AMD FX-8120P benchmark from Coolaler

[Dresdenboy]

From what I recall JFAMD kept saying IPC improved from Phenom II (and certainly not that it declined). So how can a 3.2ghz 8 core BD be slower than the X6 1100T if IPC improved according to him in a multi-threaded bench?

"JF-AMD posting: IPC increases!!!!!!! How many times did I tell you!!!"

One problem with that IPC statement is the missing background info (the backup slides so to say). Does IPC improve only when code is recompiled? Is this for synthetic benches running out of the cache or is this already including the improved IMC etc.? W/o knowing that it's still too early to "prove" something wrong.

Ya, I was there. They said BD1 is a stop-gap CPU. They are waiting for 22nm before they can truly unleash the power of BD architecture. From what I saw, I am going to skip over BD altogether because BD2 (Next Generation Bulldozer) is going to be amazing!

This wonderfully crafted slide's results are not representative of the final silicon, since we don't know the version of AGESA

 

[tulx]

One problem with that IPC statement is the missing background info (the backup slides so to say). Does IPC improve only when code is recompiled? Is this for synthetic benches running out of the cache or is this already including the improved IMC etc.? W/o knowing that it's still too early to "prove" something wrong.


This wonderfully crafted slide's results are not representative of the final silicon, since we don't know the version of AGESA

Hm, I remember seeing similar charts that claimed BD to be far superior to Deneb/Thuban/Sandy Bridge. Now it seems that won't be the case.
And FX2 was supposed to be a socket FM2 platform? In that case, Thuban is probably the pinnacle of AM3 and AM3+ is pretty much useless, should FX underperform Thuban...
Still, I'll wait for rock-solid reviews of the final FX to judge it. I'm OK with it being slower than a 2500 on 4 threads and less, but being actually slower than Thuban would be just outragous.

 

[psolord]

So umm...what was that great news that AMD was so excited about in Tweeter? It shouldn't be the OC event in Austin I think. This was on the same day they tweeted, that "tomorrow" they would announce something great. What did I miss? :S

 

[Ancalagon44]

I think it was a demo of trinity.

 

[Dresdenboy]

@tulx:
That slide has apparently been faked by RussianSensation

@psolord:
Choose one:
- Guinness World Record
- Southern Islands 28nm graphics demo
- Trinity demo (which has been demoed in summer already)

 

[PreferLinux]

Another thing I think we should all consider with these early benchmarks is the compiler and optimizations.

I suspect (being a new arch, having avx, etc.) that we could see some slow/odd behavior with certain software and potentially large improvements as things get updated and software that wasn't compiled with 1yr+ old compilers gets deployed

Sure, but how many people run Gentoo Linux? Or Linux at all? Or even uses Windows with performance-critical open source software? Other than the server or HPC markets, very few. So who can get their software done with the latest compiler to support it? Even less of those very few. So how relevant are benchmarks done with the latest compiler? Not at all to those who want to know how it will perform, but good for those of us that actually want to know just how fast it can be. People on here still cry foul when AVX is used...

@tulx:
That slide has apparently been faked by RussianSensation

@psolord:
Choose one:
- Guinness World Record
- Southern Islands 28nm graphics demo
- Trinity demo (which has been demoed in summer already)

Definitely. Note the date – Fall 2013.

 

[Ancalagon44]

Definitely. Note the date – Fall 2013.

Remember, its showing supposed Bulldozer 2 performance comparisons.

However, it doesnt make sense that they compare BD2 against Phenom II. They would be comparing it to BD1 - even if BD1 were slower than Phenom II, it would show how much of an improvement BD2 is over BD1. Showing how much faster it will be compared to Ph2 is pointless.

 

[BlueBlazer]

EVEN if that is true above, taking data from post #32 in this thread:

Even if that is true, that $300 from AMD FX contest, and $266 from online pre-order pricing starts to make sense. :hmm:

Remember JF-AMD's FAQ about drivers? Seems there will be drivers for Windows 8.....

The new drivers will cover the AMD Fusion Z-Series for tablets, C- and E-Series for ultrathin and mainstream notebooks and desktops, A-Series for high-performance laptops and desktops, and the upcoming AMD FX eight-core Black Edition for high-performance desktops.

 

[AtenRa]

First they showed a comparison between a new unspecified AMD FX processor and an unspecified Intel processor i5 Sandy Bridge, with the help of the program Handbrake a video of 5 minutes is converted to H.264 video in SD resolution. The AMD FX processor with eight cores perform this function with an average of 223 frames per second, the i5 with four cores came in at 188 fps. Both systems will be comparable in price according to AMD, which it wants to show that AMD a better price / performance offering. That may be the case, but who has a more negative view would conclude that AMD is looking to double the number of cores needs to less than 20% better performance available.

http://translate.google.nl/translat...officiele-benchmarks-amd-fx-processor&act=url

 

[Riek]

http://translate.google.nl/translat...officiele-benchmarks-amd-fx-processor&act=url

The thing about handbrake is that HT isn't really helpfull...

