AMD FX-8120P benchmark from Coolaler

[Accord99]

The ignorance just amazes me. All AMD needed to do is take 2 llano dies, strip out the GPU, and tie together what is left, and they would have an 8 core chip that is faster AND smaller than this BD. They would have known this for at least 6 months, but they did not do what I just suggested (as far as we know). The only logical conclusion is that BD has to be much faster than what is being shown here. Obviously. Billion dollar companies do not shovel out garbage when it would be much cheaper and faster ttm for them to produce a cookie cutter 8 core llano. You people who cant understand that need to get over yourselves.

It's been said that the semiconductor industry is like playing Russian Roulette, except it takes 3+ years to know whether you shot out your brains or not. It's quite conceivable by the time BD was ready and performance was known, it was too late to make a 8-core Llano. And Llano has pretty high power consumption as is, making it a tough starting point.

Even Intel can't move so fast, as when it was forced to use Prescott even as Dothan would have made a far better consumer processor.

 

[acx]

lol hogwash, a bad idea is a bad idea. You can't blame business folk for being convinced accredited engineers are leading them in the wrong direction.

Hindsight is 20/20 and sometimes you only see that ideas didn't work out after you've implemented them. Architectures are planned out 4-5 years in advance. Assumptions about business conditions, manufacturing capability, and market timing are all vague estimates at that time. Sometimes your assumptions were wrong and things just don't turn out as you had hoped. Look at Intel's Itanium project. At the time it was developed, it was supposed to take us past x86. But it turned out x86 ISA was so entrenched Intel couldn't kill off it's own ISA. And the super smart compiler ideas never quite panned out. Look at Intel's Pentium 4. They projected that manufacturing advances would keep power in check. But those advances didn't happen and Intel ran into a power wall. At the time of Itanium and Pentium 4's architectural inception could the engineers have identified those assumptions as bad? Who takes the blame when something doesn't pan out? Are you going to fire the entire engineering staff or the project managers? Is it their fault if the information they were given were incomplete or they can't predict the future with 100% accuracy?

 

[RussianSensation]

It is fairly obvious that this is a server chip trying to be a desktop part, when it is not.
This also falls in line with what JFAMD has been saying.

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!!!"

 

[inf64]

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?

Well maybe at that time,that was the info they had from engineering team? Now something has changed(problems in design?) and Bulldozer performs worse.
Until launch,everything is speculation anyway. What we will know soon is Opteron's performance. If it mimics Zambezi's behavior in legacy SIMD workloads,then it won't be much(if at all) of an improvement over MC.

 

[nyker96]

somehow i am in disbelief, one one hand we got these terrible leaked benches, on other hand AMD is actually pricing the fx8xxx series on par w/ 2500k and 2600k. So if the benches are so bad, i somehow doubt AMD will even try to price it close to the 2x00ks. still all the delay points to some serious engineering problems. What the hell did they do in the last 3-4 years in R&D?

 

[Gloomy]

What the hell did they do in the last 3-4 years in R&D?

Their R&D department has been taking hip hop dance classes for the past 5 years. All but one of them are now capable of head spinning for hours on end.

This is preparation for an upcoming dance competition, or "battle" as it is known in the world of CPU design. Intel is going to get served.

 

[Crap Daddy]

somehow i am in disbelief, one one hand we got these terrible leaked benches, on other hand AMD is actually pricing the fx8xxx series on par w/ 2500k and 2600k. So if the benches are so bad, i somehow doubt AMD will even try to price it close to the 2x00ks. still all the delay points to some serious engineering problems. What the hell did they do in the last 3-4 years in R&D?

Where have you seen on par pricing with 2600K? From this:

http://www.anandtech.com/show/4767/pricing-revealed-for-first-amd-bulldozer-fx-chips

we know the top AMD FX is 50$ cheaper. So guess what will be the performance?

 

[Abwx]

Where have you seen on par pricing with 2600K? From this:

http://www.anandtech.com/show/4767/pricing-revealed-for-first-amd-bulldozer-fx-chips

we know the top AMD FX is 50$ cheaper. So guess what will be the performance?

Because AMD has as habit to sell at the same prices for equivalent perfs ??..

 

[beginner99]

Bulldozer is not and 8 core chip in any sense of the word. It's a quad core, which is why it performs like one. It's looking the the Pentium 4 all over again, sell them on cores and or frequency and screw them on performance.

this. marketing fail. If the would have gone with the module ( eg 4 module) name it would not look that bad. It looks like floating point is much more used than AMD thought or at least in cinebench?

Considering that for FP it is a 4-core chip and all the other downisdes 80% of a dual core for a module seems a lot. HT seem to be the better choice in terms of performance gain per die area and watt.

 

[badb0y]

LOL Thuban>Bulldozer. Jakes on you AMD.

 

[Elixer]

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!!!"

