Crysis CPU benchmark: QX6850 VS QX9650 VS PhenomX4

[STaRGaZeR]

Well, how about the fact that a fully overclocked Phenom @ 3GHz is still 7.5% slower than a STOCK 3GHz C2D/C2Q that can be overclocked to 3.5GHz+ for 65nm and 4GHz+ for 45nm?

I'd say AMD has had its ass handed to it on a silver platter.

QFT. OCed X4 vs stock C2D and Yorkfield. And still pwned. Intel wins in both sides, clock per clock perfomance AND clock speeds, by a wide margin.

This sucks, who gave me my future Yorkfield at good prices then?

[Shadowmage]

why is everyone so obsessed with the fsb... if it brings more than 1-2% performance increase i would be surprised. (notice how much core2 gained from 266 to 333...)

It's not the FSB. The NB controls important factors like the L3 cache speed. In this case, the L3 cache is effectively running at ~50% speed (3GHz/1.6GHz).

[AbelJemka]

Well, how about the fact that a fully overclocked Phenom @ 3GHz is still 7.5% slower than a STOCK 3GHz C2D/C2Q that can be overclocked to 3.5GHz+ for 65nm and 4GHz+ for 45nm?

I'd say AMD has had its ass handed to it on a silver platter.

QFT. OCed X4 vs stock C2D and Yorkfield. And still pwned. Intel wins in both sides, clock per clock perfomance AND clock speeds, by a wide margin.

This sucks, who gave me my future Yorkfield at good prices then?

1. Your assumption is based on the fact that 3GHz is the max overclocking of Phenom : we don't know.
2. Those numbers are based on the fact that the test is doing with final platform for all : we don't know.

So with two hypothesis you draw a conclusion, that's funny.

[freeloader]

Disappointing performance by AMD.

Note that cache and DIV/SQRT/SHUF seem to have no effect in Farcry. The only thing I can think of is branch prediction

Under 5% difference, and that's disappointing.

[freeloader]

great, it works! but isn't that what you'd expect from a multi-billion dollar company? or is there something special about amd

I'd expect it to work and seems to be only 5% or less of an Intel quadcore. If there was a $100 difference between an Intel quadcore and AMD quadcore, i'd suck up the 5% and grab the AMD. 5% isn't even noticeable in 90% of the applications out there.

[Epsilon84]

1. Your assumption is based on the fact that 3GHz is the max overclocking of Phenom : we don't know.
2. Those numbers are based on the fact that the test is doing with final platform for all : we don't know.

So with two hypothesis you draw a conclusion, that's funny.

1. Yeah, cause Phenom is gonna clock to 4GHz+ with the next miracle stepping! Anyone care for a dance in the aisles with our friend Charlie from TheINQ?

2. Both are pre-release chips... so either AMD is leaving performance tuning very late or you are just clutching at straws here. Do you really expect a new BIOS to give a 10% clock for clock gain for K10?!

You know what else is funny? You clinging to the hope that an overclocked Phenom will have any chance against an overclocked Yorkfield.

Clock for clock disadvantage / lower headroom != performance parity

[hollo]

amd fanpoi: yorkfield's only 5% faster than k10 @ 3ghz
intel fanboy: yorkfield overclocks to well over 4ghz
amd fanpoi: this is madness..
intel fanboy: madness? THIS IS XS!!! *kick

[TimoneX]

Additionally pricing & power consumption are vital factors in this equation and they're virtual unknowns at this point. Price/performance is what really matters to most, not just performance. So what if AMD is a little slower than Intel in the next round? They've been behind for over a year. If AMD can somehow price phenom so that it's competitive w/ Intel relative to the performance then phenom will be at least partially successful. We ALL need AMD to get competitive again or Intel will cut their product lines substantially, jack up the prices, and stop innovating. It's silly to want AMD to fail as it would almost certainly hurt all us technofiends.

[AbelJemka]

1. Yeah, cause Phenom is gonna clock to 4GHz+ with the next miracle stepping! Anyone care for a dance in the aisles with our friend Charlie from TheINQ?

2. Both are pre-release chips... so either AMD is leaving performance tuning very late or you are just clutching at straws here. Do you really expect a new BIOS to give a 10% clock for clock gain for K10?!

You know what else is funny? You clinging to the hope that an overclocked Phenom will have any chance against an overclocked Yorkfield.

Clock for clock disadvantage / lower headroom != performance parity

1. Cause u know someone with a Phenom rigth now who can prove that this can't happen?

2. Penryn is out right now no? And u carefully read this thread some already talk about the fact that the platform is far from optimum.

U go on being funny for sure. I'm not hope because I have actually seen nothing of Phenom.

[mongoled]

blah blah blah blah blah

man how far stuck up our own arses are some of us



The arguments in this thread are so ridiculously stupid that its hard to swallow that XS has been invaded by rabid fanbois.

amd: mines better than yours
Intel : no mines better
amd: nope mine mine mine mine
Intel: no no no its miiiiiiiiine!!!!
amd: noooooooooo

The one thing that sticks out from this benchmark is that there is a bottleneck IRRESPECTIVE of the setups used! It couldnt be plainer to see, yet, we all decide to ignore this fact and base our 'discussions' on ifs, buts and ASSUMPTIONS on things that have not been released.

How stupid is that?

Lets look at the numbers.......

If we pretend that the test was only between Intel CPU's then we would see the following

Benchmark---Intel E6850 ---------Intel QX6850--- Intel QX9650
Frequency---3Ghz (333mhz x 9)---(333mhz x 9)---(333mhz x 9)
Run0--------44.01----------------45.14----------45.46
Run1--------50.42----------------51.37----------51.24
Run2--------51.12----------------51.63----------51.55
Run3--------51.21----------------51.55----------51.56
Average-----49.19----------------49.92----------49.95

HELLOOOOOOOOOOOOOOOO

How obvious is the bottleneck is NOT in the CPU.

Yet still just because the results of a K10 were added that where at best 5% slower there are people posting here as if amd had released a K10 with the performance of a K6.



Way to go einsteins........

Hopefully some sanity will return to XS, doubtfull as there are too many tub thumpers here trying to cloud neutrals opinions on unreleased hardware. Personally its getting tiring and saddening, XS never used to be like this



Once the K10 platform has been out for 6 months will we really know what the true picture is, at the moment some people are so hell bent on muddying the waters and i cant figure out why.....

[STaRGaZeR]

1. Your assumption is based on the fact that 3GHz is the max overclocking of Phenom : we don't know.
2. Those numbers are based on the fact that the test is doing with final platform for all : we don't know.

So with two hypothesis you draw a conclusion, that's funny.

1. My assumption is based on the fact that Intel processors are at stock with a huge OC headroom (we already know this) and the AMD processor is overclocked and we know that there wont be more than 2,5Ghz Barcelonas at launch, when Intel has its 3Ghz pieces out there now.
2. Agree, we don´t know details, but this is the same for both sides. So no problems here.

The funny thing is that you and a bunch of people act like blind people. Phenom will fail. Not even close to a year+ old uArch, Intel Core, and far away Penryn in power consumption and dissipated heat. (and clock per clock perfomance of course)

[RaZz!]

blah blah blah blah blah

man how far stuck up our own arses are some of us



The arguments in this thread are so ridiculously stupid that its hard to swallow that XS has been invaded by rabid fanbois.

amd: mines better than yours
Intel : no mines better
amd: nope mine mine mine mine
Intel: no no no its miiiiiiiiine!!!!
amd: noooooooooo

The one thing that sticks out from this benchmark is that there is a bottleneck IRRESPECTIVE of the setups used! I couldnt be plainer to see, yet, we all decide to ignore this fact and base our 'discussions' on ifs, buts and ASSUMPTIONS on things that have not been released.

How stupid is that?

Lets look at the numbers.......

If we pretend that the test was only between Intel CPU's then we would see the following

Benchmark---Intel E6850 ---------Intel QX6850--- Intel QX9650
Frequency---3Ghz (333mhz x 9)---(333mhz x 9)---(333mhz x 9)
Run0--------44.01----------------45.14----------45.46
Run1--------50.42----------------51.37----------51.24
Run2--------51.12----------------51.63----------51.55
Run3--------51.21----------------51.55----------51.56
Average-----49.19----------------49.92----------49.95

HELLOOOOOOOOOOOOOOOO

How obvious is the bottleneck is NOT in the CPU.

Yet still just because the results of a K10 were added that where at best 5% slower there are people posting here as if amd had released a K10 with the performance of a K6.



Way to go einsteins........

