My 8150 Bulldozer experience - so far!

[DeeDot78]

Just FYI today, the 8120 price has dropped on several sites to 144 to 149. Price cuts taking affect.

 

[Mallibu]

Newegg prices

FX8120 = $159,99
FX8150 = $189,99

Core i5 3470 = $199,99

FX8120/8150 at 4.7GHz is way faster and cheaper than Core i5 3470 at 4GHz in MT apps.

So yes, Intel is more expensive and perform worse than AMD at the same Price points.

Do i get the medal now ??? :biggrin:

If by "better in MT apps" you mean that "BD is better by 4-5% at certain 6-7 apps that 99% of userbase doesn't use so often to care about, and falls behind by 10 to 50% compared to Sandy Bridge in all the others" then yeh sure, we agree.

Do yourself a favor and stop doing this:

 

[CHADBOGA]

If by "better in MT apps" you mean that "BD is better by 4-5% at certain 6-7 apps that 99% of userbase doesn't use so often to care about, and falls behind by 10 to 50% compared to Sandy Bridge in all the others" then yeh sure, we agree.

Do yourself a favor and stop doing this:

Hah, great post.

 

[AtenRa]

If by "better in MT apps" you mean that "BD is better by 4-5% at certain 6-7 apps that 99% of userbase doesn't use so often to care about, and falls behind by 10 to 50% compared to Sandy Bridge in all the others" then yeh sure, we agree.

1. FX8120/8150 at 4.7GHz is faster than 5-6% at those applications against the 3570K at 4.5GHz. Now, imagine how much master FX8170 at 4.7GHz will be against the 3470 at 4GHz, it will obliterate it.

http://forums.anandtech.com/showthread.php?t=2241712&highlight=

Posts 1 and 21

Cinebench 11.5 (MT)
FX8150 @ 4.7GHz ~3.8% faster than Core i5 3570K @ 4.5GHz

7-zip
FX8150 @ 4.7GHz ~27% faster than Core i5 3570K @ 4.5GHz

x264 HD4.0 (Second Pass)
FX8150 @ 4.7GHz 12% faster than Core i5 3570K @ 4.5GHz

TrueCrypt 7,1a (AES)
FX8150 @ 4.7GHz ~20% faster than Core i5 3570K @ 4.5GHz


2. Everyone have judged the Bulldozer from the same 6-7 apps that Anandtech uses at every CPU review the last year or so. Now, because the 3470 cost more and is slower than the Bulldozer at those apps you telling us that those apps are not important.
When this is happening, it only means that the opposition has a better product at the same or lower price and people trying to downplay the importance of the apps that shows their favorite CPU in a worst position.

3. So far you are only words, you haven't provided a single benchmark chart to support your arguments. So, take off the blue glasses and see beyond your nose, there are more choices than Intel at the same or lower prices.

 

[inf64]

Some people just don't want to accept the fact that Bulldozer does good in numerous multithreaded workloads. It does this with only 4 FP units(although it does have 8 integer cores) and many "problems" in design which were identified by many reviewers/analysts. The real problem of the design is anemic ST performance and AMD should be fixing some of this in SR core(nobody knows how much faster it will be,but 15%-20% is possible). If process node gives them another 10% with higher clocks,then SR core might become the "real" Bulldozer. The problem is that by that time intel will have newer products so it's a moving target for AMD. But it's good they do acknowledge the problems and are working on fixing them.

 

[Mallibu]

Cinebench 11.5 (MT)
FX8150 @ 4.7GHz ~3.8% faster than Core i5 3570K @ 4.5GHz

7-zip
FX8150 @ 4.7GHz ~27% faster than Core i5 3570K @ 4.5GHz

The 3.8%faster in Cinebench is acceptable (no big deal though, Bulldozers main strong point is these apps and a 3.8% is not enough).

About 7zip,here's another one:

x264 HD4.0 (Second Pass)
FX8150 @ 4.7GHz 12% faster than Core i5 3570K @ 4.5GHz

TrueCrypt 7,1a (AES)
FX8150 @ 4.7GHz ~20% faster than Core i5 3570K @ 4.5GHz

Here's another from x264

I'll give you Truecrypt though, but I don't think it's enough to convince me about FX superiority.

2. Everyone have judged the Bulldozer from the same 6-7 apps that Anandtech uses at every CPU review the last year or so. Now, because the 3470 cost more and is slower than the Bulldozer at those apps you telling us that those apps are not important.
When this is happening, it only means that the opposition has a better product at the same or lower price and people trying to downplay the importance of the apps that shows their favorite CPU in a worst position.

