Sandy Bridge vs Nehalem vs Bulldozer vs Piledriver {BENCHMARKS!!!}

[Artas1984]

Remember this?

That was a review targeted to find out which CPU is the best for gaming in the specific price range. It was a great article, but not enough for me, as i wanted to find out how would the processors compare at the same speed, never mind pricing. A year passed since then and since Bulldozer failed us all, i now update that situation with the new Piledriver and look at the things again from other perspective.

So now you have opened a benchmark thread where 4 different processors will battle each other in various tests. All processors will be clocked at the same speed of 4 GHz - this means that the benchmark is not about CPU vs CPU, but rather architecture vs architecture.

This benchmark came out of my personal need to find out which processor i will be keeping for myself - which CPU fits my needs most, does not bottleneck my gaming experience at 1920X1200 and deals with my programs the fastest!

Representing Sandy Bridge - Core i7 3820



Representing Nehalem - Xeon W3520 (Core i7 920)



Representing Bulldozer - FX6200



Representing Piledriver - FX6300



Why not AMD FX 8 core CPUs? Because in the future i might want to throw the Phenom II X6 and keep it competitive with AMD FX 6 core CPUs.

Most of hardware used for CPU testing was kept the same for all processors:

GeForce GTX670 4 Gb video card .
4 DIMMs of 2 Gb DDR3 1333 MHz CL7 RAM & 4 DIMMs of 2 Gb DDR3 1600 MHz CL8 RAM.
Hitachi Deskstar 1 Tb 7200 rpm sata2 hard drive
Dell 24 inch 1920x1200 resolution monitor

All the software used for testing was identical for all 4 processors:

Windows 7 Pro X64
Forceware 306.97
Passmark 7 X64
Winzip
Irfan View
Adobe Photoshop CS2
Virtual Dub X64
20 games - 20 different engines

-----------------------------------------------------------------------------------------------------------------------------------------------------------------

Let's start with a synthetic benchmark and see what prognoses does it tell for the processors.

PASSMARK 7 X64 - calculating CPU, all tests, fast



According to Passmark 7, the new Piledriver will wipe the floor with both Bulldozer and Nehalem, even beat Sandy Bridge...!...

-----------------------------------------------------------------------------------------------------------------------------------------------------------------

Now let's toss in some real life applications which i myself use!

WINZIP - custom file compression



Looks like the Piledriver is a minor upgrade to the Bulldozer, but both are nowhere close to Intel processors.

-----------------------------------------------------------------------------------------------------------------------------------------------------------------------

Let's now extract that same compressed file!

WINZIP - custom file extraction



That was really impressive on AMD part! But most impressive is the consistency in time! I guess here extra cores do matter...

------------------------------------------------------------------------------------------------------------------------------------------------------------------------

I use Irfan View as my main photo viewing program. For it's small same it is the most powerful program that i could think of. One of the tasks i had to do with it was to convert a massive photo archive from jpeg to gif.

IRFAN VIEW - batch conversion from jpeg to gif



Piledriver is a massive upgrade to Bulldozer here, but not really in the league with Intel processors.

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------

I use the good old Adobe Photoshop CS2 for the most difficult photo tasks and it is enough for me. One of the more simple tasks i wrote in Photoshop CS2 for this benchmark was a simple photo editing command, that included filtering to diffuse glow, resizing and rotating the custom photo archive made of 950 photos.

ADOBE PHOTOSHOP CS2 - photo archive editing



The biggest surprise here is the incredible performance of my Nehalem based Xeon, which beated the newer Sandy Bridge CPU considerably. I have repeated this several times to the same result. Finally the Piledriver here is an upgrade to Bulldozer and scores the same as the Sandy Bride.

-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Most of all programs i use Virtual Dub to compress raw gaming videos. X264 and ffdshow codecs are my top 2. Let's see how the processors deal with this task.

VIRTUAL DUB X64 - compressing raw to X264



Really disappointed with the Piledriver here.. Perhaps it just does not like X264 codec...

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Typically for avi i use ffdshow, not X264, so let's see how the processors deal with it. Perhaps Piledriver will redeem itself here?

VIRTUAL DUB X64 - compressing raw to ffdshow



There is not really much difference here, though Sandy Bridge is clearly superior to all.

-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Let's game!
Here are included 20 games and each one of them uses a different engine! Now, typically when you want to find out which processor is the best in gaming, you lower the resolution and disable all effects right? But actually who games at 800X600?.....................
I game on 1920x1200 and since my GTX670 can handle that resolution and all effects easy, my biggest concern (as should be yours) is - will the processors be able to keep up with GTX670 at the resolution and settings i want to play! All games have been tested on max settings, 4X AA, 1920X1200 resolution.


ALIENS VS. PREDATOR 3 (Asura engine)



First game is a fluke - there is no real difference between the processors. This is one of the few "equilibrium" exceptions

ALAN WAKE (MAX-FX 3.0 engine)



The Piledriver offers absolutely no improvement on Bulldozer and both AMD processors get beaten by the 4 year old Nehalem. Even further from the rest sits the Sandy Bridge with it's mighty minimal frame rate!

AVATAR (Dunia engine)



The Piledriver is a small improvement on Bulldozer, but both AMD processors lie quite further from Intel.

CRYOSTASIS (AtmosFear 4 engine)



The Piledriver gains on Bulldozer and keeps up with Intel processors!

CRYSIS 2 (Cry 3 engine)



Bulldozer really suffered in Crysis 2 and Piledriver managed to redeem that shame by keeping up with Intel processors!

DEUS EX HUMAN REVOLUTION (Crystal engine)



Nothing to comment here really.

FEAR 3 (Despair engine)



Another game that could not care less about CPU

Formula 1 2011 (EGO 2.5 engine)



The Piledriver is a definitive improvement over Bulldozer, but not quite up with Intel processors.

HAWX 2 (unknown engine)



The Piledriver is more than i could have asked for in this game! Unfortunately for AMD this is the only game where the Piledriver manages to beat both Intel processors! Bulldozer is just pathetic here!

HARD RESET (Road Hog engine)



Not only does the Piledriver offer no improvement over Bulldozer, but it get's totally trashed by both Intel processors.

HOMEFRONT (Unreal 3 engine)



Once again the Piledriver offers no improvement over Bulldozer, and once again it get's trashed by both Intel processors. The difference, having in mind that Nehelem is so much older, is just scary!

LOST PLANET 2 (MT Framework)



The Piledriver just slightly improves Bulldozer's frame rate, but not in the class of Intel.

METRO 2033 (4A engine)



I regret adding this game to the list. Worst benchmark game of all time.

NECROVISION (Pain engine)



Older engine games are just fine for testing processors as well!
Piledriver is not much of difference from Intel, but clearly no improvement over Bulldozer.

PORTAL 2 (Source engine)



Another older engine game makes life easy for both Intel processors with absolute max performance possible out of this engine, yet AMD struggles! At least the Piledriver is an improvement over Bulldozer!

RED FACTION ARMAGEDDON (Geo-Mod 2.5)



The Piledriver is a small improvement over Bulldozer, yet once gain does not have the steam to push the Intel from the crown.

SERIOUS SAM 3 (Serious 3.5 engine)



Piledriver tries hard to get the shame off from it's predecessor by adding at least some extra FPS, but quite far from Intel.

SYNDICATE (Starbreeze engine)



Piledriver very well done in this game, being quite far from Bulldozer and not far from Intel!

STALKER CLEAR SKY (X-ray 1.5 engine)



The difference between Bulldozer and Piledriver is the difference between unplayable and playable as Piledriver manages to keep that critical minimal FPS where it needs to be!

WORLD IN CONFLICT (MassTech engine)



Piledriver offers no improvement over Bulldozer and gets trashed by Nehalem in those so so important minimal frame rates. Sandy Bridge is even further ahead!

---------------------------------------------------------------------------------------------------------------------------------------

RELATIVE PERFORMANCE






Conclusion: in my opinion the Piledriver FX adds way too little improvements to be competitive clock per clock with both Nehalem and Sandy Bridge Core i7 processors, and though it is a must upgrade from a Bulldozer FX CPU, no way in hell should it be an upgrade from Nehalem based Core i7 processors.

[Tonim89]

This review only confirms what everybody knows... The AMD's archtitecture should never have left the servers/workstations territory.

Great Job, bro. I'll thank you for this post, for sure!

[Protagonist]

Great Job, and now this just confirms to me how very many game developers are lazy and they don't make properly optimized game engines, it should not be a matter of CPU & GPU only, game developers need to step up in PC games and optimize their game engines, is not that they cant its just that they don't want .

I know they can do a better job with the PC game engines because they are the same people who highly optimize their engines for consoles, while consoles have less CPU & GPU muscle than most mid range and high end PCs.

[Frag Maniac]

Art, you must understand that ever since "P" ratings, CPUs should NOT be compared at same speed. Some CPUs are designed to run at different speeds than others.

The only way to truly compare CPUs is to run them at stock speeds and test them on some benches and games, then do the same with the highest possible OC.

They should also be run on the same exact test bed, or as closely as possible, because you're never going to get an Intel and AMD MB to run exactly the same.

If you run the test that way, then it will show what each CPU can do better. For one, it doesn't really account for the evolution of die shrinking from Nehalem to Sandy Bridge at all, let alone architecture improvements, both of which allow SB chips to run at higher speeds.

I'm sure Intel is still going to win on the gaming side though, at least on most games that aren't heavily threaded. AMD seems best for bang for buck workstation use.

This is about the same as the way the CPU world record speeds are done. They only talk about the clock they're able to achieve. There's no standardized benchmarks to compare actual performance accurately.

You may as well have said you're going to test them all at the same voltage too. I understand wanting to compare architecture, but if you don't test each architecture with the speed it's capable of, there's no real point to it.

So bottom line, a valiant intent, but the method seems flawed. While this test does blow smoke up my 950's butt some, making it look like it can hang with a SB, I know full well it would be left in the dust were both to be run stock and with best OC.

I also don't get get why you chose 1266MHz 7-7-7-20 for the RAM vs something like 1600MHz 9-9-9-24. Is your RAM not rated for 1600 speed? That Vcore also seems a bit low on the 920. Try running it on the default 1 hr Linack bench built into OCCT and see if you get no errors.

[drdeathx]

That's a lot of work. Great job. BF3 and COD would be in order too.

[Novulux]

Of course it shouldn't be an upgrade, not that people who purchased a top of the line i7-920 (2-4 years ago)would be expecting better performance from a middling AMD proccessor that costs much less...
On the other hand, based off this review, it also seems sad that even a similarly priced (~at release) Sandy Bridge-E processor does not offer much more performance since there would be bottlenecks elsewhere in most cases.

[Melvis]

Great job i must say, very well done.

Considering that a FX6300 is half the price of a 3820 i think it just shows how well the 6300 is to be honest?

[Lionheart]

This deserves more attention, great job

Seems like the AMD CPU's aren't as bad as I thought & I liked how you actually tested the games at full settings and at a decent resolution, I find it pointless when reviewer's test CPU's at extremely low resolutions & low settings but I guess it's for bottleneck reason's

[Artas1984]

Originally Posted by drdeathx

That's a lot of work. Great job. BF3 and COD would be in order too.

Ha! That's what i expected to hear! Most popular games don't exactly mean they have to make it in my list as i do not play any of them.

Originally Posted by Frag Maniac

The only way to truly compare CPUs is to run them at stock speeds and test them on some benches and games, then do the same with the highest possible OC.

That was not my intention at all! This theory of yours relies also on the fact, that for the max possible OC custom RAM are required for each CPU, and OC in general depends on the quality of motherboard. You are absolutely wrong - that would be more likely a comparison between several different computers or processor models rather than processor architectures.

Originally Posted by Frag Maniac

They should also be run on the same exact test bed, or as closely as possible, because you're never going to get an Intel and AMD MB to run exactly the same.

Huh? So that's what i did, did you even read the description? I've picked up the parts identical for all cpu's - same HDD, VGA, RAM. Now, Nehalem was running at 633 CL7, because that was the only possible setting to OC it at 4 GHz with 19X multiplier. The latencies don't really differ that much to compromise the benchmarks:

0,0105 for Piledriver and Bulldozer
0,0110 for Nehalem
0,0100 for Sandy Bridge

Originally Posted by Frag Maniac

I understand wanting to compare architecture, but if you don't test each architecture with the speed it's capable of, there's no real point to it.

No! That was wizzard's review, it's been done already, so why the hell would i do the same thing again? My review was exactly an architecture comparison, targeted at the relative performance instead of maximum performance. What you wanted me to do is to compare the architectures in the maximum performance criteria - which would include max oc possibilities, minimum voltage requirement, max temperature comparison - i understand that, but that was not my goal... My goal was a relative performance comparison and that i exactly did.

Disappointed comments from you Frag, i could have used a Core i7 960 and FX 8100, so now the Intel would be 400 MHz higher clocked and your words would not make any god damn sense - "test default speed and then max possible oc" - see how wrong you'd have been?

So just have in mind, that cpu architecture comparisons can be done in relative and maximum criteria and knowing that you should not have a problem with this review.

[postfranksa]

yeah,That's a lot of work. Great job. BF3 and COD would be in order too. thanks for your sharing

[Frag Maniac]

Originally Posted by Artas1984

So just have in mind, that cpu architecture comparisons can be done in relative and maximum criteria and knowing that you should not have a problem with this review.

I'm not aware that W1zzard did a review of these 4 CPUs with Piledriver included. If he hasn't, then it still would have been a useful comparison of what each one can do.

It's not so much a gripe as an observation that we already know Intel's solid core architecture beats AMD's floating modules (in games anyway), and that SB is obviously superior to Nehalem.

This test makes my 950 look like it could hang with an SB chip though, and we all know very well it can't.

So in a way it just points out the obvious of what we already know about superior architectures. Hope you see what I'm saying now.

I know you obviously put a lot of time into this, it just seems it's all to point out something that is already clear, whereas a real world Piledriver comparison might have been more useful and informative.

Chill Art, no need to get riled over one benchmark series.

[Artas1984]

Originally Posted by Frag Maniac

This test makes my 950 look like it could hang with an SB chip though, and we all know very well it can't.

But it does Frag! Listen! Your Core i7 950 3060 MHz would be just about 5 % slower in games than let's say a Core i5 2400 3100 MHz due to architecture and in real world programs your Core i7 950 would be faster due to 8 threads! This is what i wanted to tell you!!! Don't look at Core i7 920 and Core i7 3820 - the difference between them is a massive 940 MHz! Of course they are not close! I could have pushed the Core i7 3820 to 5 GHz!

[ChristTheGreat]

Originally Posted by Artas1984

But it does Frag! Listen! Your Core i7 950 3060 MHz would be just about 5 % slower in games than let's say a Core i5 2400 3100 MHz due to architecture and in real world programs your Core i7 950 would be faster due to 8 threads! This is what i wanted to tell you!!! Don't look at Core i7 920 and Core i7 3820 - the difference between them is a massive 940 MHz! Of course they are not close! I could have pushed the Core i7 3820 to 5 GHz!

But the i5 2400 is like what almost half price of what was the i7 950? and having power consumption way less.. Architecture isn't the same also

Thanks for the review , the only thing, you can't compare really clock per clock as the architecture isn't the same. 400mhz more on the i7 3820, is still 400mhz more.

AMD CPU are not that super for gaming, where intel exceed always in the single thread speed. But AMD made some good for multi-tasking. I would say also that it should stay in the server class, but at the price it is, why not . For Power consumption, if someone crunching, I guess buying a 100$ more CPU, will get maybe a 100$ less Bill of electricity at the end of the year (running CPU to 100%). Instead, AMD CPu are doing great, and we see it with the PileDriver.

Thanks again man for that review!

[Frag Maniac]

Originally Posted by ChristTheGreat

But the i5 2400 is like what almost half price of what was the i7 950? and having power consumption way less.. Architecture isn't the same also

Thanks for the review , the only thing, you can't compare really clock per clock as the architecture isn't the same. 400mhz more on the i7 3820, is still 400mhz more.

That's exactly what I was trying to say, and if you compare CPUs at the same price point (my 950 was about $300), it gets even more obvious that SB is superior. Hell, you can get a Sandy Bridge-E i7 3820 for that same $300, and it will smoke a 950 stock, and kick it's butt even worse when both are OCed.

I would be comparing an i5 2500k though, because in reality, it doesn't take much more than $200 now to get a great gaming CPU. Just the fact that you can get way better hardware for a lot less money than you could a couple years ago makes it obvious the technology in CPUs has gotten a lot better.

I mean anything in PC tech is that way, but esp when it comes to Intel CPUs in the last couple years. I've seen plenty of real world benches on even the 2500k with a 950 in the test, and those easily beat it too. Anyways, like I said, a valiant effort Art. No hard feelings I hope. Didn't mean to upset ya.

