Here comes the spin machine! (Our Take on AMD FX)

[lifeblood]

In JFAMD's defense, he didn't state any rediculous claims. He vehemently stated IPC would NOT decrease.

AMDs goal was to keep IPC the same as PhII. His vehement insistence that IPC was not worse so late in the game when the facts must have been known is, in my book, a ridiculous claim. He is senior enough to have known the truth.

His only hope for salvation is the fact he is talking about the server BD, not the desktop BD. The server version of BD may actually maintain IPC.

 

[SlowSpyder]

AMDs goal was to keep IPC the same as PhII. His vehement insistence that IPC was not worse so late in the game when the facts must have been known is, in my book, a ridiculous claim. He is senior enough to have known the truth.

His only hope for salvation is the fact he is talking about the server BD, not the desktop BD. The server version of BD may actually maintain IPC.

I wonder if there will be some sort of BD-optimizer that will keep threads 1-4 on modules and not bring the individual cores into play until threads 5-8. When looking at a per modual basis, maybe IPC will be better on the desktop and better than PhII. But as it is, if threads 1-2 are all on a single module, well, not so much.

 

[Genx87]

So AMD created BD, not for todays Apps, but future apps. Ok, I can understand that. If they think their module/core architecture is superior and can give real benefits to the user, I'm all for it. But in order for this to work you need the software. AMD should have been working on MS to bring out a Win7 scheduler patch at launch and not wait for Win8. They should have worked with all the other vendors whose software would benefit from the new design and have patches for them as well. Maybe not even patches to take advantage of everything, but at least some things.

nVidia once described itself as a software, not a hardware company. AMD would do well to learn that lesson and work better with software vendors to support their hardware.

This is AMD we are talking about. Where they like to talk the talk, but not walk the walk. If software support is needed. AMD wont be there. I am surprised they havent blamed Intel for software not being written for Bulldozer.

If they think Windows8 will vindicate them. Why the hell would they release the chip a solid 12-15 months before the launch of the OS?

I am with keys. I think they are going to go out and find somebody to buy them. I'd think from a patent perspective Apple would eye AMD as a target.

 

[BlueBlazer]

Overclocking PR disaster at Newegg TV: AMD FX-Series Bulldozer Overclocking with J.J. from ASUS. After overclocking to 4.8GHz, watch carefully the Cinebench R11.5 score. 😀

 

[Vdubchaos]

Wait, was that an AMD engineer running 580GTX?

😕

 

[zlejedi]

Wait, was that an AMD engineer running 580GTX?

😕

Maybe they needed something with good drivers or proper gpu programming support 😉

 

[carnage10]

Wait, was that an AMD engineer running 580GTX?

😕

Nah he works for Asus not AMD.

 

[Dresdenboy]

maybe we'll start seeing a required CPU Driver to make it work better 🙂

A CPU driver won't be enough. If everything in the hardware is working as intended (BIOS is fine), some core affinity tooling might help a bit, similar to the Win 8 scheduler.

But what I see being important is optimized code. We will soon conduct some tests in this regard. This radically different architecture is much more sensible to wrongly optimized code. This could be a result of optimizing logic for code following certain rules w/o overprovisioning of HW ressources to be good at many different cases.

 

[Idontcare]

Your right. I am done here . its been a pleasure. You guys take care and watch out that the truth doesn't set you free or you end up like me . According to what I just read . There be no place left in this world for truth . Bye guys be safe live long and die well.

Good bye Nemesis 1

Drama much?

Not sure what I posted that got you so defensive. Was trying to give you some advice by suggesting your current approach is doing more to undermine you than it is undermining JF.

In the court of public opinion the onus is on you to provide the links you are referring to when making the claims you do in your post. When has this not been the case?

 

[BlueBlazer]

Wait, was that an AMD engineer running 580GTX?