3. So far you are only words, you haven't provided a single benchmark chart to support your arguments. So, take off the blue glasses and see beyond your nose, there are more choices than Intel at the same or lower prices.

Noone is downplaying those apps that FX is better it's just that even though most people zip files and encode videos, they won't buy a CPU that specialises in that and sucks in other workloads when there are alternatives that offer all-around better performance (i5 3570k.)

About those benchmarks you asked, I though it was common sense since every review site on the web is showing the i5 is better, but since you're lacking that sense here's some quick pictures for your joy (tell me if you need more)

 

[Mallibu]

Some more:

Spoiler

 

[Mallibu]

And some more:

Spoiler

 

[Mallibu]

And some last ones..

And all these, while FX consumes 2x more power at stock and 3x more power when O/Ced.

But yeah, I see your point, FX is totally better :awe:

 

[inf64]

Aten RA was using MTed workloads. I can't see how you missed that one "minor" part . Prime example of this is when you used 1st pass for x264,games,lame compression etc. Also sysmark,pcmark,3dmark.. Seriously? Who cares about these BS synthetic tests?

 

[AtenRa]

The 3.8%faster in Cinebench is acceptable (no big deal though, Bulldozers main strong point is these apps and a 3.8% is not enough).

You forget that this 3.8% is against Core i5 3570K @ 4.5GHz that cost even more than 3470.

But Ill keep that in mind for feature quoting

Noone is downplaying those apps that FX is better it's just that even though most people zip files and encode videos, they won't buy a CPU that specialises in that and sucks in other workloads when there are alternatives that offer all-around better performance (i5 3570k.)

So, the 3470 was not enough and you are now using the even more expensive 3570K. What happened to 3470 ??

I fail to see where is the FX8120/8150 @ 4.7GHz against the Core i5 3470 @ 4GHz among all those charts you have quoted ???

 

[Mallibu]

Aten RA was using MTed workloads. I can't see how you missed that one "minor" part . Prime example of this is when you used 1st pass for x264,games,lame compression etc. Also sysmark,pcmark,3dmark.. Seriously? Who cares about these BS synthetic tests?

Sure, let's take out those 3 benchmarks as useless (like cinebench isn't, but I see you linking it). There are like ~40 other benchmarks linked above that I'm sure you'll find interesting.

 

[exar333]

Just FYI today, the 8120 price has dropped on several sites to 144 to 149. Price cuts taking affect.

Damn, I realize how spoiled those are who live by a MC. The 2500K is still only $159...

Grabbed a 3570K + GB MB for ~$230 yesterday for a family member's budget gaming build.

 

[inf64]

Sure, let's take out those 3 benchmarks as useless (like cinebench isn't, but I see you linking it). There are like ~40 other benchmarks linked above that I'm sure you'll find interesting.

Multithreaded benchmarks. I will leave you to ponder over this a while.

 

[Mallibu]

Multithreaded benchmarks. I will leave you to ponder over this a while.

I'm sure all of the benchmarks above are fairly Multithreaded.
If by multithreaded you mean using all 8 threads, then I have some bad news for you about todays and tommorows software.

 

[Yuriman]

Those benches are typically of a stock FX-81x0 against a stock 2500k. When you overclock, 3.6ghz -> 4.7ghz and 3.4ghz -> 4.5ghz, it's a pretty similar percent increase so you can probably compare stock to stock with overclocked to overclocked with some safety. Also consider that Ivy generally has around a 5% IPC advantage over Sandy.

There are indeed tests where Bulldozer shines, but I don't really see it as a significantly cheaper alternative if you're going to crank it up to 4.7ghz due to higher cooling and power needs over an Intel i5.

 

[inf64]

The ones that Aten was using as example were very well multithreaded. The benchmarks you used are mostly dual threaded and we know that in low thread workloads FX is not competitive (even to lowly i3 since even lowly i3s perform close to i5/i7 in those cases,difference only being that they are locked frequency wise and i5/i7 can Turbo up).
Cinebench ,mainconcept,x264 2nd pass ,povray,solidworks,C-ray,truecrypt,photoshop cs 5.1,handbrake,abbyy finereader,7-zip are all examples where FX>i5 and sometimes even i7 2600K. There are real world applications in client segment that are well mulithreaded and as time goes by, they will become the dominant part of test suits with ST ones being minority.

