Core 2 vs Piledriver FX?

[Dent1]

The memory specs are in the FX 4350's screenshot. it's 4GB 933 8-9-8-24. Also that FX 4350 isn't even close to the Phenom II that's not something that tweaking the NB and RAM will fix.

Again, Cinebench R11.5 is one benchmark, hardly the best example for showing single threaded performance.

That screenshot is too unofficial, nobody can't verify that its accurate or authentic.

Even if that screenshot is authentic, the timings on the Phenom II rig significantly tighter.

I'm not dismissing your results. Its interesting, but more interested in a review from a proper reviewer. (TPU, HardOCP, Toms Hardware, Guru 3D etc)

 

[buildzoid]

Again, Cinebench R11.5 is one benchmark, hardly the best example for showing single threaded performance.

That screenshot is too unofficial, nobody can't verify that its accurate or authentic.

Even if that screenshot is authentic, the timings on the Phenom II rig significantly tighter.

I'm not dismissing your results. Its interesting, but more interested in a review from a proper reviewer.

Memory is not important enough in Cinebench to get you more than 0.3(1333 7-7-7-21 vs 2133 9-11-11-31) points so the FX 4350 is not able to catch up. I will admit however that FX chips fully loaded have less IPC per core than when only 1 core is being used because how badly the cache is setup. If AMD wanted to show of IPC prowess they just need to build an 8 core Kaveri chip and disable all the odd cores. Because running 1 core per module boost the FX's IPC by about 25%. However the moment the other core in the module starts getting some load that IPC quickly disappears so if you want IPC and good multithreaded performance on an FX chip buy the 8350 and disable half the cores

 

[Dent1]

Memory is not important enough in Cinebench to get you more than 0.3 points so the FX 4350 is not able to catch up. I will admit however that FX chips fully loaded have less IPC per core than when only 1 core is being used because how badly the cache is setup. If AMD wanted to show of IPC prowess they just need to build an 8 core Kaveri chip and disable all the odd cores. Because running 1 core per module boost the FX's IPC by about 25%.

This may be true. But for the results to be valid it needs to be from a credible source. Not a community source. The configuration can't be a complete mismatch.

Also seems crazy you are basing this judgement on one benchmark, and a point based benchmark opposed to a FPS benchmark. Considering this thread is about single threaded performance (mostly gaming).

Look back at my previous post what do you think of the E8400 vs Phenom II X2 550's gaming results? - Seems like everyone is dodging responding to it!
http://www.techpowerup.com/forums/threads/core-2-vs-piledriver-fx.207620/page-3#post-3201425

In response to keeping the analysis apples vs apples. I will compare dual cores only at the same clock speed.
Phenom II X2 550 BE (3GHz) and E8400 (3Ghz) - Neoseeker review
http://www.neoseeker.com/Articles/Hardware/Reviews/pii_555/8.html
Bioshock
Phenom II X2 550 BE
186 FPS
149 FPS
121 FPS
Core 2 Duo E8400
180 FPS
141 FPS
114 FPS
Farcry 2
Phenom II X2 550 BE
58 FPS
59 FPS
55 FPS
Core 2 Duo E8400
57 FPS
58 FPS
55 FPS
Crysis Warhead:
Phenom II X2 550 BE
43 FPS
37 FPS
33 FPS
Core 2 Duo E8400
47 FPS
40 FPS
33 FPS
Left4Dead
Phenom II X2 550 BE
113 FPS
110 FPS
84 FPS
Core 2 Duo E8400
103 FPS
97 FPS
85 FPS
Call of Jarez
Phenom II X2 550 BE
108 FPS
Core 2 Duo E8400
109 FPS
World of Conflict
Phenom II X2 550 BE
162 FPS
Core 2 Duo E8400
186 FPS
The Phenom II X2 is faster in Bioshock, Farcry and Left4dead. Whereas the Core 2 Duo E8400 is faster in Crysis Wars and World in Conflict. A draw for Call of Juarez. The results are too narrow to conclusively say which processor performed better overall.
The only stand out distinction between the too was that World of Conflict performed significantly better on the E8400 but this was balanced out by the Phenom II X2's narrow but somewhat wide victories in Bioshock and Left4dead and being generally consistent in either narrowing winning or narrowing losing in the other games.
Neoseekers reviews seem to support Anandtech and Bit Tech. All three demonstrate the C3 Phenom II series being on par with the Core 2 Duo in single threaded gaming.
Would it be a stretch to assume If the Phenom II X2 and E8400 are equal (according to 3 reviews) the FX 6300 which is IPC is up to 25% better than its predecessor would fair well?
I open the floor for responses.

