Overclocking: Core i7 Vs. Phenom II

[sohei]

you have an A

 

[totenkopf]

I don't mean to be an ungrateful critic but even though I do still like Tom's, I must admit that this article is teh poopoo.

Okay, they thought that they were giving us what we wanted with this article... how?! What we wanted was a meaningful comparison between the two CPU architectures (even though we know that core i7 is faster). They might as well have hit the p2 with a hammer before installing it.

First of all, as everyone has already said, the memory problem. Don't need to whip that horse. However, I am going to whip the other dead horse: the cooler. "Wah wah wah! we couldn't find a way to attach the TRUE to the am2 and my cat's breath smells like cat food!". How weak is that? Your argument is no longer sound. You invalidated your argument because your key premise stopped being true as soon as the systems stopped being as similar as they could be. Hell, If you couldn't use the TRUE for the am2 than you should have picked ANYTHING you had "laying around" that DID work on both CPUs instead of just handicapping one of them. Afterall, we are comparing the CPUs... I think. And don't throw that crap at me about the p2 having acceptable temps regardless, the test is still flawed.

Perhaps more disturbing is that you weren't willing to buy a ten dollar part (adapter/bracket for TRUE to am2) to protect the integrity of Toms's and more importantly your own integrity. Instead you release this drivel and try to pass it of as some kind of objective, well thought out experiment. Hire a tech editor... or at least fire the one you have.

Also, about the abysmal OC on the p2... if you don't have a representative sample of the p2's OC ability, either get one, or don't release the article. It makes Tom's look bad, and closer resembles libel than scientific method.

All of this is really moot because when am3 comes out in a couple months you'll get the test right, right? Or will there still be extraneous, and avoidable, variables in your "fair" test like different coolers, memory amount/speed, and two year old MOBO's? Or next time are you really gonna just hit the p2 with a hammer before you drop it in?

Unrelated to all of this: Just take down your charts. If they aren't up to date, they sort of lose their usefulness.

ps. don't hate me because you disappoint me. Also, I love reading the comments here (a lot of smart, informed people), Tom's audience is now the best part of their site (hint: start hiring, Tom's!).

 

[sohei]

A+

 

[chaos23]

Also, about the abysmal OC on the p2... if you don't have a representative sample of the p2's OC ability, either get one, or don't release the article. It makes Tom's look bad, and closer resembles libel than scientific method.

That's what most of us are trying to say. The first review/benchmark re: P2 matched up against c2d and core i7 was way better than this one (although they only used ddr2-800 memory modules).

Tom's should make another article/review regarding the OC capabilities of the Phenom using a decent HSF, a 790fx motherboard and a good overclocker, one that can really OVERCLOCK. Overclocking by adjusting only the multiplier, geez.. what do you think of that.

 

[Naw-yi]

the only test we should look at is the crysis game benchmark......
phenom wins it with a lower clock speed, ddr 2 and lower price.

this is the game that we are all trying to get the most fps possible isnt it? and amd just won this test....

 

[Naw-yi]

if all things were equal, the p2 would win.

 

[randomizer]

[citation][nom]Naw-yi[/nom]the only test we should look at is the crysis game benchmark......phenom wins it with a lower clock speed, ddr 2 and lower price.this is the game that we are all trying to get the most fps possible isnt it? and amd just won this test....[/citation]
People buy processors just for Crysis?

 

[Naw-yi]

yes they do.

 

[Naw-yi]

check the forums for builders asking if there new build will play crysis.

 

[randomizer]

So you'd buy a Phenom II soley because Crysis is bottlenecked by the GPU and therefore the CPU doesn't come into play as much? Are you trying to tell me that almost everyone uses their PC just to see if they can benchmark Crysis? You've got to be on something. People actually use their PCs for real work, maybe you should go check the forums for that too.

 

[Naw-yi]

dude, your saying that the 4870x2 is a bottleneck? is that why the i7 is slower? and this is tomshardware.com not openoffice.com, builders post here for their killer rigs that run the most demanding games, not there killer rigs that can render a resume the fastest at 1600x1200. get a clue.

 

[randomizer]

i7 is slower within margin of variation. Obviously you've never actually benchmarked Crysis, or you'd know that it varies quite alot with each run. And yes, I am saying that the 4870X2 is bottlenecking, that's why they perform almost exactly the same. Or did you look at the wrong graph?

Also, I guess you've never heard of video encoding, you know, things that people do besides playing Crysis

 

[spearhead]

[citation][nom]dagger[/nom]Are you implying core 2 quad is slower than 9950?[/citation]
Yes it is but by a very small margin in not even all benchmarks.
but the q6600 runs at stock 2,4ghz which is able to being overclocked to 3,6ghz vs the phenom 9950 2,6ghz at stock and can hit around 3,2ghz overclock. so this makes q6600 faster as an overclocker and scales better clock by clock. but 9950 and q6600 at stock are quite evenly matched.