Hopefully some sanity will return to XS, doubtfull as there are too many tub thumpers here trying to cloud neutrals opinions on unreleased hardware. Personally its getting tiring and saddening, XS never used to be like this



Once the K10 platform has been out for 6 months will we really know what the true picture is, at the moment some people are so hell bent on muddying the waters and i cant figure out why.....

well said.

seems like bashing amd is kind of a temporary fashion or smth...

back then i joined xs due to it's mostly objective and informative posts and threads. i mean, there always were fanboys etc, but tbh, in the last months/year this increased exponentially and became absolutely annoying.

and no, i'm not an amd fanboy or smth, but it really hurts me if people say smth like "amd is no more due to 2012 LOLZ!... ROFL AMD, DIE!".
wtf, i'd never say something like that about neither intel nor amd. grow up.
furthermore, whoever of them, amd or intel, closes down, we, the customer, would lose. that's for sure...

[BeardyMan]

AbelJemka,

Yorkfield is not released, the release date is 12th of November. In the benchmark they were using P35 and DDR2 800. There are mainboards with X38 chipset which is faster, as well as DDR3 1333MHz CL5 and DDR2 1066 CL4(supported by both P35 and X38). There is going to be another generation of Intel chipset for LGA775. So the platform used for Yorkfield ES is not even close to final.

K10 is good on 2P or more, just because the IMCs and the HTT links.
Face it AMD failed to deliver anything that can cause a little headache to Intel on the desktop.
Phenom will sux compared to Kentsfield, clock for clock. It'll never compete against Yorkfield. Phenom will need at least 2.7GHz to compete against the slowest Yorkfield(the 2.5GHz with 2x3MB L2) in performance. Yorkfield has no competition for energy efficiency and overclocking ability.

Yorkfield failed its a next nextgen product at 45nm and can't deliver more worthy performance then Phenom 65nm.

[NapalmV5]

STaRGaZeR, phenom will fail ?

whats tomorrows lottery numbers please ?

[DuceGT]

Pre-release CPUs and pre-release chipsets and a bunch premature adults on an internet forum.

[Jacky]

and no, i'm not an amd fanboy or smth, but it really hurts me if people say smth like "amd is no more due to 2012 LOLZ!... ROFL AMD, DIE!".
wtf, i'd never say something like that about neither intel nor amd. grow up.

even if it was true? Oh man are you blind or smth? Don't you see they are rapidly running out of money. If AMD doesn't start to deliver very soon, they have HUGE PROBLEMS. that I can promise you.

Q3 was a better than average quartal for Intel&AMD. Still AMD managed a net loss of 260-299M$ (don't remember the exact number) excluding the ATI purchase.
ATI can't make any money even though Q3 should have been quite good, they had almost their full line up ready, lower production node compared to nvidia, hd 2900xt getting better and better drivers - still they didn't even manage to break even! With the 8800gt they will hardly do much better..

The same applies to AMD, they probably already included some major Barcelona sales and still made a big loss.

The arguments in this thread are so ridiculously stupid that its hard to swallow that XS has been invaded by rabid fanbois.

No argument. We know that K10 won't be faster than penryn on the desktop. However it will lack SSE4, consume more power, overclock worse, bla, bla...

Barc is strong in 2p+ and in BW intensive apps. NOT on desktop/notebook (we have enough benchmarks to prove that, spec_fp/int, anandtech's extrapolation).
AMD may break even in Q3/Q4 2008 at the earliest.. that's their main problem.

[BrowncoatGR]

Power consumption increases with the square of V.

That's utter BS.SOI and metal gates address different factors of leakage , SOI is becoming less useful once you get to 65nm and lower.

The fact that SOI doesn't adress gate leakage , the biggest problem at 65/45/32nm means that AMD had to use thicker gates which equal slower transistors.

You dont know that. There is more to process design and materials than just fancy terms. AMD is supposed to soon implemen a new variant of SOI called SGOI that incorporates an eSiGe substrate and is supposed to have 40% increased electron mobility. AMD plans to use high-k and metal gates either later at 45nm or at the latest at 32nm. As for SOI runnng out of steam thats just an assumption(at best) you have made and no, AMD evaluating all of their options does not mean anything really. Any tech company will always do research before committing itself to any technology. It would have been shocking if AMD/IBM weren't evaluating all the options available to them

Huh ? Where do you get this info ?