😕

He was from ASUS. Its either the ASUS tuning software failed to overclock properly, or the AMD FX processor was throttling. 😛

 

[frostedflakes]

In JFAMD's defense, he didn't state any rediculous claims. He vehemently stated IPC would NOT decrease. Doesn't that sound reasonable? Why would this seem outrageous for a brand-new product?

I think as he knew more and more about issues relating to performance, he stuck to his story. That's just my thoughts on what happened, definitely no proof here. I would like to hear his side of the story, if he comes back to the forum.

If you'll notice his more recent comments, he doesn't stick to that story. In the Bulldozer pre-launch FAQ he posted, for example, he brushes off IPC as irrelevant instead. Back in 2010 or whenever he made the earlier comments about IPC that enthusiasts latched on to, that's probably what he had been told and what had been expected of Bulldozer by people inside AMD. But as they started to get their hands on actual silicon and get closer to launch, what they expected and what Bulldozer actually delivered may have been quite different.

Q. Is IPC higher on bulldozer? All I care about is IPC.
A. IPC is simply a measurement. What if IPC was 2X what it is today, but clock speed was 500MHz. Is that what you want? You are getting double IPC, right? IPC is only one measure. The people that are telling you IPC is the only thing that matters have an agenda. Taking only one measurement, out of context, is like trying to say that a person's weight is all that matters. I weigh 195. Does that make me fat? Does that make me skinny? It is impossible to say unless you know my height. IPC is like weight - it tells you something in context to other factors, but is meaningless on its own.

 

[Vdubchaos]

That video should've ended with Newegg guy saying "riveting tale chap"

 

[Idontcare]

He was from ASUS. Its either the ASUS tuning software failed to overclock properly, or the AMD FX processor was throttling. 😛

Did they mention temps, or monitor them?

 

[BlueBlazer]

Did they mention temps, or monitor them?

Nopes, they never bothered since they were busy talking about overclocking. 😀

 

[mrbrout]

I was very interested that several of you here posted the uncovered incorrect AMD sales video claiming to compare Bulldozer to the I7-980E when in fact it was the I-5 2500..

So I mentioned this fact over at the AMD users forums, http://forums.amd.com And for that single FYI post my post was not only deleted my entire forums account there was banned. I had gone so far in that post to mention it was not a rage or some fanboy post but something I felt people should know about. I know as a computer enthusiast I would be interested in falsified benchmarks from AMD marketing.

I though amd.com was a legitimate users group?

 

[Idontcare]

In JFAMD's defense, he didn't state any rediculous claims. He vehemently stated IPC would NOT decrease. Doesn't that sound reasonable? Why would this seem outrageous for a brand-new product?

I think as he knew more and more about issues relating to performance, he stuck to his story. That's just my thoughts on what happened, definitely no proof here. I would like to hear his side of the story, if he comes back to the forum.

If you'll notice his more recent comments, he doesn't stick to that story. In the Bulldozer pre-launch FAQ he posted, for example, he brushes off IPC as irrelevant instead. Back in 2010 or whenever he made the earlier comments about IPC that enthusiasts latched on to, that's probably what he had been told and what had been expected of Bulldozer by people inside AMD. But as they started to get their hands on actual silicon and get closer to launch, what they expected and what Bulldozer actually delivered may have been quite different.

Q. Is IPC higher on bulldozer? All I care about is IPC.
A. IPC is simply a measurement. What if IPC was 2X what it is today, but clock speed was 500MHz. Is that what you want? You are getting double IPC, right? IPC is only one measure. The people that are telling you IPC is the only thing that matters have an agenda. Taking only one measurement, out of context, is like trying to say that a person's weight is all that matters. I weigh 195. Does that make me fat? Does that make me skinny? It is impossible to say unless you know my height. IPC is like weight - it tells you something in context to other factors, but is meaningless on its own.

Sorry to say but that is classic "redirection" on behalf of JFAMD and there is a reason people hate it and despise it when the other side of the discussion falls back and resorts to it.

Saying "IPC doesn't matter" is a strawman. IPC matters, as does clockspeed. We all get this, its swag (stuff we all get).

