[TOM'S] AMD inter-generational CPU shootout

[Sweepr]

Not trying to get into a hyperbole war, but I think that your wrong. It might not be by much But I think that the 6k would be faster with it's two extra cores if an application is using it's full power.

Here's a list of highly MT applications (7-Zip, x64, WinRAR...) running on both. The 4C i5-4670K is usually ~30% faster than the 3M/6C FX6300 and comes repectably close to the 4M/8C FX8350 (~3.6% slower on average) with much lower power comsumption on load.

 

[Abwx]

Here's a list of highly MT applications (7-Zip, x64, WinRAR...) running on both. The 4C i5-4670K is usually ~30% faster than the 3M/6C FX6300 and comes repectably close to the 4M/8C FX8350 (~3.6% slower on average) with much lower power comsumption on load.

HFR average include all their MT AND ST applications ,
so the 3.6% is totaly irrelevant when looking specificaly
at MThreaded environments.

We can clearly see this with the 2600 being only
1% better than the relevant i5 while it clearly
best it in MT scenarii.

 

[Sweepr]

HFR average include all their MT AND ST applications ,
so the 3.6% is totaly irrelevant when looking specificaly
at MThreaded environments.

We can clearly see this with the 2600 being only
1% better than the relevant i5 while it clearly
best it in MT scenarii.

There is no ST test. The beauty of Sandy Bridge to Haswell IPC improvements is that an i5-4670K 4C/4T is very close to i7-2600K 4C/8T performance in MT tasks @ equal clocks. Pretty cool huh.

 

[Gikaseixas]

Like which?

Edit:

From what I remember, PoV the i5 is faster, Cinebench the i5 is faster, Handbrake the i5 is faster.

What else have we compared?

wow not worth answering really

 

[BallaTheFeared]

@Abwx I said "our tests" meaning the 8350 no longer has the 400MHz speed advantage, though I'm sure at the clocks we run it still has a tremendous power consumption advantage

wow not worth answering really

Whats not worth answering?

 

[Abwx]

There is no ST test. The beauty of Sandy Bridge to Haswell IPC improvements is that an i5-4670K 4C/4T is very close to i7-2600K 4C/8T performance in MT tasks @ equal clocks. Pretty cool huh.

No , there are some tests that are not fully MTs ,
here the worst of that list in matter of perfs
delta in function of threads scaling..

http://www.hardware.fr/articles/897-17/cpu-compression-7-zip-winrar.html

 

[BallaTheFeared]

abwx you're cherry picking results that have nothing to do with the context of the discussion.


What does 7zip have to do with games?



Is WinRar really that much better than 7z?

They said they compressed the same file, but look at the difference in results.. 7z is slow compared to WinRar...

I wonder what the output looked like though?

That is actually quite interesting, despite it's lack of context!

 

[Phynaz]

It is slower in all thoses tests...

Unless the i5 has more compute throughput than a 3770 or 4770.

Hmmm...Wondering why you are posting benchmarks from the 6800K review instead of the FX review that you originally linked to. Cherry picking again?

Anyway, here's the charts from hitting "next page" on your original link. Please take note that I didn't cherry pick and left the x264 second pass benchmark in there. After all I wouldn't want you to think I'm being dishonest.

 

[Abwx]

Hmmm...Wondering why you are posting benchmarks from the 6800K review instead of the FX review that you originally linked to. Cherry picking again?

Anyway, here's the charts from hitting "next page" on your original link. Please take note that I didn't cherry pick and left the x264 second pass benchmark in there. After all I wouldn't want you to think I'm being dishonest.

Cherry picking such that we have the most recent intel CPUs
scores since balla has a HW , so much for your straw.

Well , you re just forgetting that we were talking
of total computational power and the second page has
a few benches that do not uses all the available cores
as aknowledged by the few difference between CPUs
that have half the core.

If we should follow you in your interpretations there would
be no substancial delta between intel cpus that are known
to be vastly distants perfs wise in really Mthreaded cases.

But while we are at it , why not go also to the third page
of the "non cherry picked" review , with FX8XXX performing
slightly better in CB than in the review i "cherry picked"...

Surprisingly , not a good score in Euler3D , isnt it.?..

Well , it is optimized , with this notice :

User's Guide
3
Optimization Notice
Intel’s compilers may or may not optimize to the same degre
e for non
-
Intel microprocessors for
optimizations that are not unique to Intel microprocessors. These optimizations include SSE2, SSE3,
and SSSE3 instruction sets and other optimizations. Intel does not guarantee the availability,
functionality, or effecti
veness of any optimization on microprocessors not manufactured by Intel.
Microprocessor
-
dependent optimizations in this product are intended for use with Intel
microprocessors. Certain optimizations not specific to Intel microarchitecture are reserved for
Intel
microprocessors. Please refer to the applicable product User and Reference Guides for more
information regarding the specific instruction sets covered by this notice.

http://software.intel.com/sites/default/files/m/d/4/1/d/8/ShallowWater.pdf

 

[Phynaz]

Well , it is optimized , with this notice :

http://software.intel.com/sites/default/files/m/d/4/1/d/8/ShallowWater.pdf

So what? It's real application that real people buy.

Should we continue onto the game benchmarks and other pages from the same article you attempted to cherry pick from?

 

[Vesku]

The FX 6300 looks pretty good overall if you aren't looking to minimize power use. Clocked into the low 4GHz range it falls between a G2020 and i3-3225 in one of the worst gaming scenarios for AMD, StarCraft 2. The FX 6350 stock is faster (mild OC target for the FX 6300) than the i3-3225 in games like Crysis 3 and Borderlands 2, in the THG reviews. If the ~$120 difference between the FX-6300 and the 4670K means you can get more GPU for a gaming machine I'd take that option.