 

[apache_lives]

[citation][nom]sohei[/nom]i tell you why....because amd cpu in this case will OC much futher and at 4+ ghz will be much closer or even past it the performance of nehalem ...and a amd cpu with a 2 yers old soket and with DDR2, OC better than I7 ,whit same performance or better?...can u imagine? all the triple channel fancy mobo ,cooler, RAM etc will worth nothing[/citation]

Gee lets see, it looses at stock to lower clocked intel chips using socket 775 - same age at the AM2, and as other benchmarks in the past have shown, the difference between ddr1 and ddr2, or ddr2 and ddr3 etc isnt much, if not a little loss and TRI channel - barely any gain from dual to tri channel - your scratching around looking for a reason to down intels win.

First it was the high caches of a core 2 duo, then the 45nm, now its tri channel memory - intel has owned performance for years now, excuses are crap.

On a different note, yes that AMD is competing much better then the original flop, but i still would prefer 8 threads - i know how it feels to have them as does any P4-HT and above owner (in the past).

 

[Guest]

I would really love to know what difference it would have made, if you'd put 4GB instead of 6GB of memory into the Core i7 test system (with 2*1GB + 4*512MB you still would reach triple channel memory access) (or at least put in 6GB into the AMD system, too).

It is really astonishing how thg favours the intel setup (high end vs medium end board, 3.8 vs 3.64 GHz, and the memory of course)... but on the over hand, I guess, intel just paid more...

 

[apache_lives]

Chris986 you seem to forget the concepts of ram - least amount of sticks the better, and the most matched etc, only an idiot would use that setup you listed.

Second if you have the $$$ for the intel system you wouldnt go 4gb, and the apps benchmarked there are mostly 32 bit so 4gb limited, plus on fresh installs so the difference would be minimal.

If the rigs were the same you fanboys would be bitching about something else.

Go check out the Core 2 Quad vs Phenom II benchmarks and see the AMD loose there and find another excuse to complain, and while your there see why the Phenom II @ 45nm and an extra 600mhz barely beats a 2 year old Q6600 etc.

 

[jwoollis]

I am suitably impressed with the new Phenom II processor but I think there is a small problem here. Comparing the AMD platform to the Intel Core i7 Platform is like comparing an Average Family Saloon to a Porche!

While comparing the best that both camps have is all that can be done it really isn't a fair comparisson.

If AMD could push out samples of it's Socket AM3 Phenom II processor and motherboard and these were tested with the best comparable resources e.g. memory GPU etc, then I think we would probably see a different story.

At present the AMD system is bottlenecked by bandwidth limitations where the Intel Core i7 is not so the AMD system will not overclock so well. Increase the headroom so AMD and Intel or on an even playing field and AMD may well thrash Intel.

But careful, AMD has an excellent opportunity to recover and possibly even win back the performance crown. However if AMD does not push faster and get these latest technologies out of the door yesterday, then Intel is going to have time to refine and improve it's own technology and make another huge leap furthur ahead of AMD.

AMD has to speed up the release of it's new technologies, at least until it has a confortable lead on Intel - and that means they should already be working on 32nm technologies like Intel rather than remaining 1 or 2 generations behind Intel.

 

[apache_lives]

[citation][nom]jwoollis[/nom]I am suitably impressed with the new Phenom II processor but I think there is a small problem here. Comparing the AMD platform to the Intel Core i7 Platform is like comparing an Average Family Saloon to a Porche!While comparing the best that both camps have is all that can be done it really isn't a fair comparisson.If AMD could push out samples of it's Socket AM3 Phenom II processor and motherboard and these were tested with the best comparable resources e.g. memory GPU etc, then I think we would probably see a different story.At present the AMD system is bottlenecked by bandwidth limitations where the Intel Core i7 is not so the AMD system will not overclock so well. Increase the headroom so AMD and Intel or on an even playing field and AMD may well thrash Intel.But careful, AMD has an excellent opportunity to recover and possibly even win back the performance crown. However if AMD does not push faster and get these latest technologies out of the door yesterday, then Intel is going to have time to refine and improve it's own technology and make another huge leap furthur ahead of AMD.AMD has to speed up the release of it's new technologies, at least until it has a confortable lead on Intel - and that means they should already be working on 32nm technologies like Intel rather than remaining 1 or 2 generations behind Intel.[/citation]

AMD doesnt really have bottleknecks, thats probably there whole problem - they cant just change something obvious to get performance like change the bus or integrate something else etc

 

[jwoollis]

AMD doesnt really have bottleknecks, thats probably there whole problem - they cant just change something obvious to get performance like change the bus or integrate something else etc

I don't agree - If AMD did not have bottlenecks they would win everything.

Here you are comparing a 1st Gen 45nm AMD CPU on a chipset which is 1 generation behind Intel based on Dual Channel DDR2 Memory whereas Intels CPU is 3rd Generation 45nm, a complete redesign of the core architecture on a completely new chipset which removed many of the old core architecture limitations making it closer to AMD's fundamental design.