We had good estimates of where clockspeed was going (plenty of "30% higher targeted clockspeed" comments all over the interwebz)...the BIG unknown was IPC.

Why were we focused on IPC? Because to the untrained laymen, bulldozer was looking like netburst all over again, thus the reasoning for making the assurances in the first place that IPC was not going to decrease.

Had we not been told, steadfastly, repeatedly, that IPC did not decrease then we, as a community, would have had VERY different discussions in regards to bulldozer expectations this past year.

IPC matters, if it didn't then JFAMD would never have been compelled to speak to IPC not decreasing in the first place, he would have said "IPC doesn't matter, clocks are king, didn't Willamette teach you nothing?".

Coming back now, attempting to rewrite history, is just insulting to the community that forced itself to accommodate the IPC comments as fact while we also reconciled it with respect to all the other stuff we get (clockspeeds, gate first vs last, SOI vs bulk, voltage, etc).

It was a critical linchpin in THE debate, and John knew it which is why he went to such lengths to make sure everyone heard him and his statements about IPC not decreasing.

 

[Keysplayr]

This is AMD we are talking about. Where they like to talk the talk, but not walk the walk. If software support is needed. AMD wont be there. I am surprised they havent blamed Intel for software not being written for Bulldozer.

If they think Windows8 will vindicate them. Why the hell would they release the chip a solid 12-15 months before the launch of the OS?

I am with keys. I think they are going to go out and find somebody to buy them. I'd think from a patent perspective Apple would eye AMD as a target.

Apple would be my fist choice. But they are doing so well as they are, about to surprise exxon mobil for the wealthiest company in the world. They might not want to mess with that. But who know now with Jobs gone. I'm with you though.

 

[BlueBlazer]

I was very interested that several of you here posted the uncovered incorrect AMD sales video claiming to compare Bulldozer to the I7-980E when in fact it was the I-5 2500..

So I mentioned this fact over at the AMD users forums, http://forums.amd.com And for that single FYI post my post was not only deleted my entire forums account there was banned. I had gone so far in that post to mention it was not a rage or some fanboy post but something I felt people should know about. I know as a computer enthusiast I would be interested in falsified benchmarks from AMD marketing.

I though amd.com was a legitimate users group?

Could possibly be the same reason AMD shutdown the comments section on their Youtube video. :hmm:

Maybe they needed something with good drivers or proper gpu programming support 😉

I think its more likely the Cooler Master Hyper 212+ could not cool down the processor sufficiently when overclocking. Even the OC3D guy mentioned about the enormous heat when overclocking. :hmm:

 

[mrbrout]

Kudos Idontcare. Civility should always precede a position but that doesn't negate the demand for integrity and accountability. I think you illustrated this above quite well.

 

[Keysplayr]

Kudos Idontcare. Civility should always precede a position but that doesn't negate the demand for integrity and accountability. I think you illustrated this above quite well.

Welcome. So they didn't treat you so well at amd forums? Do you have any links? Or did they make it so you had never existed there?

 

[mrbrout]

They made it as if I never existed! No warning nothing.. just deleted post and my account was deleted. Thing is, It wasn't a troll post there.. it was just a mention of the i7-980 false benchmark in AMD's video. I linked the screenshot proof and even another website's Cinebench score showing AMD's bench was swapped. So it wasn't like I just went in making blanket unfounded accusations. And I prefaced it as an FYI to clarify I was not some Intel Fanboy troll there to start trouble....

Then.. Bam! no warning.. nothing.. post deleted and my entire account deleted as if I never existed. Looks like Blatant censorship to me.. and though I don't like to tatel, I felt it worth mentioning here since you guys broke this false amd benchmark video.

 

[exar333]

Sorry to say but that is classic "redirection" on behalf of JFAMD and there is a reason people hate it and despise it when the other side of the discussion falls back and resorts to it.

Saying "IPC doesn't matter" is a strawman. IPC matters, as does clockspeed. We all get this, its swag (stuff we all get).