So that FX sample would also best the 2600K by >10%. Basically what you get is an FX8120? that would be as fast as an 2600K @ 3.85GHz. (given Handbrake is pretty much SSE stuff i guess? this would be rather impressive imho.) But according to that they should have run FX8150 vs 990x in that test... because my estimate is they would perform about equally here. (unless the FX8150 is the one priced at the i5 level ofcourse.)

 

[Piano Man]

http://translate.google.nl/translat...officiele-benchmarks-amd-fx-processor&act=url

No bueno. First they are cherry picking, like any company would do. The first test I'm assuming is the FX-8120 vs. the i5-2500k since both are right at $220 doing something that would definitely benefit more cores over anything.

Then the 2nd test was most likely the FX-8150 vs the i7-980x which we all know loses to the i7 2600k in gaming at stock settings. The FX gets a 2-3% win at a crazy high resolution on a game that is most likely cherry picked.

And these are official benchies....

Ughhh. Either AMD is pulling this biggest sandbagging scheme ever, or it really doesn't look good at all for BD.

 

[Dresdenboy]

Remember, its showing supposed Bulldozer 2 performance comparisons.

However, it doesnt make sense that they compare BD2 against Phenom II. They would be comparing it to BD1 - even if BD1 were slower than Phenom II, it would show how much of an improvement BD2 is over BD1. Showing how much faster it will be compared to Ph2 is pointless.

According to what some people think, the speedups vs. BD1 should be even larger then Is RS indirectly admitting (if this is his work) that he chose PhII to make the speedup appear larger since he assumes BD1 to be faster than PhII X6?

BTW is there any known fact against AMD pulling a K7 type pre-launch/launch performance info as noted earlier in this forum?

 

[BlueBlazer]

http://translate.google.nl/translat...officiele-benchmarks-amd-fx-processor&act=url

Assuming that is a Core i5 2500K (highest Core i5 processor), and using this reference.....

For Core i5 2500K to Core i7 2600K >> (188 * 21.6) / 16.32 = 249 fps

That's faster than FX's 223 fps... :hmm:

The thing about handbrake is that HT isn't really helpfull...

So that FX sample would also best the 2600K by >10%. Basically what you get is an FX8120? that would be as fast as an 2600K @ 3.85GHz. (given Handbrake is pretty much SSE stuff i guess? this would be rather impressive imho.) But according to that they should have run FX8150 vs 990x in that test... because my estimate is they would perform about equally here. (unless the FX8150 is the one priced at the i5 level ofcourse.)

Huh? AMD is pitting the FX against a Core i5 processor (and not Core i7 2600K). Look above for my performance conclusion.

Then the 2nd test was most likely the FX-8150 vs the i7-980x which we all know loses to the i7 2600k in gaming at stock settings. The FX gets a 2-3% win at a crazy high resolution on a game that is most likely cherry picked.

It seems there's a subtle reason they have been demonstrating Dirt only all along. :hmm:

 

[Arzachel]

No bueno. First they are cherry picking, like any company would do. The first test I'm assuming is the FX-8120 vs. the i5-2500k since both are right at $220 doing something that would definitely benefit more cores over anything.

Then the 2nd test was most likely the FX-8150 vs the i7-980x which we all know loses to the i7 2600k in gaming at stock settings. The FX gets a 2-3% win at a crazy high resolution on a game that is most likely cherry picked.

And these are official benchies....

Ughhh. Either AMD is pulling this biggest sandbagging scheme ever, or it really doesn't look good at all for BD.

Cherry picking aside, if BD can tie Nehalem in gaming performance, that bodes well for its IPC and turbo. BRING ON THE PROPER BENCHES!

 

[BlueBlazer]

Cherry picking aside, if BD can tie Nehalem in gaming performance, that bodes well for its IPC and turbo. BRING ON THE PROPER BENCHES!

I've mentioned it before, that AMD FX just matched a 3 year old Nehalem. But if compared clock-to-clock speed parity (considering AMD FX's much higher clock speeds), it will not match Nehalem on IPC basis.

 

[AtenRa]

The title of the article was made by the author, so IMHO it is wrong because i believe AMD only show them ES parts

 

[Soulkeeper]

Sure, but how many people run Gentoo Linux? Or Linux at all? Or even uses Windows with performance-critical open source software? Other than the server or HPC markets, very few. So who can get their software done with the latest compiler to support it? Even less of those very few. So how relevant are benchmarks done with the latest compiler? Not at all to those who want to know how it will perform, but good for those of us that actually want to know just how fast it can be. People on here still cry foul when AVX is used...


Definitely. Note the date – Fall 2013.

It seems you've zeroed in on my sig and assumed some odd things ...
"So who can get their software done with the latest compiler to support it?" <--- I have no energy to point out how this is wrong.

