the box feature and is always on.It is just as SMT ,a built in feature.

Second,your premise is false.CPU effect is not = 0.Therefore your conclusion is invalid.CPU effect is not as important in 16x10 as it is in 800x600(in which btw no1 games).In low res. you could probably see some differences that come from game engine optimizations(compilers,datasets used etc.).

In which games PhII doesn't fare any better than Phenom I?In the GPU limited ones where all CPUs have closely grouped scores?

As for the last question about slides and pricing,did it occur to you Q9650 is not priced the same as PhII 940?? AMD done-if the slide is from AMD in the first place- the price per. comparison by comparing systems that will cost similarly.PhII 940 will not cost the same as Q9650 and is logical not to compare it with that model. Have you heard of Radeon 4850/70 card? You know that 4870 model is at 80-85% the perf. of GT200(GTX280),somewhere even higher ,while costing A LOT less. That's called great value.

It looks like CPU does count for something then. YOU'VE BEEN EXPOSED. YOU WERE MAKING A CLOCK FOR CLOCK ARGUMENT, NOW WHEN IT SUITS YOU, YOU SWITCH TO A "BANG FOR BUCK, BUT HEY, LOOK OUT FOR THE Q9550. IT'S GOING TO OFFER BETTER VALUE THAN PII 940 THE WAY THINGS LOOKING, UNLESS AMD IS GOING TO PRICE THE PII 940 @ $250.

Sorry for caps, my keyboard was acting up.

[informal]

It looks like CPU does count for something then. YOU'VE BEEN EXPOSED. YOU WERE MAKING A CLOCK FOR CLOCK ARGUMENT, NOW WHEN IT SUITS YOU, YOU SWITCH TO A "BANG FOR BUCK, BUT HEY, LOOK OUT FOR THE Q9550. IT'S GOING TO OFFER BETTER VALUE THAN PII 940 THE WAY THINGS LOOKING, UNLESS AMD IS GOING TO PRICE THE PII 940 @ $250.

Sorry for caps, my keyboard was acting up.

Calm down,are you getting excited over there?

You may have exposed yourself.I made a clock for clock comparison because you asked me to do it(remember the math comment?).You seem to not understand that games at 16x10 are not CPU limited.CPU helps though,but as you can see,a cache+core uplift means more than pure Mhz uplift(PhII vs PhI).

Second,again you asked me why no Q9650 in the comparisons and when i give you logical response you go all wild with *closed ears with both hands* na-na-na-na comment .
I told you why there is no comparison with Q9650.I can tell you it would fare better in those games but 6% uplift in freq. wouldn't lead to that bigger scores,so it would fall in line with Phenom II @ 3Ghz(again back to GPU limitation @ 16x10,for 100th time).

How about we wait for real reviews instead this one which isn't online anymore? I want to see how all 4 CPU generations perfrom in real gaming situations,with SLI/CF or 4870X2/GTX280,in high res. and full details.hat's how you can measure how one chip performs and will it suit your needs in this type of usage.

[Zucker2k]

Calm down,are you getting excited over there?

You may have exposed yourself.I made a clock for clock comparison because you asked me to do it(remember the math comment?).You seem to not understand that games at 16x10 are not CPU limited.CPU helps though,but as you can see,a cache+core uplift means more than pure Mhz uplift(PhII vs PhI).

Second,again you asked me why no Q9650 in the comparisons and when i give you logical response you go all wild with *closed ears with both hands* na-na-na-na comment .
I told you why there is no comparison with Q9650.I can tell you it would fare better in those games but 6% uplift in freq. wouldn't lead to that bigger scores,so it would fall in line with Phenom II @ 3Ghz(again back to GPU limitation @ 16x10,for 100th time).

What bigger? Are you pretending to not see the fractions between PII 940 and Q9550? Price-wise, the Ci7 940 is not in the same price segment as the Q9550. There, your price argument is out the window what next?

Do you see fractions too?

STALKER CLEAR SKY, 1680x1050

Phenom II x4 940 3Ghz = 41.7
C2Q Q9550 2.83Ghz = 41.2

.5

CRYSIS WARHEAD, 1680x1050

Phenom II x4 940 3Ghz = 35.7
C2Q Q9550 2.83Ghz = 35.0

.7

DEAD SPACE, 1680x1050

Phenom II x4 940 3Ghz = 219.0
C2Q Q9550 2.83Ghz = 214.4

4.4 (about 1.8%)

FAR CRY 2, 1680x1050

Phenom II x4 940 3Ghz = 54.4
C2Q Q9550 2.83Ghz = 46.2

8.2 (16%)

