ExtremeTech: AMD Bulldozer FX pricing revealed: a lot cheaper than Sandy Bridge

[Arzachel]

Me, I never trust anyone in marketing. Standard policy for any company.

When the alternative is angry internet men, I'll take my chances. There is always some truth beneath the spin, which can't be said about baseless speculation. It's certanly a fun time waster though.

 

[formulav8]

JFAMD also confirmed several times that BD wasn't delayed.

Apparently he's right since server BD is shipping within the 3rd 1/4 time frame. (unless he stated differently?)

He is only in the know on the server side. I think its possible that BD is clock - per -clock faster than Ph2 with server workloads. Desktop workloads may tell a different story.

Server requirements is majorly different than desktop requirements.

Anyways, tomorrow is supposed to be AMD's big announcment. I think it has something to do with a major partnership or something maybe graphics related or something like that. Just a few hours will tell I guess.


Jason

 

[chihlidog]

JF may not have been intentionally misleading, but I think it has been made crystal clear that anything he says about BD needs to be taken with a grain of salt.

 

[jhu]

What is this BD everyone speaks of?

I'm more convinced that Bigfoot exists than I am that Bulldozer exists. I wonder which one we'll have proof of first?

Bulldozer has already shipped.

 

[choliscott]

I think (s)he is talking about the desktop version, not the server version

Bulldozer has already shipped.

 

[BlueBlazer]

@Everyone: How can you compare something thats down to mostly GF's 32nm process with something that has been set in stone from the day the design was finalised?

Blame Hector for going fabless ("fablite" as it is known).

And yes, I trust a representative of the company far more, than people that base their opinions on uncomfirmed leaks. I thought that would seem logical?

It was "unconfirmed" leaks back in May that hint at delays, that AMD (and JF-AMD) denied till the last minute. It was "unconfirmed" leaks that seems to correlate with the pre-order pricing discussed in this thread. Its beginning to look like "unconfirmed" leaks have better information than coming from either AMD or JF-AMD.

Apparently he's right since server BD is shipping within the 3rd 1/4 time frame. (unless he stated differently?)

He did state differently....

Everybody calm down. It is a few weeks difference in schedules. We are shipping server in Q3 like we said.

That schedule delay pushed the launch of client into Q3. Nothing sinister. But keep in mind that launches align around events and partners, not production schedules.

And on the next statement, he had gall the to say....

The rumors of isssues aren't worth the time to address, they aren't true. Too much conspiracy theory going around.

After the the rumors of the delays were proven correct!

When the alternative is angry internet men, I'll take my chances. There is always some truth beneath the spin, which can't be said about baseless speculation. It's certanly a fun time waster though.

What truth? Almost everything he said was disproven (see above)....

 

[LOL_Wut_Axel]

]Why does it need to have IPC of Nehalem? How about 85% of Nehalem's IPC(same core count of course),on average, and 4.2Ghz Turbo on 4 cores? [/B]
Look here. Take as an example i7 950 and i7 960. Use in between overall score for hypothetical 4C Zambezi and apply 0.85x - you get 154pts. 50% more cores on desktop,with intel,brings you 16% more performance in that chart.Hence 100% cores (4C Zambezi => 8C Zambezi) brings you 32% more performance: 154x1.32=204pts in the chart. Now this is for 3.1Ghz/4Ghz. 3.6Ghz/4.2Ghz should therefore score roughly 10% more (accounting for high turbo on 8120) : 204*1.1=224pts. That's a hair slower than 990x and a bit faster than 2600K.
Remember this is possible performance of Zambezi that assumes it is 17% slower than Nehalem at the same clock.

Because 85% of Nehalem's IPC would put it at only around 5% faster than Llano, which is unacceptable. We don't want CPUs that are total dogs in anything that's not highly multi-threaded.

 

[Mopetar]

JF may not have been intentionally misleading, but I think it has been made crystal clear that anything he says about BD needs to be taken with a grain of salt.