I wasn't speaking of end users. I was referring to the software developers.
You know ... those people behind the curtains that code/compile your magic benchmarks/games/apps.
All binaries (the magic stuff in your self extracting .exe) had to be compiled at one point or another and to expect them to have all the latest optimizations in an older binary you downloaded off some random site is not wise.

All I'm saying is none of these early benchmarks are likely to be worth a damn.
And performance, as with any completely new arch, should increase with time.

 

[BlueBlazer]

The title of the article was made by the author, so IMHO it is wrong because i believe AMD only show them ES parts

ES again (at this late stage)? I don't see anywhere AMD mentions that in the article (except "AMD FX"). The performance seen here is inline with leaked ES benchmarks posted earlier (slower than Phenom II X6 1100T, as rumors stated).....

For Core i5 2500K to Phenom II X6 1100T >> (188 * 19.35) / 16.32 = 224.6 fps

That's still faster than FX's 223 fps.... :hmm:

Performance projections from leaked ES benchmarks, pre-order pricing and now these official benchmarks are now coming into light.

 

[AtenRa]

ES again (at this late stage)? I don't see anywhere AMD mentions that in the article (except "AMD FX"). The performance seen here is inline with leaked ES benchmarks posted earlier (slower than Phenom II X6 1100T, as rumors stated).....

No CPU info was released (no specs no models) (Edit: except it was an 8 core) , if you insist that BD ES CPUs have the same performance as retail BD CPUs thats your call, but IMHO they dont.

 

[Arzachel]

ES again (at this late stage)?

The overclocking WR was done on engineering samples, so it wouldn't surprise me.

 

[Soulkeeper]

Damn it, I think I just stepped into a flame-war thread

maybe I should throw a bunch of in my posts to avoid insult

 

[Riek]

Assuming that is a Core i5 2500K (highest Core i5 processor), and using this reference.....

For Core i5 2500K to Core i7 2600K >> (188 * 21.6) / 16.32 = 249 fps

That's faster than FX's 223 fps... :hmm:

Huh? AMD is pitting the FX against a Core i5 processor (and not Core i7 2600K). Look above for my performance conclusion.

Edit: It seems there's a subtle reason they have been demonstrating Dirt only all along. :hmm:

The problem is handbrake... You have results from certain conversions where 2600 is 32% faster. I have results where 2600 is <5% faster and 11% faster. This is a big difference.

Teh rsults closed to those framerates have actually <11% difference instead of the 32% according to your link. Which means the opposite.

With your number 2600 is 10% faster with my number it is 10% slower.
With my number the FX8150 is 223*1.17 = 260~ 980x188*1.45 = 272

 

[BlueBlazer]

No CPU info was released (no specs no models) (Edit: except it was an 8 core) , if you insist that BD ES CPUs have the same performance as retail BD CPUs thats your call, but IMHO they dont.

My early assessments is that the performance of B0 stepping to B2 stepping (from leaks) hardly improved. Thus IMHO unlikely any magical 50% to 100% improvements in the last minute, for a processor that is priced at $300 and lower. :hmm:

The overclocking WR was done on engineering samples, so it wouldn't surprise me.

If you insist on that, then can we expect more delays (because silicon re-spin takes time, in months)? Also I happen to mention in another thread about lack of any announcement on shipping dates, as well as launch dates. Hopefully it will not be another last minute "60-90 days" as shown by AMD back in June. :hmm:

 

[ransomlist]

If only BD had come out on time in the first place. How the beep are AMD going to make up for lost ground between BD1 and BD2 or enhanced when Intel is going to be surging ahead with Ivy Bridge and haskell.

At the end of the day, if AMD match mid-enthusiast performance at lower prices i'm OK as a consumer. Lets be honest, typical PC use doesn't get more intensive than gaming and as current even budget processors aren't stretched by the latest games and probably wont be until a new gen of consoles allows developers to make a few steps forwards.

Still, as far as competition across its entire range of offerings, AMD isn't in a good place with Bulldozer leaving Intel uncontested at the high end...

 

[BlueBlazer]

The problem is handbrake... You have results from certain conversions where 2600 is 32&#37; faster. I have results where 2600 is <5% faster and 11% faster. This is a big difference.

That reference was based on fps with multi-threaded video transcoding. Its not based on trancoding time, such as this one here. :hmm:

Teh rsults closed to those framerates have actually <11% difference instead of the 32% according to your link. Which means the opposite.

With your number 2600 is 10% faster with my number it is 10% slower.
With my number the FX8150 is 223*1.17 = 260~ 980x188*1.45 = 272

Huh? The article says the AMD FX processor result is 223 fps, then why do you need to multiply the result by 1.17 to inflate it to 260? Where does that 1.17 come from? I'm using Core i5 result of 188 fps from that article as a base.. :hmm: