Bulldozer vs SandyBridge

[Cs342]

I'm planning on building a new system in January, but I can't decide whether I should build an AMD or Intel system. Which of these processors do you think will offer the best value for my money and why? Also what are the release dates for these 2 new processor lines, and should I just build a 1366 or AM3 system instead of waiting??

 

[halfcalf]

Now what is hilarious is reading some of the other enthusiast site forums that are linked in this thread. One guy is posting that he is absolutely sure that Bulldozer is going to bulldoze Sandy Bridge, given the name "Bulldozer", and then in the next sentence admits he knows nothing about Sandy Bridge..

Well, that's pretty obvious... Bulldozer is no match for Sandy Bridge. Let's take the Cat D11T for example. At an operating weight of 230581 lbs., there is no way that any Bridge built on a Sandy basis is going to be able to support the weight of this Bulldozer!

 

[keithlm]

Bulldozer will be based on 32 nm fabrication process.While ivy bridge will be based on 22nm fabrication process.That difference is enough for any one to say that ivy bridge will destroy bulldozer.Bulldozer will compete with sandybridge.

Anyone that comes to that conclusion without any available evidence or benchmarks is clueless.

 

[halfcalf]

Anyone that comes to that conclusion without any available evidence or benchmarks is clueless.

Yes, but definitely NOT dense.

 

[Cs342]

So if I want to build a gaming rig, will Socket 1155 be a good choice or will I have to wait for IVy Bridge? I've heard rumors that Ivy Bridge CPUs will all be Xeons but I'm not sure if that's true. And if 115/1365 are anything like 1156/1366 then 1365 will be much better than 1155, although it might not be worth the wait. The Bulldozer CPUs should launch around the same time as Sandy Bridge, so I have no idea why AMD hasn't released a single benchmark yet.

 

[jf-amd]

1. AMD does not have a "new" PR policy. It is standard corporate policy not to release benchmarks until launch. Nothing has changed. We also won't comment on how we would perform relative to any unanounced competitive product, that is just crazy because no official benchmarks exist.

2. 32nm. Any 28nm is a crap rumor. Period.

3. SB launches in Q1. Bulldozer (client) launches in Q2. To say that Bulldozer will launch closer to ivy bridge says that someone really doesn't understand either company's schedules.

IF, for some crazy reason, intel is going to launch SB and IB within a 3 quarter span, it would say that intel was really concerned that BD was going to be a lot faster. It is extremely expensive to move to new processes, you don't rush that unless you are really under the gun. If you believe SB is faster, then intel would have no reason to spend the extra money to pull IB in. Take your pick on that one.

4. If you are going to bring up Randy and Barcelona, then you might was well bring up intel's claim that the Pentium 4 will hit 10GHz. Remember that Randy was in the server business (he was my boss) and he would not be talking about C2D ever because he was not running client.

 

[halfcalf]

I'm always pleased to see Mr. Fruehe comment as it really shows that he is listening to what the enthusiasts are saying, and that is an overwhelming positive. Therefore thank you.

In the discussion of where Bulldozer fits in the overall scheme of CPU performance, though, I believe that Mr. Fruehe has given us a very oblique indication in this comment (as it is naturally all he can say at this time).

His comments discussing Bulldozer versus Sandy/Ivy Bridges "seem" to indicate that Bulldozer is not designed to be just another midrange also-ran, but is aiming for the top of the heap. That is certainly gratifying and definitely calls for a thumbs up!

2011 is going to be a great CPU year! Let's toast true competition as it makes everything better!

 

[cjl]

You donot need rocket science knowledge to come to that conclusion.Anyone with basic chip knowledge will know a series with smaller die will always beat a series with larger die.

Always?

Why does an Athlon 64 beat an Atom then? Granted that's an extreme example, but it does show my point - chip speed is a combination of process and architecture, and process alone cannot tell you which chip will be the fastest.

 

[harna]

SO WHAT SHOULD I DO?!!!!! lol

...may I presume that if you can wait...that maybe you don't need to upgrade.

...As for me. I cycle between mobo/cpu/RAM and the GPU...12 months a piece. I find that 2 years gives enough time for significant development in architecture to justify an upgrade. By then the products have been tested, matured and most bugs ironed out....that's unless one likes to iron out bugs of course....