A lot of what he's said has come from his articles (or another person from the same/similar department) posted on AMD's website. I believe that the legal department is fairly strict about the type of information that can be posted and that performance claims need to be verifiable. As such, it's doubtful that anything he's posted is an outright lie.

Honestly, I'd take his word on a topic more than most others here. Everyone else is working with a lot of second, third, or fourth hand knowledge from sources of questionable credibility. If you're going to take JF's word with a grain of salt, you'd need at least a barrel for anyone else.

 

[Stoneburner]

I've suffered long enough without upgrading. I need a processor that will shave time off my DT rush in SC2. For too long I have suffered ignominious defeat because my slow DT's run headfirst into cannons, overseers, and alligator filled pits. Somehow alligators can detect cloaked units.

Looks like the 2500k for me.

 

[Idontcare]

A lot of what he's said has come from his articles (or another person from the same/similar department) posted on AMD's website. I believe that the legal department is fairly strict about the type of information that can be posted and that performance claims need to be verifiable. As such, it's doubtful that anything he's posted is an outright lie.

Honestly, I'd take his word on a topic more than most others here. Everyone else is working with a lot of second, third, or fourth hand knowledge from sources of questionable credibility. If you're going to take JF's word with a grain of salt, you'd need at least a barrel for anyone else.

this

 

[LOL_Wut_Axel]

Where in the world did you come up with 8C/8T = 6C/12T? Based on what data? Scaling doesn't change with more threads if the program actually uses 8 threads. An 8 core i5 760 would be 90-100% faster in a program that uses 8 threads instead of 4 vs. a i5 760. If you argue that few programs use 8 threads, then I agree, which is why BD being slower in 1-4 threads is critical.

Simple maths:

Speedup, given perfect scaling:
1 to 2 cores: 100%
2 to 4 cores: 100%
4 to 6 cores: 50%
6 to 8 cores: 25%

If an application scales to almost the max it can, from 4 to 6 cores you'll see a speedup of 45-50%, but going higher than 6 cores has almost always been more detrimental, so a speedup of 20% is more typical. Take into account that HyperThreading adds on average 20% performance in multi-threaded applications, and there you go. Even if we were to take into account perfect scaling, that's 25%, so only a bit more than putting HT on a Six-Core chip.


i7-990X is a not a SB architecture and it only has 6 cores. So how would it be 2x faster than a 4 core SB? Your comparison makes no sense.

How does me comparing using another Nehalem chip, only difference being it has HyperThreading, not make any sense? How does yours make more sense? It makes perfect sense. On a perfectly multi-threaded application a Core i7-990X would be 50% faster than a Core i5 750 at the same clock speed.

^^

Anyway, it's pointless to argue now. Bulldozer is a dud. All I have to wonder now is if I'll go Sandy Bridge or wait for Ivy Bridge and the next-gen GPUs plus PCIe 3.0.

 

[grimpr]

Looks some overclockers from Sweden busted a world overclock record with a processor from the future.

http://www.facebook.com/photo.php?f...1651816575.204516.203438921575&type=1&theater

 

[looncraz]

JFAMD also confirmed several times that BD wasn't delayed.

JF-AMD was referring to the server side, which he stated many many times (he didn't know anything about delays on desktop SKUs - and couldn't say anything if he did).

Server side was scheduled to ship in Q3, and it shipped in Q3.

 

[LOL_Wut_Axel]

Hmm, what about Folding@home? You need 8 threads for -bigadv. At around $200, wouldn't this make a great CPU just for that purpose? I'm sure it'd push all the cores to their max.

 

[looncraz]

It was "unconfirmed" leaks back in May that hint at delays, that AMD (and JF-AMD) denied till the last minute. It was "unconfirmed" leaks that seems to correlate with the pre-order pricing discussed in this thread. Its beginning to look like "unconfirmed" leaks have better information than coming from either AMD or JF-AMD.