1.5V is huge for a 65nm process ; 90nm used 1.35-1.4V.

I agree that 1.5V is huge for 65nm in fact i think its nowhere near default voltage and i doubt they tweaked their overclock at all. He is right about desktop Yorkfields though, but that is only because Intel is aggressively binning all good dies for Xeons so that they can keep the very low 80WTDP for them. So i see this TDP improving very quickly once production increases. This is a clear sign though that 45nm will not be high volume till at least Q2

[STaRGaZeR]

STaRGaZeR, phenom will fail ?

whats tomorrows lottery numbers please ?

I´m talking about final perfomance (CPC and clock speeds here), power consumption and heat disipated. No, it won´t fail in the meaning of "fail" you´re using, but they will be for mainstream. We want the best perfomance, and will never buy one of those if we want that perfomance. If you are a fanboy, well, you will buy every sh|t AMD throws at you. Every "bench" we have seen points in that direction. If you don´t thrust them, you will see it in a few months with your own eyes.

[Shintai]

blah blah blah blah blah

man how far stuck up our own arses are some of us



The arguments in this thread are so ridiculously stupid that its hard to swallow that XS has been invaded by rabid fanbois.

amd: mines better than yours
Intel : no mines better
amd: nope mine mine mine mine
Intel: no no no its miiiiiiiiine!!!!
amd: noooooooooo

The one thing that sticks out from this benchmark is that there is a bottleneck IRRESPECTIVE of the setups used! It couldnt be plainer to see, yet, we all decide to ignore this fact and base our 'discussions' on ifs, buts and ASSUMPTIONS on things that have not been released.

How stupid is that?

Lets look at the numbers.......

If we pretend that the test was only between Intel CPU's then we would see the following

Benchmark---Intel E6850 ---------Intel QX6850--- Intel QX9650
Frequency---3Ghz (333mhz x 9)---(333mhz x 9)---(333mhz x 9)
Run0--------44.01----------------45.14----------45.46
Run1--------50.42----------------51.37----------51.24
Run2--------51.12----------------51.63----------51.55
Run3--------51.21----------------51.55----------51.56
Average-----49.19----------------49.92----------49.95

HELLOOOOOOOOOOOOOOOO

How obvious is the bottleneck is NOT in the CPU.

Yet still just because the results of a K10 were added that where at best 5% slower there are people posting here as if amd had released a K10 with the performance of a K6.



Way to go einsteins........

Hopefully some sanity will return to XS, doubtfull as there are too many tub thumpers here trying to cloud neutrals opinions on unreleased hardware. Personally its getting tiring and saddening, XS never used to be like this



Once the K10 platform has been out for 6 months will we really know what the true picture is, at the moment some people are so hell bent on muddying the waters and i cant figure out why.....

How dare you post the obvious!

[mongoled]

How dare you post the obvious!

Hello Mr Shintai, who joined XS on Aug 2006 and has already racked up over 2000 posts.

Your persona fits perfectly in line with those who in my opinion have devalued the essense of XS. Although your knowledge evidently extends far beyond whats green and blue you also are the most frustrating type of poster.

How can someone who obviously has a respectable level of IQ be so hell bent on offering opinions that are so one sided?

What is your perogative?

In the past there was a discussion on XS regarding the alliances that certain forum members had with regards to companies that manufacture, market and sell certain products.

It seems that we have come round in a full circle and the same thing is happening again.

I cant accept the fact that someone as your self can be so biased towards a company without having a hidden agenda. I just wish there was a way that peeps posting with hidden agendas could be ousted and shamed for what they are doing to this community.

Only those responsible for XS can help put a stop to this. Otherwise more and more peeps who can discuss things on a logical level will stop posting here.......

Please note and READ that I have not directly said that YOU have a hidden agenda, I am sure though that there are peeps here who will know if this is the case or not.

[savantu]

sorry, 333fsb vs 200 ? nice try by anand

higher fsb on amd still makes a difference

also, dont forget, demo @ 1 thread

Huh ?

The 1600MHz HT links offers 6.4GBs of BW and low latency.That's more than enough since the memory traffic isn't there.

This is simply grasping for straws...

Another nice one : RAM running at 800 CL5.

Hello world : 99% of DDR2 800 in this world runs at CL5.
Hello world take 2 : using better RAM improves the scores.But this is also true for the Intel platform.