RESPONSE REQUIRED!

 

[TRWOV]

In my experience Cinebench loads the integer cores quite nicely since they scale as they should. Disabling Turbo core and running 3 threads on a FX6300 gives you about half the points.

Intel Burn Test on the other hand loads the FPUs and as such if you run 3 threads on a 6300 you get 90% of the performance running 6 threads:


FX6300 IBT 6 threads: ~52Gflops

FX6300 IBT 3 threads: ~48Gflops

CB11 FX6300 6 threads: 4.53pts

CB11 FX6300 3 threads: 2.71pts

 

[buildzoid]

In my experience Cinebench loads the integer cores quite nicely since they scale as they should. Disabling Turbo core and running 3 threads on a FX6300 gives you about half the points.

Intel Burn Test on the other hand loads the FPUs and as such if you run 3 threads on a 6300 you get 90% of the performance running 6 threads:


FX6300 IBT 6 threads: ~52Gflops

FX6300 IBT 3 threads: ~48Gflops

CB11 FX6300 6 threads: 4.53pts

CB11 FX6300 3 threads: 2.71pts

How are you setting the threads up? because my Asus Sabertooth can't disable cores in modules just whole modules. Also using the OS to control threads to run on 1 module each on FX chips is impossible

 

[Steevo]

AMD's Windows update tells windows to use even cores first instead of loading random cores to seemingly improve FX performance, and also fixes the issue with parking cores http://support2.microsoft.com/?kbid=2645594

The "module" issue is the cause of the performance decrease, due to the shared resources it stalls the pipeline more than the already slow cache does when two cores are sharing them.

Long story short AMD made a server chip, spun it off to make desktop chips, and it had less performance per core than their old architecture at first, later versions corrected this to some degree, but not enough to catch up, so they turned their focus to APU's which are good chips, and do wonders as all around work horses.

 

[RCoon]

The E8400 performs similarly to a Phenom II in most games relevant to it's timeline. If you check out per core performance, the FX line was largely similar to the Phenom II in FPU orientated performance. This means an FX 4xxx would trade blows with an E8400, the 6xxx series however has a single extra FPU at its disposal, so it should perform (very marginally) better than an E8400. And so it should. The E8400 is almost as old as my niece, and the FX is very recent. I don't see why anyone is debating over a chip that old. Who in their right mind would buy an E series intel processor instead of just buying a G3258 or an i3?

 

[HumanSmoke]

The E8400 is almost as old as my niece, and the FX is very recent. I don't see why anyone is debating over a chip that old. Who in their right mind would buy an E series intel processor instead of just buying a G3258 or an i3?

Exactly. Most of this thread is devoted to speculation and assumption basically because direct benchmark comparisons are almost non-existent. That in itself should be ample proof of the extent of the generational gap between the two architectures.

 

[Dent1]

The E8400 performs similarly to a Phenom II in most games relevant to it's timeline. If you check out per core performance, the FX line was largely similar to the Phenom II in FPU orientated performance. This means an FX 4xxx would trade blows with an E8400, the 6xxx series however has a single extra FPU at its disposal, so it should perform (very marginally) better than an E8400. And so it should. The E8400 is almost as old as my niece, and the FX is very recent. I don't see why anyone is debating over a chip that old. Who in their right mind would buy an E series intel processor instead of just buying a G3258 or an i3?

Your post is exactly what all the reviews suggest. Unfortunately nobody will listen to you

Your line of thinking is inline with Trusted Reviews, Hard OCP, Extreme Overclocking and Anandtech.