We had good estimates of where clockspeed was going (plenty of "30% higher targeted clockspeed" comments all over the interwebz)...the BIG unknown was IPC.

Why were we focused on IPC? Because to the untrained laymen, bulldozer was looking like netburst all over again, thus the reasoning for making the assurances in the first place that IPC was not going to decrease.

Had we not been told, steadfastly, repeatedly, that IPC did not decrease then we, as a community, would have had VERY different discussions in regards to bulldozer expectations this past year.

IPC matters, if it didn't then JFAMD would never have been compelled to speak to IPC not decreasing in the first place, he would have said "IPC doesn't matter, clocks are king, didn't Willamette teach you nothing?".

Coming back now, attempting to rewrite history, is just insulting to the community that forced itself to accommodate the IPC comments as fact while we also reconciled it with respect to all the other stuff we get (clockspeeds, gate first vs last, SOI vs bulk, voltage, etc).

It was a critical linchpin in THE debate, and John knew it which is why he went to such lengths to make sure everyone heard him and his statements about IPC not decreasing.

I agree 100 percent. What I was trying to say is that JFAMD made no ridiculous or un-plausible comments. He said what most here wanted to hear, that BD was 'good'. I don't remember him saying it was amazing and would blow us away.

Did he know that IPC would go down? Probably. Did he mention this? No, he tried to use mis-direction. Could you imagine our surprise here (and many other places) if he went on record to say IPC went DOWN!? People here and on many other forums were expecting Nehalem-like IPC with awesome MT capabilities.

I am not covering for him, but I am saying what he said was very easy to listen and was what we were all HOPING for. A solid performer, no slower in IPC than PhII and better than the 2600k for most multi-threaded titles.

As soon as the early benchmarks came out, and we saw the non-confirmed numbers, then we really started to question the performance, power usage, and clocks.

A debacle in all sense of the word.

 

[drizek]

Nobody was expecting Nehalem-like IPC. I think everyone was expecting 10% better than Phenom II with 5GHz+ overclocks on 8 cores with 180% scaling.

We were expecting the 8150 to be 50% faster with 33% more cores relative to the 1090T.

With power consumption we were expecting lower idle power draw and equal load power draw to Thuban.

Bulldozer is a disappointment across the board.

It's funny, we all assumed that AMD had been keeping an extremely tight lid on leaks of BD performance. They weren't. It's just that their CPU was so crap that none of us could believed the benchmarks when we saw them.

 

[intangir]

Nobody was expecting Nehalem-like IPC.

There were certainly people arguing for it!

http://forums.anandtech.com/showpost.php?p=32259594&postcount=1650

Luckily, no one now disputes that AMD was aiming for Phenom II IPC. If they do, I have at least 3 Bulldozer reviews I could use as a secondary source that state plainly that's what AMD told them. The ISSCC 2011 paper I used as a primary source wasn't enough proof to convince some people, apparently.

On the bright side, the module penalty does seem to be about -10% between 2M/2C and 1M/2C mode, resulting in 180% or more scaling in most circumstances. Too bad good performance/thread scalability and good single-threaded performance are opposing goals. I believe Linus Torvalds had a rather notorious rant about this. I could dig that up too, if anyone would like to argue that point!

 

[Lonbjerg]

Nobody was expecting Nehalem-like IPC. I think everyone was expecting 10% better than Phenom II with 5GHz+ overclocks on 8 cores with 180% scaling.

We were expecting the 8150 to be 50% faster with 33% more cores relative to the 1090T.

With power consumption we were expecting lower idle power draw and equal load power draw to Thuban.

Bulldozer is a disappointment across the board.

It's funny, we all assumed that AMD had been keeping an extremely tight lid on leaks of BD performance. They weren't. It's just that their CPU was so crap that none of us could believed the benchmarks when we saw them.

Some of us wasn't very surpised...the leaks were very consistant and when JFAMD suddenly didn't mean that IPC mattered...I knew performance was an issuse.