[savantu]

You dont know that. There is more to process design and materials than just fancy terms. AMD is supposed to soon implemen a new variant of SOI called SGOI that incorporates an eSiGe substrate and is supposed to have 40% increased electron mobility. AMD plans to use high-k and metal gates either later at 45nm or at the latest at 32nm. As for SOI runnng out of steam thats just an assumption(at best) you have made and no, AMD evaluating all of their options does not mean anything really. Any tech company will always do research before committing itself to any technology. It would have been shocking if AMD/IBM weren't evaluating all the options available to them

This is funny ( or pretty sad depending on how you see it ).

AMD plans to ramp 45nm in H1 2008 and is building CPUs as we speak on 45nm ( if we were to believe Ruiz ) , yet , 3 months ago they didn't know which way to go on 45nm.

As for the 40% claim , nothing is free my dear Watson.The ceteris paribus condition isn't fulfilled.They might get 40% , but the trade offs aren't shown.

As for High-K and metal gates , that's the most significant change in silicon technology in 40 years ; it's a turning point in addressing future scaling.
Claiming that it doesn't matter and will somehow be mitigated by herculean efforts with SGOI and low-k is just a refusal to accept the reality : IBM/AMD did not have the R&D , money and brainpower to implement this features in a timely manner.

I agree that 1.5V is huge for 65nm in fact i think its nowhere near default voltage and i doubt they tweaked their overclock at all. He is right about desktop Yorkfields though, but that is only because Intel is aggressively binning all good dies for Xeons so that they can keep the very low 80WTDP for them. So i see this TDP improving very quickly once production increases. This is a clear sign though that 45nm will not be high volume till at least Q2

Actually Xeons are binned conservatively , closet to their worst case scenario.Intel also uses slower , less leakier transistors because their offer better reliability.

As for the "clear sign" this might be obvious only to you since Intel plans to crossover to 45nm in Q3 2008.
Compared to AMD's volume , Intel's 45nm process will be high volume starting with this Q.

[Gambit_2K]

Huh ?

The 1600MHz HT links offers 6.4GBs of BW and low latency.That's more than enough since the memory traffic isn't there.

This is simply grasping for straws...

Another nice one : RAM running at 800 CL5.

Hello world : 99% of DDR2 800 in this world runs at CL5.
Hello world take 2 : using better RAM improves the scores.But this is also true for the Intel platform.

But it's also true that timings almost do nothing on an intel plattform but can do miracles for the AMD platform.

[savantu]

It's not the FSB. The NB controls important factors like the L3 cache speed. In this case, the L3 cache is effectively running at ~50% speed (3GHz/1.6GHz).

Is HT link speed=NB speed ?

AFAIK all K10 on Ht 2.0 will run their NB at 1.6GHz.

[BrowncoatGR]

This is funny ( or pretty sad depending on how you see it ).

AMD plans to ramp 45nm in H1 2008 and is building CPUs as we speak on 45nm ( if we were to believe Ruiz ) , yet , 3 months ago they didn't know which way to go on 45nm.

As for the 40% claim , nothing is free my dear Watson.The ceteris paribus condition isn't fulfilled.They might get 40% , but the trade offs aren't shown.

As for High-K and metal gates , that's the most significant change in silicon technology in 40 years ; it's a turning point in addressing future scaling.
Claiming that it doesn't matter and will somehow be mitigated by herculean efforts with SGOI and low-k is just a refusal to accept the reality : IBM/AMD did not have the R&D , money and brainpower to implement this features in a timely manner.

Please reread what i wrote. I did not say that metal gates and high-k do not matter. I said just because Intel has them does not make their process superior by default. There is a lot more to it. Also care to elaborate on the drawbacks of SGOI? The only one i can think of is increased manufacturing complexity. As for the last part

Actually Xeons are binned conservatively , closet to their worst case scenario.Intel also uses slower , less leakier transistors because their offer better reliability.

As for the "clear sign" this might be obvious only to you since Intel plans to crossover to 45nm in Q3 2008.
Compared to AMD's volume , Intel's 45nm process will be high volume starting with this Q.

What i meant is that Intel is using all the good dies for Xeons and therefore desktop products have a higher TDP. And Intel needs to have more fabs online to have high volume this quarter. It is a fact that they will start 45nm production at 2 more fabs this Q so Q1 should see increased availability and in Q2 they should be easy to find.