That is the problem, it is all if..if..if. If these benchmarks are real, and it is starting to look like that they are all fakes, or at least they are using gimped CPUs or BIOS issues or...
Looking around at other forums, we see posts from people who were invited by AMD to check out what they have, and, his remarks say otherwise from what these benchmarks show.

http://www.xtremesystems.org/forums...-at-8429-MHz&p=4949106&viewfull=1#post4949106

I cheer good products. Blue,Green or pink, doesn't matter.
Now there are things I can't talk about yet but I can say I think they have a winner here.
I saw the benches...

JFAMD also has been posting about fake cpu-z shots here:
http://www.overclock.net/amd-cpus/791495-bulldozer-blog-live-514.html#post14932458

Which basically all boils down with, we just don't know, until AMD wants us to know.

 

[Idontcare]

No, at least I don't.

Daimon

http://en.wikipedia.org/wiki/Dirk_Meyer#Career

He was a co-architect of the Alpha 21064 and Alpha 21264 microprocessors during his employment at DEC and also worked at Intel in its microprocessor design group.[2]
Meyer joined AMD in 1996, where he personally led the team that designed and developed the Athlon processor.
Dirk Meyer was formerly president and chief executive officer of AMD. At one time, he was the chief operating officer. In this role, he shared leadership and management of AMD with former Chief Executive Officer and Chairman of AMD Hector Ruiz.[3] Prior to this role, Meyer served as president and chief operating officer of AMD’s Microprocessor Solutions Sector where he had overall responsibility for AMD’s microprocessor business, including product development, manufacturing, operations and product marketing.
On July 18, 2008, Dirk Meyer replaced Hector Ruiz as the CEO of AMD.

^ you aren't going to fool Dirk into going with a polished turd of a microarchitecture.

If Bulldozer is borked, if the results in this thread are truly indicative of the performance levels of an 8120P, then it has to be because of some risks the design team took in their efforts to make an aggressive microarchitecture and those risks were probably viewed as acceptable by Dirk. Captain of the ship and all that.

 

[Black96ws6]

More info, if that leak is legit, then why does this guy say what he says in this post?

http://www.xtremesystems.org/forums...-at-8429-MHz&p=4949106&viewfull=1#post4949106

 

[badb0y]

More info, if that leak is legit, then why does this guy say what he says in this post?

http://www.xtremesystems.org/forums...-at-8429-MHz&p=4949106&viewfull=1#post4949106

Who is this guy?

 

[RussianSensation]

Looking around at other forums, we see posts from people who were invited by AMD to check out what they have.

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!

 

[Black96ws6]

Who is this guy?

No idea! But if you keep reading in that thread (http://www.xtremesystems.org/forums...-at-8429-MHz&p=4949106&viewfull=1#post4949106), everyone else seems to think very highly of him, and he does have over 30,000 posts!

Also looks like he had his hands on the latest silicon and saw actual benchmarks...

 

[Arkaign]

Hahaha, apparently BD2 is 6X faster than BD1? Bravo AMD, bravo.

 

[sm625]

It's quite conceivable by the time BD was ready and performance was known, it was too late to make a 8-core Llano. And Llano has pretty high power consumption as is, making it a tough starting point.

Llano really doesnt have that high a power consumption. AMD just dumped the chip on the market without making any attempt to bin for lower voltage.

There is no reason to believe AMD couldnt bring a K10 style Athlon X8 @ 2.6GHz and 1.2VID (95W TDP) at yields good enough to sell them for $150 apiece in less than 6 months. It would be no bigger than thuban, and die harvested parts could sell as faster quads and hexes. They could easily have made a 1035T replacement that cost no extra money for 2 extra cores. They could have easily done this 6 months ago, yet chose not to, in favor of a design that would... perform worse and sell less and make less money and yet cost more to produce? Yeah right.

A year from now even the release stepping BD will outscore an 1100T in cinebench by at least 40% clock for clock. Bookmark this if you dont believe me.

 

[badb0y]

BD2 release date 2013 2016

 

[Eeqmcsq]

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!

Hahahaha! Nice photoshop.

 

[GaiaHunter]

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!

Movieman is an Xtreme Legend and he actually was there.

 

[lol123]

http://en.wikipedia.org/wiki/Dirk_Meyer#Career



^ you aren't going to fool Dirk into going with a polished turd of a microarchitecture.

If Bulldozer is borked, if the results in this thread are truly indicative of the performance levels of an 8120P, then it has to be because of some risks the design team took in their efforts to make an aggressive microarchitecture and those risks were probably viewed as acceptable by Dirk. Captain of the ship and all that.

If that's the case and there was nothing to suggest that the architecture was going to end up as a failure, then why was Dirk fired?