[Frick]

Originally Posted by Frag Maniac

I would be comparing an i5 2500k though, because in reality, it doesn't take much more than $200 now to get a great gaming CPU. Just the fact that you can get way better hardware for a lot less money than you could a couple years ago makes it obvious the technology in CPUs has gotten a lot better.

For pure gaming even $200 is a bit overkill.

[wolf]

What an excellent comparison dude, you did a great job so a big shoutout to you!

[claylomax]

Great job! I like the fact you didn't include BF3 and COD games because I don't play them and it's pissing off some people.

[Frag Maniac]

Originally Posted by Frick

For pure gaming even $200 is a bit overkill.

Actually it's a pretty good sweet spot price wise for a gaming CPU, even a budget one. More and more games are very CPU intense now too, and the sheer volume of i5 2500k chips sold tends to indicate many others feel the same.

I guess everyone has their own opinion though. There's always people looking to spend less, then they wonder why their cheaper chips don't perform as well as they'd like them to.

You have to add a little wiggle room for the oft embellished benches that are so common on CPUs.

[Frick]

Originally Posted by claylomax

Great job! I like the fact you didn't include BF3 and COD games because I don't play them and it's pissing off some people.

I don't play any games w1z has in his reviews so I hate his reviews.

I don't know if you're serious or not (I'm thinking not) but that was just stupid.

[claylomax]

Originally Posted by Frick

I don't play any games w1z has in his reviews so I hate his reviews.

I don't know if you're serious or not (I'm thinking not) but that was just stupid.

How is that stupid?
I'm sick of all these modern warfare games; whatever happened to them old shooters?

[Tatty_One]

Nice review and a lot of work for our benefit..... thank you. I used to have an old 920 C0 stepping, it was a pretty good one but I replaced it last month with a brand new 930 OEM which cost £70, thats considerably cheaper than the cheapest 4 core Piledriver..... it gives me 200mhz more 24/7 than my old 920...... hows that for bang for buck then!

[Frick]

Originally Posted by claylomax

How is that stupid?
I'm sick of all these modern warfare games; whatever happened to them old shooters?

You said you liked that he didn't include games many people play because you didn't play them. If we go by that measure I would only like reviews that include Heroes of Might and magic 3 performance.

[Artas1984]

Originally Posted by Protagonist

Great Job, and now this just confirms to me how very many game developers are lazy and they don't make properly optimized game engines

Hu? I think everyone knows that the developers deliberately do this, so that there would be the need to buy powerful video cards! It is a conspiracy with NVIDIA and AMD i think! They do this, so we had the need to buy SLI and CF video cards

Originally Posted by Tatty_One

hows that for bang for buck then!

I bet you now feel really lucky that you have bought the Core i7 930 cheap and not some AMD FX Piledriver?

Originally Posted by Frick

You said you liked that he didn't include games many people play because you didn't play them. If we go by that measure I would only like reviews that include Heroes of Might and magic 3 performance.

OF TOPIC: if you like Heroes 3, you must try this as well as some of my multiplayer maps.

You should know me if you are a celestial heavens member - i made many contributions in this game and have presented them there.

ON TOPIC: everything is ok, Battlefield and Call Of Duty series are well deserved great titles and should be included in any gaming CPU or VGA benchmark; however, there are plenty of benchmarks including those games already, and now for once you see a benchmark with some "nontraditional" lesser known games like Avatar, Cryostasis or Necrovision and that is also good, because some people would like to know how fare the processors in those games too.

[Artas1984]

I've just searched through the net a found a pretty impressive comparison between a Core i7 920 @ 4200 ΜΗz vs Core i7 3770 @ 4200 MHz

http://alienbabeltech.com/main/ivy-b...0-at-4-2ghz/5/

The processors are amazingly neck to neck in gaming! DK if it's year old, it's a must see!

[danwat1234]

Great work on the testing.

Looks the single threaded performance isn't much better, maybe 10% clock for clock with Nehalem vs Sandy bridge and from what I've seen of Sandy vs Ivy it's another 10% and Ivy vs Haswell is the same thing at around 10% boost.
With multitasking/SMP I'm sure it's much better.

I think I'll hold onto my G50VT with an overclocked X9100 and SSD for a few more years
Penryn to Nehalem from what I've read is about 20% clock for clock in single threaded comparisons.