I like to gain extra features with an upgrade...not just speed..so for me at this point...affordable hexacore, eyefinity, fusion, unlocked multies, upgrade portability and software support of vendor hardware give AMD "the nod," in my book.

....Whether SB can outrun BD on some/any or all benchmarks whilst supported by an AMD/Nvidia product is of no real significance....

..but as an active participant in the ICT upgrading stakes, it does appear that AMD is continuing to successfully broaden it's user appeal and product base....more important than BD will be the reception of Bob Cat into the lappy market...and by the look of the new dual graphic support laptops being brought to the table by the big manufacturers, it does look to me like the Bobby is gonna keep the Itanic Warehouse over-stocked.

 

[jf-amd]

Ok Jf amd how come tons of benchmark started leaking before the launch of ati 6000 series?

Well, I am not in charge of GPU so I have no idea. And, when you say leaking, it probably has to do with the web site releasing data. There is a difference between leaking and someone breaking NDA. How soon before launch did reviews start popping up?

 

[Lian]

I kept my eyes on the 6xxx benchmarks, popped up a few of weeks before launch.

And as far as i know it didnt "leak", it was released by the manufacturers... Same as when the manufacturers received the 1120 shader 6850. Look for those benchies and you will find the "more or less" date to when the figures where released.

 

[halfcalf]

Well, I am not in charge of GPU so I have no idea. And, when you say leaking, it probably has to do with the web site releasing data. There is a difference between leaking and someone breaking NDA. How soon before launch did reviews start popping up?

The current Bulldozer vs. Sandy Bridge benchmarks secrecy is a bit of a WikiLeaks situation. Country A says that Country B must never be told of weapon XYZ in a leaked cable. Country B publicly reacts with shock that Country A has such a weapon. The reality is that both countries already know everything about each other's weapons programs right down to the brand of coffee in the weapon lab cafeteria. Both Intel and AMD are multizillion dollar companies. They are both very well informed about each other's development programs. If they were as in the dark as the public, then heads should roll in the boardroom. However, it's a game that they have to play publicly. That's another reason why I respect and thank Mr. Fruehe for continuing to hang out here!

 

[fazers_on_stun]

1. AMD does not have a "new" PR policy. It is standard corporate policy not to release benchmarks until launch. Nothing has changed. We also won't comment on how we would perform relative to any unanounced competitive product, that is just crazy because no official benchmarks exist.

2. 32nm. Any 28nm is a crap rumor. Period.

3. SB launches in Q1. Bulldozer (client) launches in Q2. To say that Bulldozer will launch closer to ivy bridge says that someone really doesn't understand either company's schedules.

IF, for some crazy reason, intel is going to launch SB and IB within a 3 quarter span, it would say that intel was really concerned that BD was going to be a lot faster. It is extremely expensive to move to new processes, you don't rush that unless you are really under the gun. If you believe SB is faster, then intel would have no reason to spend the extra money to pull IB in. Take your pick on that one.

4. If you are going to bring up Randy and Barcelona, then you might was well bring up intel's claim that the Pentium 4 will hit 10GHz. Remember that Randy was in the server business (he was my boss) and he would not be talking about C2D ever because he was not running client.

Hmm, my bad - from the youtube video, he was apparently talking about the server version of Core2 (which is what I said - not "C2D"), aka Clovertown:

[flash=480,385]http://www.youtube.com/v/G_n3wvsfq4Y?fs=1&hl=en_US&rel=0[/flash]

Re the Intel claim, IIRC a P4 did hit over 8GHz on LN2 - holds the world's record for overclocking I think. So Intel was ~80% correct in their prediction. I don't want to do the math on Mr. Allen's claim but I'm willing to bet it is considerably lower than 80% correct, in desktop apps anyway .

 

[halfcalf]

This reminds me of a policy dictated to me by the CEO of a fairly large corp I worked for in the 80s: Keep Your Big Fat Mouth Shut Until It Ships! He also liked to say Speculation Is Ma$turbati0n!