 

[BlueBlazer]

That is the problem, it is all if..if..if. If these benchmarks are real, and it is starting to look like that they are all fakes, or at least they are using gimped CPUs or BIOS issues or...
Looking around at other forums, we see posts from people who were invited by AMD to check out what they have, and, his remarks say otherwise from what these benchmarks show.

http://www.xtremesystems.org/forums...-at-8429-MHz&p=4949106&viewfull=1#post4949106

Eh? Kyle at [H] gave some better hint than that...

I think you will be happy with it. Not ultrafuckingblown out the water with it, but after OCing I think you will be good with your purchase.

And also this hint....

This is the board we will be using for BD overclocking launch articles.

Revelations? 🙂

JFAMD also has been posting about fake cpu-z shots here:
http://www.overclock.net/amd-cpus/791495-bulldozer-blog-live-514.html#post14932458

Which basically all boils down with, we just don't know, until AMD wants us to know.

JF-AMD again? He belittles (tries to wave them off) every information even though some IMHO are reliable (while he tries to paint them all are "fakes"). I think some of his statements are down right "insinuating" (also has cause furors with others). I could give you one recently here....

Yes, back in august there were all kinds of benchmarks on "final" silicon. I could not mention that we had not reached production silicon either.

The funniest thing is that I could give 2 people (an intel fanboy and an amd fanboy) the exact same piece of silicon and both could give you different benchmark results. Why would anyone assume that all of these fake benchmarks, if they were real, were actually on an optimized system? Couldn't you run a 4200RPM HD, 800MHz memory, and play all kinds of BIOS tricks? I know that I could probably get about 80% of the optimizations done to get the best performance out of a platform, but I know I could do 100% of the things that would sabotage performance. If you think about it, AMD is not releasing info, so the stuff that is floating around (if it is even real) is probably from people not friendly to AMD. And somehow everyone is trusting those results. Nobody ever questions the motives of the people supplying this info, but to be fair, because it is third or fourth hand (some guy I know who has a friend who has an uncle that knows a guy in...) how can anyone vouch for the results.

It can be interpreted many ways (depending on your mindset). Have other examples from other boards. :hmm:

 

[BD231]

You do know Dirk's background, right? What he did before going to AMD, what he did when he got to AMD, and why he was promoted up the chain to eventually become the CEO, yes?

I do know his background cost him his job ...

Hindsight is 20/20 and sometimes you only see that ideas didn't work out after you've implemented them. Architectures are planned out 4-5 years in advance. Assumptions about business conditions, manufacturing capability, and market timing are all vague estimates at that time. Sometimes your assumptions were wrong and things just don't turn out as you had hoped. Look at Intel's Itanium project. At the time it was developed, it was supposed to take us past x86. But it turned out x86 ISA was so entrenched Intel couldn't kill off it's own ISA. And the super smart compiler ideas never quite panned out. Look at Intel's Pentium 4. They projected that manufacturing advances would keep power in check. But those advances didn't happen and Intel ran into a power wall. At the time of Itanium and Pentium 4's architectural inception could the engineers have identified those assumptions as bad? Who takes the blame when something doesn't pan out? Are you going to fire the entire engineering staff or the project managers? Is it their fault if the information they were given were incomplete or they can't predict the future with 100% accuracy?

Do I know why Bulldozer is a mess? No. Dose Mr. Dirk have a clue? Of course not, he wasn't the one working on the chips and that wasn't ever his job with AMD.

To be fair new arc's are hit or miss which I can definitely agree on but With llano showing design issues as well I don't buy a lack of direction alone. Simply put you don't go asking a design team to make a cpu that requires massive advances in FPU/IPC performance just to support the additional hardware you're tacking on as your main competing product.

In my eyes they shot themselves in the foot by banking on what is essentially their weakest area of expertise compared to the competition (FPU/IPC), and I'd like to think these multi-billion dollar companies work as a collective when it comes to making those decisions.

The issue is that these blunders are sent down the chain and people still buy the sh t. AMD came out hardcore with Multi-Core cpu's back in the day, now all of the sudden they're cheaping out by selling non-traditional multicore CPU's that will be advertised as something more than they are. It was clearly a cost cutting move from the start.

 

[Idontcare]

If that's the case and there was nothing to suggest that the architecture was going to end up as a failure, then why was Dirk fired?

That's not what I am saying, I'm saying there might have been a lot of red flags about the design goals but Dirk stuck his neck out.

Given his legacy at AMD, and the fact he was CEO, who was going to tell him no? But if the thing showed up from the fab and was broken in ways that gave a lot of naysayers the opportunity to say "I told you so" then the BoD's may have felt push came to shove.

At any rate, I take MM's word over Collaler's leaked benchmark. If MM has seen the benches and he's willing to go public to say they will impress then that is all I need to hear as reassurance that BD is still going to deliver as JFAMD guided us to expect (IPC increases, not decreases).

Leaked benches must be bogus or useless for any number of reasons. Time to move on from this thread.