He did state differently....And on the next statement, he had gall the to say....After the the rumors of the delays were proven correct!

What truth? Almost everything he said was disproven (see above)....

JF-AMD never said the client version wasn't delayed, he said that server was more or less on time for Q3 - and we was telling the truth.

The rumors to which he was referring had nothing to do with the delay, but the cause: most were saying bulldozer must be buggy, or the clock speeds couldn't be reached - JF-AMD called those hogwash, and eluded that things were going as expected and the delay was for non-technical reasons (though that doesn't mean engineers wouldn't use that opportunity to improve matters more).

He also said the following about Bulldozer's performance as it pertains to server:

50% better throughput :hmm:
35% better performance
Lower than expected module overhead (7-8% vs 10-15%)
Server chips ship in Q3
Doesn't know / can't say anything about client side.:ninja:
IPC increases vs phenom II
Energy efficiency and die savings were primary focus. :'(
Not trying to match Sandy Bridge in IPC. D:

We can't say he lied about anything...

--The loon

 

[Jdmathew]

I only see good news. When did I become accustom to paying 3-400 for a home PC processor. Personally Im rooting for AMD to compete just to drive down Intel pricess, which I actually buy at this point.

 

[sequoia464]

What is this BD everyone speaks of?

I'm more convinced that Bigfoot exists than I am that Bulldozer exists. I wonder which one we'll have proof of first?

I had tequila shooters with Bigfoot one night - pretty sure he exists.

 

[gmaster456]

Looks interesting. And $155 isn't bad at all. If it outperforms the i5-2400 I'll be happy. If not, I'll probably just go with Sandy Bridge. I've been waiting long enough and am not all that interested in waiting AGAIN for revision 2 next year.

 

[ElFenix]

Looks some overclockers from Sweden busted a world overclock record with a processor from the future.

http://www.facebook.com/photo.php?f...1651816575.204516.203438921575&type=1&theater

holy crap 8 mhz!

 

[Riek]

Because 85% of Nehalem's IPC would put it at only around 5% faster than Llano, which is unacceptable. We don't want CPUs that are total dogs in anything that's not highly multi-threaded.

and where would a 4GHz llano +5% performance end up you think? with the atoms?

 

[eternalone]

holy crap 8 mhz!

I love this because no one still knows how bulldozer performs. This is literally making people crazy on tech forums. LOL

 

[lol123]

Looks some overclockers from Sweden busted a world overclock record with a processor from the future.

http://www.facebook.com/photo.php?f...1651816575.204516.203438921575&type=1&theater

I don't understand why people go on to assume that this will be on a new processor. The current world record is on a Celeron or a Pentium 4 if I'm not mistaken and it's unlikely that the new world record will be on either a Sandy Bridge or a Bulldozer. With all those cores they simply won't overclock to the same frequencies as the old single core CPUs.

Edit: My bad, it actually says "with a CPU from the upcoming...". Now this is really exciting.

 

[jhu]

holy crap 8 mhz!

Wow! It's almost twice as fast as my IBM PC XT!

 

[notty22]

Here are some cpu-z overclock records.
http://valid.canardpc.com/records.php
I don't think anyone called JF a liar. No more than people are putting words in his mouth concerning details, desktop offerings.

 

[ransomlist]

I don't understand why people go on to assume that this will be on a new processor. The current world record is on a Celeron or a Pentium 4 if I'm not mistaken and it's unlikely that the new world record will be on either a Sandy Bridge or a Bulldozer. With all those cores they simply won't overclock to the same frequencies as the old single core CPUs.

Edit: My bad, it actually says "with a CPU from the upcoming...". Now this is really exciting.

I want more details!

I wonder if its possible to temporarily disable bulldozer modules on command for an 8 'core' model and emulate the fx 4-xxxx for example (with higher clocks possible as a consequence).

Bah who are we kidding, it must be SB-E:'(