Here is hardware.fr average from their application test suite:
http://www.hardware.fr/articles/863-22/moyennes.html
FX 8150 @ stock -150.7 pts
i5 2500K @ stock -136.4pts

 

[Mallibu]

The ones that Aten was using as example were very well multithreaded. The benchmarks you used are mostly dual threaded and we know that in low thread workloads FX is not competitive (even to lowly i3 since even lowly i3s perform close to i5/i7 in those cases,difference only being that they are locked frequency wise and i5/i7 can Turbo up).
Cinebench ,mainconcept,x264 2nd pass ,povray,solidworks,C-ray,truecrypt,photoshop cs 5.1,handbrake,abbyy finereader,7-zip are all examples where FX>i5 and sometimes even i7 2600K. There are real world applications in client segment that are well mulithreaded and as time goes by, they will become the dominant part of test suits with ST ones being minority.

The ones I linked, are multithreaded and use 2-4-6 threads and represent todays software usage.
Don't expect the 8 thread usage to become mainstream anytime soon in the following years, since 50% of people have Dual Core and 45% a Quad Core (from Steam statistics). We're in the 2->4 threads transition and we will be in many years to come.

So, we can compare the two processors in RL usage (benchmarks linked above that use 2-4 or even 6 threads), or the 3-4 programs that use 8 threads and represent a very special niche that probably noone cares enough about (statisticaly).

 

[inf64]

I have provided the average application scores from hardware.fr for both FX and 2500K. FX is 10% faster on average. And not all of the tests are well MTEd,so it's not the best case for FX either.

 

[Mallibu]

The ones that Aten was using as example were very well multithreaded. The benchmarks you used are mostly dual threaded and we know that in low thread workloads FX is not competitive (even to lowly i3 since even lowly i3s perform close to i5/i7 in those cases,difference only being that they are locked frequency wise and i5/i7 can Turbo up).
Cinebench ,mainconcept,x264 2nd pass ,povray,solidworks,C-ray,truecrypt,photoshop cs 5.1,handbrake,abbyy finereader,7-zip are all examples where FX>i5 and sometimes even i7 2600K. There are real world applications in client segment that are well mulithreaded and as time goes by, they will become the dominant part of test suits with ST ones being minority.

Here is hardware.fr average from their application test suite:
http://www.hardware.fr/articles/863-22/moyennes.html
FX 8150 @ stock -150.7 pts
i5 2500K @ stock -136.4pts

First, it's 148.8 vs 150, since it's the 3570 we're talking about.
And let's see the whole story from the review you linked, shall we?

Why not mention this?

The fact that you managed from the whole review that shows FX is so far behind, to take an isolated image that shows a 1% difference and ignore the 30-40% differences linked above, makes you look very desperate.

edit. Images appear to be dead atm, so here's the link for some reading
http://www.hardware.fr/articles/863-21/jeux-3d-total-war-shogun-2-starcraft-ii-anno-1404.html
http://www.hardware.fr/articles/863-20/jeux-3d-rise-of-flight-f1-2011.html
http://www.hardware.fr/articles/863-19/jeux-3d-crysis-2-arma-ii-oa.html
Read the whole review before trying to reference something from it next time please.

 

[inf64]

Because we were talking about APPLICATION performance,not gaming. If you want to game on a budget,you can buy a cheap i3 and forget about i5 and other similar CPUs since they bring you little over i3 (for now). In productive work (a thing that brings you money) ,FX is better than i5 in many situations. If you want to take entertainment (games) and base your whole argument on a trivial case like that, then you are more desperate than you seem .

 

[AtenRa]

Nobody said the FX is an all around better CPU than the 3470, but at those MT apps it is faster and cheaper. On the other hand, you are using a way more expensive 3570K (FX8120 $159 vs 3570K $229) to prove that it is faster in 1-2 threaded apps.
Not only that, but now that Intel have lost the performance/price in heavy multithreaded apps, you only care about 1-2 threaded apps because most of users use dual or quad core CPUs.

Next time i will compare the 8120@ 4.7GHz to a Core i3 at default clocks, i dont believe people will mind as they do the same with FX vs 3570K/3770K. ()

 

[AtenRa]

First, it's 148.8 vs 150, since it's the 3570 we're talking about.

No, we were talking about the FX8120/8150 @ 4.7GHz against the Core i5 3470 @ 4GHz.

 

[Mallibu]

Nope,I care about todays software which is 2 to 4 threaded. (and to a little extend 6 threaded.) I do not care about 2-3 specific programs that are 8 threaded, unless I plan to use them and only them.

So in todays software the i5 is better, as shown by the 50+ benchmarks above.

 

[inf64]

Yes ok ,whatever fits your dream world...