AMD Zambezi news, info, fans !

[informal]

i tried to get clarity on what the 80% was because its clearly confusing people,
i got an answer, and i trust that answer until we are told it was a mistake.

I have no problem with that . But you have hints now,by a several users (chew* being the most respected one) that this "8C" part performs more like 4C/8T part,which is perfectly in line with Analyst Day presentation. And you have some benchmarks too,although on non final hardware,that shows similar behavior (100x mentioned cinenebench). Note that the cinebench score from far east matches what *M* said in the press briefing some weeks ago.

BTW Manicdan,from the same link you posted ,I have found this:

So, let me take on these statements:

“Aren’t these two premises in conflict?”
No, actually they are not. You are looking at 2 different comparisons. The “80%” number is not performance relative to previous generation. 80% is the throughput of two threads running on a single module vs. 2 threads running on two completely independent cores.

There you go,even JF said it.

[Solus Corvus]

Frankly, I think you guys are trying to extrapolate too far from too few data points. These statements like "80%" and "IPC increases" aren't enough information to derive a statistically useful prediction, IMO.

[demonkevy666]

bulldozers front end isn't bigger then a phenom II x4.

it still the same 2 way 64kbytes. In the instruction cache.

.....

[Opteron146]

Get it here, if you have the AsRock 990FX Extreme4 board:

http://www.asrock.com/mb/download.as...xtreme4&o=BIOS
Thx@Ronny145.

Should be now really the final version, last version that I saw in the wild was V0.0.9.2

I hope all the testers will get that an Update soon, too.

@xsecret:
Which Agesa were you using? Did you already have a beta BIOS for your board with V1.1.0.0?

[Manicdan]

I have no problem with that . But you have hints now,by a several users (chew* being the most respected one) that this "8C" part performs more like 4C/8T part,which is perfectly in line with Analyst Day presentation. And you have some benchmarks too,although on non final hardware,that shows similar behavior (100x mentioned cinenebench). Note that the cinebench score from far east matches what *M* said in the press briefing some weeks ago.

BTW Manicdan,from the same link you posted ,I have found this:


There you go,even JF said it.

the reply to my post was after that one.

[informal]

Get it here, if you have the AsRock 990FX Extreme4 board:

http://www.asrock.com/mb/download.as...xtreme4&o=BIOS
Thx@Ronny145.

Should be now really the final version, last version that I saw in the wild was V0.0.9.2

I hope all the testers will get that an Update soon, too.

@xsecret:
Which Agesa were you using? Did you already have a beta BIOS for your board with V1.1.0.0?

Do you think new Agesa version is going to bring any difference? How much of a difference did the previous change bring? (sorry I didn't track the numbers but I know chiphell user did post something related to this;IIRC he got somewhat better numbers,primarily due to Turbo functioning better).

[Opteron146]

Do you think new Agesa version is going to bring any difference? How much of a difference did the previous change bring? (sorry I didn't track the numbers but I know chiphell user did post something related to this;IIRC he got somewhat better numbers,primarily due to Turbo functioning better).

The first leaks were done with an Agesa 0.0.4.X, these were the ones around April, cinebench score was @4.8X with that. Currently we are around a score of 5.8, with 0.0.9.X. So it gave a nice speed-up, however there was also the change from B0 to B2 silicon. So it is not clear if it was due to the BIOS code or due to the revision.

In any way, I do not expect lower scores ^^

Furthermore, do not forget Movienan's comment here:
http://www.xtremesystems.org/forums/...=1#post4950183

Would fit ;-)

[informal]

Ok at least that is encouraging . Further lowering of the scores would be indeed bad^^. We are now at thuban's level or somewhat above it. Not very good but not slower at least.

[Mechanical Man]

I have no problem with that . But you have hints now,by a several users (chew* being the most respected one) that this "8C" part performs more like 4C/8T part,which is perfectly in line with Analyst Day presentation. And you have some benchmarks too,although on non final hardware,that shows similar behavior (100x mentioned cinenebench). Note that the cinebench score from far east matches what *M* said in the press briefing some weeks ago.

Actually, in my opinion that slide tells the same, second core adds 80% compared to cmp on average. Think about average usage case, its not using full power of cpu really.

BTW Manicdan,from the same link you posted ,I have found this:


There you go,even JF said it.

You are taking it out of context.

[freeloader]

Engineers > Marketing when it comes to what's really what.

yeah! Any day of the week. Assuming 1.75x the perf. of two cores, maybe AMD should have called it a 7 core.

[PerryR]

Yes, it's an anagram of DNA. What did I win ?

So to it can be rearranged to the abbreviation of the words "do not answer?" I'm glad we've reached an agreement my friend.

For me, it's the same thing as movieman (or whatever his name) saying that AMD has a winner etc...

Except that movie-man said he "thinks" they have a winner, and didn't specify (name) any benches where the product fails/wins. Right? Right. You can gladly correct me if I'm wrong.

I've enjoyed participating in this intelligent discussion.

[FlanK3r]

From "noname man" (we know person) blog

Some Bulldozer comment from my reader Franck7511:

Well, please publish my comment in the name of "Franck7511".

First times, I didn't believe in your leaks.
But I had documents from AMD on Bulldozer performance.
Performance is so poor that it can compete with the 2600K only in multi-threaded benchs...(otherwise, Intel Sandy Bridge processors will be WAY faster !).
Despite a higher frequency, and 2 times more cores ! (A Bulldozer Module should be renamed as a "Core"...)
My estimations are that a Bulldozer Core is only 80% of a Stars Core at equal frequency : that's why a FX4 at 4.2 GHz only compete with the 955...

The overclocking is the only thing good...