AMD needs to bring it's chipset up todate by introducing AM3 which is DDR3 (not sure if it will be dual/triple/quad channel?) That will bring AMD inline with Intel Core i7

Intel's next step within 12 months or so will be to introduce 35(ish)nm CPU's and within 24 months to introduce a new design from Core i7.

AMD has to catch up and overtake Intel meaning we need AM3 now, and 35(ish)nm CPU's within 12 months.

AMD has always had more efficient designs using less cache which has meant it has been able to meet or even beat Intel with older technologies, now that Intel has Core 2 and Core i7, though AMD cannot expect to operate like that exclusively unless it can come up with a radical new design and I don't just mean integrating GPU and other external components onto the CPU.

 

[Guest]

lol amd cooling is shitsy intel gets brand
look this
http://www.custompc.co.uk/news/605226/phenom-ii-overclocked-to-6ghz.html

 

[Guest]

http://www.nordichardware.com/news,8403.html
here overclokers gets ph2 6 ghz stable..

 

[kschoche]

[citation][nom]apache_lives[/nom]Gee lets see, it looses at stock to lower clocked intel chips using socket 775 - same age at the AM2, and as other benchmarks in the past have shown, the difference between ddr1 and ddr2, or ddr2 and ddr3 etc isnt much, if not a little loss and TRI channel - barely any gain from dual to tri channel - your scratching around looking for a reason to down intels win.First it was the high caches of a core 2 duo, then the 45nm, now its tri channel memory - intel has owned performance for years now, excuses are crap.On a different note, yes that AMD is competing much better then the original flop, but i still would prefer 8 threads - i know how it feels to have them as does any P4-HT and above owner (in the past).[/citation]

While I will agree that the PhenomII was beaten hands-down. However, your premise that faster memory doesnt affect performance is entirely wrong. Every single test, save for the gpu tests and the in-core INT/"FP"/SSE/etc tests, rely heavily on memory bandwidth.
Anytime you do a computation that doesnt fit inside cpu cache, you hit the memory and especially for streaming and compression, the lowest latency and highest bandwidth for large memory accesses (DDR3 >> DDR2 >> DDR) will help immensely.

It may not account for a 100% performance increase as easily as say, 8 threads instead of 4 threads in a trivially parallelized benchmark, but it adds quite a bit.
Assuming AMD didnt botch their DDR3 controller on-die, you should be able to see a noticeable difference, though I still would put its performance below that of the Intel chips here.
Once again, its not going to beat it, and the 4G vs. 6G comments are idiotic unless tests are using more than 4G of memory (I dont see any). But the speed of the memory makes a pretty large difference.

 

[squatchman]

http://www.nordichardware.com/news,8403.html
here overclokers gets ph2 6 ghz stable..

I'm not going to defend Tom's people working with hardware, but are you really going to compare an LN2 clock speed to an Air-cooled one?

 

[jwoollis]

[citation][nom]kschoche[/nom]While I will agree that the PhenomII was beaten hands-down. However, your premise that faster memory doesnt affect performance is entirely wrong. Every single test, save for the gpu tests and the in-core INT/"FP"/SSE/etc tests, rely heavily on memory bandwidth. Anytime you do a computation that doesnt fit inside cpu cache, you hit the memory and especially for streaming and compression, the lowest latency and highest bandwidth for large memory accesses (DDR3 >> DDR2 >> DDR) will help immensely.It may not account for a 100% performance increase as easily as say, 8 threads instead of 4 threads in a trivially parallelized benchmark, but it adds quite a bit. Assuming AMD didnt botch their DDR3 controller on-die, you should be able to see a noticeable difference, though I still would put its performance below that of the Intel chips here.Once again, its not going to beat it, and the 4G vs. 6G comments are idiotic unless tests are using more than 4G of memory (I dont see any). But the speed of the memory makes a pretty large difference.[/citation]

I think comparissons of memory have to be considered not just in volume 4G vs 6G but in the technology and the manner in which it is implemented.

In this case comparing 4GB of DDR2 memory to 6GB of DDR3 memory is flawed since for the most part DDR3 memory when appropriately used and identically implemented will outperform DDR2 memory.

However in this case the memory is not identically implemented. In this case DDR2 memory is accessed in pairs doubling the potential speed of DDR2 memory whereas faster DDR3 memory is accessed in sets of three, tripling the speed of access of DDR3 memory.

I am sure that Phenom II when equiped with DDR3 memory in the same manner as Intel Core i7, then I am sure that Phenom II would overclock more substantially and reliably and there is a strong likelihood that Phenom II might also outperform Intel Core i7 at most if not all levels.

 

[jwoollis]

I neglected to suggest that since the Phenom II cannot yet be equiped to the same level as the Intel Core i7 then perhaps the tests should be repeated with the Intel Core i7 downgraded by reducing memory profile of system to match the Phenom II System i.e. 4GB memory at DDR2 speeds. This is possible as the chipset supports almost any combination of DDR3 memory.

While this defeats the purpose of overclocking the systems, it does at least provide us with an idea of how the systems would perform on an even basis.

Overclocking test may be performed when AMD Socket AM3 is released which supports DDR3 memory