WORLD AT WAR, 1680x1050

Phenom II x4 940 3Ghz = 48.7
C2Q Q9550 2.83Ghz = 57.3

8.6 (16%+)

WORLD IN CONFLICT, 1680x1050

Phenom II x4 940 3Ghz = 92.7
C2Q Q9550 2.83Ghz = 95.0

2.3

[informal]

It's like arguing with a tree.I really see no point here.You fail to understand the basics.My opinion is that Core i7 was never the intended chip to compete with from price perspective.Maybe the lowest part(i920) was.The charts we saw(they are offline now) were representing 4 CPU generations. It's clear since all chips were there apart from 65nm C2Q.Agena was there for PhII comparison (and we see the core +cache improvements nicely),C2Q and i7 were 2 latest fom intel.Last one is there simply because it is released (it would be stupid to miss out a whole gen. in one review) and is similar in clock to all the chips since 920 would be too low looking at the def. clocks.Q9550 is 6% away in clock terms and closest in price as things stand right now since PhII 940 will cost around 265-275 US Dollars.Q9550 is 250 bucks.

Again you quote yourself for who knows what time in vain.I did the math since you asked me to do it.Remember.It's 16x10,CPU plays smaller role and therefore you see smaller differentiation when chips with similar cache size and clock are in question(12MBs vs 8Mbs,2.83Ghz vs 3Ghz).Agena falls a bit short because it lacks those larger caches and CPU clock,hardly surprising eh?

Tell me,what kind of results from Q9650 you expect to see in 16x10 in these games?Do you think it will score 6% faster than Q9550 in these settings?Hmm k.
Still you failed to see RV770 vs GT200 similarity and not responded to it(yes Phenoms will be sub 280$ region,apart from early price jacking that always happen).

[qurious63ss]

First of all,we have obscure slides with no configurations at all.Turbo is out of the box feature and is always on.It is just as SMT ,a built in feature.

Second,your premise is false.CPU effect is not = 0.Therefore your conclusion is invalid.CPU effect is not as important in 16x10 as it is in 800x600(in which btw no1 games).In low res. you could probably see some differences that come from game engine optimizations(compilers,datasets used etc.).

In which games PhII doesn't fare any better than Phenom I?In the GPU limited ones where all CPUs have closely grouped scores?

As for the last question about slides and pricing,did it occur to you Q9650 is not priced the same as PhII 940?? AMD done-if the slide is from AMD in the first place- the price per. comparison by comparing systems that will cost similarly.PhII 940 will not cost the same as Q9650 and is logical not to compare it with that model. Have you heard of Radeon 4850/70 card? You know that 4870 model is at 80-85% the perf. of GT200(GTX280),somewhere even higher ,while costing A LOT less. That's called great value.

I don't think the reason they don't compare to Q9650 is because of price difference but because of performance difference. If the PII 940 was able to be competitive with the Q9650 then you would see Amd showing slides left and right that their ~$300 part can compete with a ~$600 Intel part.

On the second part, the 4870 cost less then the GT200 because the die is less then half as big 250mm2 vs 600mms, thus cost less to manufacture. Unfortunely for Amd their is no advantage there in the CPU front since the k10 is 260mm2 and the c2q is 214mm2.

[charged3800z24]

GUYS!!!! what I find most interesting about the scores, is how much better Phenom II was over Phenom.. I know. I know.. 400MHZ more... but stock.. it is finally a competive clock...that in its self is a much desirable improvement.. get AMD in the same playing field finally....Added Unlocked (again) multi...it is a no brainer OC...cheap swap in for current owners.. and cheaper then i7 ...so it will finally be a option to Intel (compitition is grand )

[Zucker2k]

It's like arguing with a tree.I really see no point here.You fail to understand the basics.My opinion is that Core i7 was never the intended chip to compete with from price perspective.Maybe the lowest part(i920) was.The charts we saw(they are offline now) were representing 4 CPU generations. It's clear since all chips were there apart from 65nm C2Q.Agena was there for PhII comparison (and we see the core +cache improvements nicely),C2Q and i7 were 2 latest fom intel.Last one is there simply because it is released (it would be stupid to miss out a whole gen. in one review) and is similar in clock to all the chips since 920 would be too low looking at the def. clocks.Q9550 is 6% away in clock terms and closest in price as things stand right now since PhII 940 will cost around 265-275 US Dollars.Q9550 is 250 bucks.

Again you quote yourself for who knows what time in vain.I did the math since you asked me to do it.Remember.It's 16x10,CPU plays smaller role and therefore you see smaller differentiation when chips with similar cache size and clock are in question(12MBs vs 8Mbs,2.83Ghz vs 3Ghz).Agena falls a bit short because it lacks those larger caches and CPU clock,hardly surprising eh?