???What do u think about it???

[metalazzo]

I think that he doesn't know how to configure his rig and that drivers/magical bios are not up to date (according to JF-AMD).

[chew*]

I think that any info other than what I or AMD have personally offered which is not much if any can be discarded until further notice.

There are still peices to the puzzle missing that I can assure you 99.9% don't have yet regarding CPU rev and bios support and agesa.

To sum up BD facts

BD is physically a 4 core 8 thread part.

It has no coldbug

Samples can bench at 5+ on stock cooler, 6+ on phase change and 8.4 on lhe.

And most important of all SATA works

Like i said before the joke is on that guy we won't name, he was sent intentionally quite possilby the worst chip ever produced. 6.4 on ln2,
or he needs to learn how to OC, 6.4 is my validation speed on phase change......he who laughs last laughs best.............

[massman]

What I remember from this long thread is Chew's comment regarding the 8 cores. It's quite funny how people look at the performance as this 'eight-core' chip and scream it's really bad, but if it were 'four-core with HT' people would say it's doing okay price/perf comparing to 2500K/2600K.

Even though it's the exact same product, opinions are different.

[PerryR]

I think that any info other than what I or AMD have personally offered which is not much if any can be discarded until further notice.

There are still peices to the puzzle missing that I can assure you 99.9% don't have yet regarding CPU rev and bios support and agesa.

To sum up BD facts

BD is physically a 4 core 8 thread part.

It has no coldbug

Samples can bench at 5+ on stock cooler, 6 on phase change and 8.4 on lhe.

And most important of all SATA works

Let me quote you for emphasis.

Even though it's the exact same product, opinions are different.

Well, you can imagine what the complaints will be on day of release/actual benches are released, "They need an 8 core to perform on the same level as a quad core!"

(sigh) Fun times lie ahead.

[Lightman]

FlanK3r, I think it makes sense.
Still, power consumption will be make or brake deal for BD. It should be better than Phenom X6 in a lot of cases, but we won't get ST killer we all hoped for.
Anyway my AM3+ socket will part away with Phenom X6 in few days, so I hope to fill it with decent performance chip soon.

[Olivon]

I think that any info other than what I or AMD have personally offered which is not much if any can be discarded until further notice.

There are still peices to the puzzle missing that I can assure you 99.9% don't have yet regarding CPU rev and bios support and agesa.

To sum up BD facts

BD is physically a 4 core 8 thread part.

It has no coldbug

Samples can bench at 5+ on stock cooler, 6 on phase change and 8.4 on lhe.

And most important of all SATA works

Like i said before the joke is on that guy we won't name, he was sent intentionally quite possilby the worst chip ever produced. 6.4 on ln2,
or he needs to learn how to OC, 6.4 is my validation speed on phase change......he who laughs last laughs best.............

The "AMD FX CPU (...) | August 31, 2011" document this guy talked about is not the final performance version chew* ?

[drfedja]

I think that any info other than what I or AMD have personally offered which is not much if any can be discarded until further notice.

There are still peices to the puzzle missing that I can assure you 99.9% don't have yet regarding CPU rev and bios support and agesa.

To sum up BD facts

BD is physically a 4 core 8 thread part.

It has no coldbug

Samples can bench at 5+ on stock cooler, 6 on phase change and 8.4 on lhe.

And most important of all SATA works

Like i said before the joke is on that guy we won't name, he was sent intentionally quite possilby the worst chip ever produced. 6.4 on ln2,
or he needs to learn how to OC, 6.4 is my validation speed on phase change......he who laughs last laughs best.............

According to VR-Zone CB10 results, BD module scales very well with threads ~70% with second thread on module.

For eg. there is xCPU result for FX8 17800 CB10 points. For 1CPU is only 3450-ish points. It is ~5.16x scaling factor. Six core Thuban without turbo has scale factor ~4.9-5. According to Amdahal's law hypothetical 8th core Thuban could scale approximately ~6.25 for eight full cores. That is 21% more than CMT, which means that CMT scale pretty well. Problem is obviously with IPC or low/inaccurate frequency or something other. Per module with single thread execution, according to VR-zone result 3.1 GHz without C6 has same or little lower performance than K10 core, per clock. To reach hypothetical X8 K10 multithread performance, BD module must have 20-22% better IPC than Stars core.

8120Matrix.png

[chew*]

The "AMD FX CPU (...) | August 31, 2011" document this guy talked about is not the final performance version chew* ?

Not sure what you are referring to.

[informal]

Not sure what you are referring to.

He is referring to reviewers guide given to press,I think.

[maltrabob]

The "AMD FX CPU (...) | August 31, 2011" document this guy talked about is not the final performance version chew* ?

Oh no, Franck7511, another Bulldozer negativist from France. No offence Olivon, really, but what`s going on? Has AMD hung a giant Bulldy billboard over Eiffel tower or what?

[chew*]

He is referring to reviewers guide given to press,I think.

So a reviewers guide meant to accompany B2 F die

[Olivon]

Oh no, Franck7511, another Bulldozer negativist from France. No offence Olivon, really, but what`s going on? Has AMD hung a giant Bulldy billboard over Eiffel tower or what?

That's not my fault

I don't know him except that's more an AMD fan usually. The guy seems disappointed from what he has seen. That's all I know I promess !

And keep cool guyz, there's no "Bulldozer vendetta" from France. That's in your imagination

[El Gappo]

Like i said before the joke is on that guy we won't name, he was sent intentionally quite possilby the worst chip ever produced. 6.4 on ln2,
or he needs to learn how to OC, 6.4 is my validation speed on phase change......he who laughs last laughs best.............

Money is on the later.