I presented numerous link evidence throughout this thread. Very few people bothered to comment on the results of the review and reverted back to blanket statements.

As far s the i3. That is a good question. I'm going to start researching single threaded gaming performance between i3 and FX 6300.

Exactly. Most of this thread is devoted to speculation and assumption basically because direct benchmark comparisons are almost non-existent. That in itself should be ample proof of the extent of the generational gap between the two architectures.

Not true. I posted many direct benchmarks.

Initially Phenom II X4 940 BE vs E8400 (both 3GHz) and then I did Phenom II X2 BE 550 vs E8400 (both 3Ghz)

The sad thing is I did the Phenom II X2 BE 550 (3Ghz) specifically as a request for Earthdog as he wanted two dual cores at the same clock. The results showed both architectures performed exactly the same but he didn't even reply to it (post #67)

Then in post #68 I took another request from Earthdog showing a direct comparison of the Bulldozer FX and Piledriver FX at the same clock to further illustrate that the 5-10% drop in single threaded performance between Bulldozer and Phenom II was superseded by the upto 25% increase in single threaded performance between the Bulldozer FX and Piledriver FX. So by process of elimination the Piledriver FX's single threaded performance would be greater than Bulldozer, Phenom II and Core 2 Duo.

 

[EarthDog]

I said thanks, what else do you want, lol!

To be clear, I didn't ask for a dual verus a dual. What I asked for was benchmarks that are not multi threaded as that clearly favors a quad core for anything over 2 threads. Instead you provided dual vs dual which essentially accomplished the same thing.

You also have only shown gaming comparisons. Cinebench is the only one that isn't a game and does a better job at isolating the CPU versus games. All you have is game data outside of cinebench right now.

As I said, I don't really disagree with your first threads assertion, just the data you are presenting really didnt prove your point. Looks like you have in games, but not yet in productivity...

 

[Tonduluboy]

lol this debate is like debating which shoes are more better in jogging "7 yrs old Adidas running shoes or 7 yrs old Nike running shoes" or something like that. Does anybody using one of these 2 chips for everyday use nowadays?

 

[HumanSmoke]

As far s the i3. That is a good question. I'm going to start researching single threaded gaming performance between i3 and FX 6300.

Don't get your hopes up. AMD themselves are distancing themselves from the current architecture as rapidly as possible. When the company itself admits that CMT isn't competitive enough going forward and they will ditch it as soon as possible, I doubt you'll make their case for them.
The fact that it requires a multi-thread application for the FX to outshine a dual core processor should be indicative. These gaming benchmarks - with and without GPU limitation would seem to highlight that fact.

Then in post #68 I took another request from Earthdog showing a direct comparison of the Bulldozer FX and Piledriver FX at the same clock to further illustrate that the 5-10% drop in single threaded performance between Bulldozer and Phenom II was superseded by the upto 25% increase in single threaded performance between the Bulldozer FX and Piledriver FX.

None of the benchmarks in that post exceed 20%, so I'm not sure how you managed to equate "up to 25%" from those figures.

 

[Dent1]

Don't get your hopes up. AMD themselves are distancing themselves from the current architecture as rapidly as possible. When the company itself admits that CMT isn't competitive enough going forward and they will ditch it as soon as possible, I doubt you'll make their case for them..

I'm going to look into i3s at a later date. There is just so much information absorb.

Don't get your hopes up. AMD themselves are distancing themselves from the current architecture as rapidly as possible. When the company itself admits that CMT isn't competitive enough going forward and they will ditch it as soon as possible, I doubt you'll make their case for them.
The fact that it requires a multi-thread application for the FX to outshine a dual core processor should be indicative. These gaming benchmarks - with and without GPU limitation would seem to highlight that fact.

None of the benchmarks in that post exceed 20%, so I'm not sure how you managed to equate "up to 25%" from those figures.

I never equated 25%. AMD claimed up to 25% increase in single threaded performance. Obviously this is a subjective claim which might not be reflected in those three games (Stalker, Final Fantasy XIV and Lost Planet). The sample would have to be much bigger and varied.

I was merely pointing out that improvements have been made between Bulldozer FX and Piledriver FX which the review clearly shows.

Even if 25% isn't an accurate average the improvement would be enough to align Piledriver with the single threading IPC of the Core 2 or Phenom II at minimum. This argument alone still nullifies the claims that Core 2 is a faster single threading architecture than Piledriver.

I was able to produce that evidence within 11 minutes of being asked. I then recommended the community calculate the percentages themselves it's unfair for me to do all the research.

Had I not spent 11 minutes and spent 1100 minutes finding a more suitable link you would have accused me of cherry picking. So I cant win.

 

[EarthDog]

It was AMD that made those claims, but, you posted them as support to your assertion... which we don't think there was close to 25% performance gains between BD and PD.

Again, I could care less about the subject matter, otherwise I would waste life and look too. I just want to see that the right metrics are coming out to come to a conclusion on this long forgone subject that already had a conclusion yet was rehashed in two threads now that both end up doing the same thing (woo hoo donuts in the hot rod.. spin dem wheels!).

So... you have games covered and proved that a newish AMD chip can hang with an old Intel... what about productivity? Cinebench seems to stray from that a bit from that one dataset, no?

 

[Dent1]

It was AMD that made those claims, but, you posted them as support to your assertion... which we don't think there was close to 25% performance gains between BD and PD.

Again, I could care less about the subject matter, otherwise I would waste life and look too. I just want to see that the right metrics are coming out to come to a conclusion on this long forgone subject that already had a conclusion yet was rehashed in two threads now that both end up doing the same thing (woo hoo donuts in the hot rod.. spin dem wheels!).

Again. Had I not spent 11 minutes and spent 1100 minutes finding a more suitable link you would have accused me of cherry picking. So I cant win.

I'm sure there is 1 or 2 benchmarks showing 25% but your mind is closed and your contribution is low.

Edit:

25% is at peak, so it only needs to achieve this once to get the stamp of approval, it isn't average. Even with a mere 8% or 10% average single threaded improvement over Bulldozer, it would equal the Phenom IIs IPC, we've already established the Phenom II equals the Core 2 Duos aruchitecturer (you haven't even denied that) so by proxy its impossible for the Core 2 architecture to outperform a Piledriver.

 

[buildzoid]

dude even Anadtech shows that the FX 4300 3.8Ghz is barely faster than a Phenom II X4 980BE 3.7Ghz. Also the FX4300 3.8Ghz is barely faster than the FX4130 3.8Ghz same as the. Now that maybe because the difference in L3 size but the 8320 and 8150 are also almost equal in performance and both have 8MB of L3. Piledriver has barely more IPC than Bulldozer but pulls much less power which is why Piledriver doesn't completely suck just mostly sucks.
Link to FX 4300 vs Phenom II
4130 vs 4300
8150 vs 8320

 

[EarthDog]

Listen boss, Im not the one on a frikkin soap box trying to correct people's perception with erroneous data (until I suggested how to show the right data to get your point across). Im trying to help get your ducks in a row...

Anyway, unsubscribed from this shitstorm. Good luck.

 

[Dent1]

dude even Anadtech shows that the FX 4300 3.8Ghz is barely faster than a Phenom II X4 980BE 3.7Ghz. Also the FX4300 3.8Ghz is barely faster than the FX4130 3.8Ghz same as the. Now that maybe because the difference in L3 size but the 8320 and 8150 are also almost equal in performance and both have 8MB of L3. Piledriver has barely more IPC than Bulldozer but pulls much less power which is why Piledriver doesn't completely suck just mostly sucks.
Link to FX 4300 vs Phenom II
4130 vs 4300
8150 vs 8320

Nobody has acknowledged my results in post #67 (Phenom II 550 vs E8400 - dual cores both 3GHz) performing exactly the same.

Why should I analyse your links if nobody cares enough to analyse my research?

Listen boss, Im not the one on a frikkin soap box trying to correct people's perception with erroneous data (until I suggested how to show the right data to get your point across). Im trying to help get your ducks in a row...
Anyway, unsubscribed from this shitstorm. Good luck.

But even with the "wrong" data or conservative estimates of 10% improvement's it would still show the IPC of the of the Core 2 Duo about equal to the Piledriver FX, not above.

Its just funny you ignored the data of the Phenom II 550 performing the same as the E8400. It seems you like to pick and chose what data you believe.

Running away just proves I'm right.

 

[EarthDog]

I need to unsub quicker, LOL!

But even with the "wrong" data or conservative estimates of 10% improvement's it would still show the IPC of the of the Core 2 Duo about equal to the Piledriver FX, not above.

And? I never disagreed with you... for the most part. About the only part I take exception to at this point is the 25% BD to PD improvements AMD claimed and you brought to the table as 'facts' to support your assertion.

Its just funny you ignored the data of the Phenom II 550 performing the same as the E8400. It seems you like to pick and chose what is factual.

JFC....... Again, I said thanks. What more do you want from me?

Ive moved on from the gaming side of the house... just waiting for you to support the productivity side of things at this point (but again, just done as I dont care... and Im getting shit on for helping you out even though I agree with most of what you say).

Incredible.

Running away just proves I'm right.

Wow... seriously? What it proves to me is that you are acting like a complete [insert insult here] towards me when I'm trying to help you out. But you keep picking at...... at..... I have no clue what you are on about. So at this point, you are going to sink or swim on your own kiddo.

 

[Dent1]

Fair enough. To be fair it was HumanSmoke which worked out it wasn't quiet 25%. Then you jumped straight in to double-team.

My point is it doesn't have to be exactly 25% to prove my point. 10% is enough to normalise the Piledriver FX to Phenom II or Core 2 IPC.


I've requested for the thread to be closed because I wanted this to be a friendly scientific debate that has been lost.

 

[HumanSmoke]

I never equated 25%. AMD claimed up to 25% increase in single threaded performance. Obviously this is a subjective claim which might not be reflected in those three games (Stalker, Final Fantasy XIV and Lost Planet). The sample would have to be much bigger and varied.
I was merely pointing out that improvements have been made between Bulldozer FX and Piledriver FX which the review clearly shows.

Nobody doubts improvements were made, but repeating PR fluff as fact doesn't help your cause.

Had I not spent 11 minutes and spent 1100 minutes finding a more suitable link you would have accused me of cherry picking. So I cant win.

That depends upon what you presented and what you chose to leave out. Presenting verifiable nonpartisan and representative figures don't constitute cherry picking as far as I'm concerned. With that in mind, I'd point you towards CPU World's comparison of Zambezi (FX-6120) with Vishera (FX-6300) at identical stock clocks. These are more in line with the actual architectural improvements.

 

[EarthDog]

LOL, Im not on a god damn team Dent. Facts are the side I am on. Sorry I happened to agree with someone else on this forum and that appeared to be piling on.

IBTC.

 

[MT Alex]

So... you have games covered and proved that a newish AMD chip can hang with an old Intel...

That the point of my previous post. I'm not sure how any of this diatribe makes AMD look any better.

 

[eidairaman1]

Why does this even matter? Core2 went EOL like what, 5 years ago? I find it pathetic that Core2 is so close to FX in performance.



What sites are those? Simple fact of the matter is that Intel cpus do outperform AMD cpus, and Nvidia has the top performing GPU right now. That is why this site is primarily pro-Intel and Nvidia. If AMD takes the performance crown in either of those subjects, this site would turn pro-AMD.

As a whole to the forum, There's no point of berating users who go against the grain. I couldn't be happier with the build I took time to piece together. I try to help where I can if people do listen whether red or blue (don't know enough of green). Some users here act like those in Ferguson-bullying.

 

[BarbaricSoul]

As a whole to the forum, There's no point of berating users who go against the grain. I couldn't be happier with the build I took time to piece together. I try to help where I can if people do listen whether red or blue (don't know enough of green). Some users here act like those in Ferguson-bullying.

Only thing I berated was the performance of AMD's FX processors. I have not berated a single user in this thread thank you.