Tell me,what kind of results from Q9650 you expect to see in 16x10 in these games?Do you think it will score 6% faster than Q9550 in these settings?Hmm k.
Still you failed to see RV770 vs GT200 similarity and not responded to it(yes Phenoms will be sub 280$ region,apart from early price jacking that always happen).

I wish objective people reading this thread would come forward and offer their opinions. It is obvious Informal is picking and choosing as he wishes. It's now a generational comparison, way to go. This is where I stop however, because your arguments are like moving targets, they're shifting by the minute. I'm out b4 this degenerates into name-calling. So why is there no 65 nm C2Q? uh uhuh.....

[informal]

Whatever.You asked me to do the math i did it for you.You asked for my opinion(that's my opinion not a fact) about Q9650 and you ignored it.You fail to understand games are mostly GPU bound at higher resolutions and that CPU plays smaller role(but yet not unimportant as you can see from the charts).Now you say you would go for name calling(i say you would go since i wouldn't for sure).I rest my case.

As for die are comment by qurious63ss ,there is more to it.It's double patterning(intc) versus immersion lito(amd) ;DFM advantage intel had before is erased when AMD moved to 45nm(they use DFM now too),although AMD used more metal layers with Barcelona and i'm not sure what is the case with Core i7 now. Penryn for sure is cheaper to make since it's MCM.Core i7 i can't say. AMD will have Propus core with no L3,a budget one if you like,and it will be a lot smaller leveraging small core size in Shanghai.This one will be the "money maker" chip since it will be small and relatively fast.

[Yokozuna]

Anybody saw this ?

http://forums.vr-zone.com/showthread.php?t=359342

http://www.freeimagehosting.net/imag...ea01696aad.jpg
http://www.freeimagehosting.net/imag...fcf31169d9.jpg
http://www.freeimagehosting.net/imag...6e1446313a.jpg
http://www.freeimagehosting.net/imag...a4e1257410.jpg
http://www.freeimagehosting.net/imag...b79b46edf1.jpg
http://www.freeimagehosting.net/imag...3bb1b138aa.jpg

Source seems to be expreview, but not longer available

These charts look as phony as the fabricated Deneb benchmarks posted exactly one week ago on the "markbench" blog, which, coincidentally(?), were also first propagated on the VR-zone forums. by a new poster.

[Frontl1ne]

^ The charts do look dodgy but the guy who posted said he got it from xpreview and that apparently they're missing from xpreview now. I don't see why anyone would want to fake those scores like that where the Deneb gets outclassed in some games, yet manages to beat the i7 in others. Kinda pointless since you're not really showing that one cpu beats another :S

[Ghostbuster]

Looks dodgy alright...

Silentjack07
Registered User

Join Date: Nov 2008
Posts: 2

And those 2 posts were all here http://forums.vr-zone.com/showthread.php?t=359342

[informal]

Looks dodgy alright...

Silentjack07
Registered User

Join Date: Nov 2008
Posts: 2

And those 2 posts were all here http://forums.vr-zone.com/showthread.php?t=359342

Good points.Seems very fishy.

[Macadamia]

^ The charts do look dodgy but the guy who posted said he got it from xpreview and that apparently they're missing from xpreview now. I don't see why anyone would want to fake those scores like that where the Deneb gets outclassed in some games, yet manages to beat the i7 in others. Kinda pointless since you're not really showing that one cpu beats another :S

But we do know i7 actually has performance issues with high(++) settings in Far Cry 2.

At first I thought it was a single reviewer's anomaly, but I'm seeing more sources being coherent in this one.

[Ghostbuster]

But we do know i7 actually has performance issues with high(++) settings in Far Cry 2.

At first I thought it was a single reviewer's anomaly, but I'm seeing more sources being coherent in this one.

Mind if you show us the sources...

[Macadamia]

http://www.bit-tech.net/hardware/200...-core-i7-920/7
http://enthusiast.hardocp.com/articl...50aHVzaWFzdA==
http://www.techspot.com/review/124-i...65/page11.html

[Ghostbuster]

This one http://www.bit-tech.net/hardware/200...-core-i7-920/7 contradicts with the so-called (fishy/dodgy) "benchmarks" http://www.freeimagehosting.net/imag...a4e1257410.jpg

Note its using the same graphics card also, GTX280, and same resolution 1680x1050.

[Motiv]

just out of interest, are these shanghais for SpecInt?

http://www.spec.org/cpu2006/results/...110-05957.html

[Zucker2k]

just out of interest, are these shanghais for SpecInt?

http://www.spec.org/cpu2006/results/...110-05957.html

Yes.