 

[Abwx]

So what? It's real application the real people buy.

Should we continue onto the game benchmarks and other pages from the same article you attempted to cherry pick from?

Real applications bought by insuspecting customers
who dont know that they bought an innefficient product ,
but whatever , that s not the subject..

Btw , where was the cherry picking since we were
talking of throughput not of current games benches
since the discussion geared toward wich would be
more efficient with games progressively going
the highly multithreaded route , wich lead me to think
that you didnt take the care to at least try to read
a few posts before coming with yours perpetuals
accustations of cherry picking that end backfiring
against your own sayings.

 

[BallaTheFeared]

The problem is AES has nothing to do with gaming performance, other than that I'm ok with your charts.

I did notice you were missing Haswell in some of yours, so I went ahead and added stock 3.6GHz and clock for clock 4GHz vs Piledriver.


Cinebench

i5-4670k @ 3.6

i5-4670k @ 4GHz

PoV Chess

i5-4670k @ 3.6

i5-4670k @ 4GHz

PoV Benchmark

i5-4670k @ 3.6

i5-4670k @ 4GHz

I'll stand by my comment, AMD needs two cores to match one Intel non hyperthreaded core - specifically in gaming/gaming type workloads. So even if games go six thread the x6 will not be faster and if it goes to eight the x8 will only reach similarity it will not suddenly be a better choice.

 

[aigomorla]

965be still rocking

+1

friend still has her 965 rocking strong.
and she hasnt felt cpu limited at all in anything.
And its probably the one of the oldest systems i still service which doesnt require much outside blowing dust.

I did notice you were missing Haswell in some of yours, so I went ahead and added stock 3.6GHz and clock for clock 4GHz vs Piledriver.

need gulftown?

 

[Enigmoid]

Surprisingly , not a good score in Euler3D , isnt it.?..

Well , it is optimized , with this notice :

http://software.intel.com/sites/default/files/m/d/4/1/d/8/ShallowWater.pdf

Euler3D also loves cache performance and bandwidth which unfortunately Piledriver sucks at.

Adding the L4 cache to iris allow roughly a 40% gain (4950 is roughly equal to 2600k in the singlethread test).

 

[BallaTheFeared]

need gulftown?

Need a eight core Haswell-E?


>insert crazy score here<

 

[aigomorla]

Need a eight core Haswell-E?


>insert crazy score here<

the beckton were supposed to come out in 10cores.. yet never saw them.

cpu's sucked tho... i think Magoney were greater in the sys benchmarks.

If i recall, again... the beckton and dunnington suffered from cache shortages which made them slower then there cousins.

 

[BallaTheFeared]

You're such a pessimist

 

[Abwx]

Euler3D also loves cache performance and bandwidth which unfortunately Piledriver sucks at.

Adding the L4 cache to iris allow roughly a 40% gain (4950 is roughly equal to 2600k in the singlethread test).

In single threaded it would be surprising that the 2600 is clocked
40% higher than the 4950.

Granted the cache perfs , Euler3D is mainly floating point
dependant and in this matter the FX is lacking even if
it has as much FPU throughput as a 2600.

 

[aigomorla]

You're such a pessimist

cuz with intel lately...

*sigh*

i always expect the worst.. but hope for the best...
while before... it was expecting for the best... getting blown away with even better...
(C2D transition up to the Sandy... )

After Sandy.. i have no idea Whose in charge of intel.. or what there even thinking sometimes...
It seems like my 2 favorate companies have pulled a 180 and started flicking people off... (microsoft + Intel)

 

[Unitary]

This was certainly an interesting article, but I don't understand their constant Athlon II x4 bashing. Sure, it was the slowest processor of the bunch, but it is a budget cpu that came out four years ago.

I'd instead offer a different conclusion, one that highlights that paired with a modern GPU, this four year old budget cpu still can play modern games (mostly) at the highest settings. Naturally, everyone has their own opinions on what FPS level is acceptable, but if I had a four year old computer that gets 30+ FPS on average at 1080p with most eye candy on, I'd be pretty happy.

 

[JimKiler]

Haha, yeah I was.

I guess you could do it, but like I said, I doubt you're CPU is bottlenecking your GPU yet. Also, I don't think that's a drop-in replacement because the FX chips require socket AM3+ while your current MoBo is probably AM3. AM3 supports AthlonII/PhenomII, AM3+ supports those + the FX line. So any upgrade at this point will require a new MoBo also...

No new mobo required, my original upgrade was a Phemon 2 and AM3 but since the board was flaky i got an AM3+ 2 or 3 years ago so I can definitely put a FX CPU in my rig. I would like to upgrade after seeing this article. Obviously a Core i5 would be better but I am too cheap to spend that much.

I think the real issue is I should have spent another $50 - $60 for a better video card than the 7770, which works fine, but i want a little more now.

 

[Enigmoid]

In single threaded it would be surprising that the 2600 is clocked
40% higher than the 4950.

Granted the cache perfs , Euler3D is mainly floating point
dependant and in this matter the FX is lacking even if
it has as much FPU throughput as a 2600.

No, the 4950 runs at a turbo on 4 cores of 3.4 ghz, slightly below the 2600k but made up by the IPC increase.

 

[escrow4]

I'm running an FX 6300 currently with a 7870, its more than sufficient with any single player game set to High at 1080p, 40+ FPS with some choice settings adjustments in some games (I'm looking at you Crysis 3). Looks like I'll drop in a 9950/9870 come Christmas. It may be entertaining to spend dozens of posts arguing which is better but in the end that FX 6300 will do the job. And yes, I am fully aware than in multi player games especially it's a good idea to go Intel. Otherwise though, for the price of an i5 I got a 6300 + 970 mobo, that there is value.