 

[jf-amd]

Hmm, my bad - from the youtube video, he was apparently talking about the server version of Core2 (which is what I said - not "C2D"), aka Clovertown:

[flash=480,385]http://www.youtube.com/v/G_n3wvsfq4Y?fs=1&hl=en_US&rel=0[/flash]

Re the Intel claim, IIRC a P4 did hit over 8GHz on LN2 - holds the world's record for overclocking I think. So Intel was ~80% correct in their prediction. I don't want to do the math on Mr. Allen's claim but I'm willing to bet it is considerably lower than 80% correct, in desktop apps anyway .

we beat clovertown by more than 40% in stream, so technically you could claim that he was right.

And Intel's specs don't count overclock and the intel claim was base clock speed, not overclock, so you can assume that they are only ~30-40% right. So it can go either way.

This is precisely why it is not a good idea for either side to talk about benchmarks prior to the product being out. Everyone is crying that they want benchmarks NOW, but the reality is that this is what you end up with when you ask people to make statements prior to final shipping silicon.

 

[halfcalf]

Well, my Xmas wishes for AMD is that Bulldozer is a technological marvel and an outperformer extraordinaire! That way everybody wins, especially the enthusiasts!

 

[keithlm]

You donot need rocket science knowledge to come to that conclusion.Anyone with basic chip knowledge will know a series with smaller die will always beat a series with larger die.

Anyone with any intelligence would be smart enough to not make blanket claims concerning new architectures when there is NO DATA available.

Your comment about "a series with a smaller die" would be completely true IF you were comparing two CPU that have identitical architectures. However, it would be very stupid to make the same claim when the architectures are distinctly different AND there is no available data.

But you just keep right on doing that... we'll bookmark this thread.

 

[jf-amd]

Blanket statements are never correct.

Wait, that is a blanket statment. Doh!

 

[Cs342]

My birthday is in March, so if both Bulldozer and Sandy Bridge are released before then then I should have a pretty good idea of what to expect.... although I've always favored AMD because their products are much more reasonably priced than Intel's.

 

[halfcalf]

jf-amd: Prince Michael Jackson II makes Blanket statements all the time!

Cs342: You'll have Sandy LGA1155 by then but that's about it.

 

[fazers_on_stun]

we beat clovertown by more than 40% in stream, so technically you could claim that he was right.

Heh, not to flagellate a deceased equine too much, but about 10 seconds into the video Mr. Allen claimed "many" workloads, and elsewhere a "wide variety of workloads".

I dunno about other enthusiasts sites, but 3 years ago the impression in many threads here was that Barcie was a dud.. And especially so for the gamers here, although they knew that Barcie was designed as a server CPU.

AMD's efforts to rewrite Barcelona history remind me of this plucky lad:

[flash=480,385]http://www.youtube.com/v/cufQD5Y31ZA?fs=1&hl=en_US&rel=0[/flash]

And Intel's specs don't count overclock and the intel claim was base clock speed, not overclock, so you can assume that they are only ~30-40% right. So it can go either way.

I'd like to see a link or quote of Intel's claim, if you have one handy. IIRC, the first gen of P4 was on a 180nm process - maybe they could have reached 10GHz stock using their present 32nm..

This is precisely why it is not a good idea for either side to talk about benchmarks prior to the product being out. Everyone is crying that they want benchmarks NOW, but the reality is that this is what you end up with when you ask people to make statements prior to final shipping silicon.

That I agree with, since it's what I've been saying for 3-4 months now - wait for the benchies .

 

[jf-amd]

I'd like to see a link or quote of Intel's claim, if you have one handy. IIRC, the first gen of P4 was on a 180nm process - maybe they could have reached 10GHz stock using their present 32nm..

http://www.design-reuse.com/news/4850/intel-building-blocks-10-ghz-processors.html

The original statement was made in 2000 IIRC, this was a follow-up at ISSC basically saying the building blocks were there to have 10GHz by the end of the decade (Dec 31, 2009).

If you just google "Intel 10GHz" you can find enough references to know what was said. I used to have the interview with the guy that made the statements, but it has been so long that I can no longer find it.

 

[halfcalf]

Sun is now saying that they're going for 5 GHz stock CPUs in the SPARC. Is thermal the only reason keeping AMD and Intel from shipping silicon at these frequencies?

 

[jf-amd]

RISC is a different architecture and Sun is in a different business model. For instance, what if you could get there with a $1000 cooler. On a $100K system that is not a big impact.

Thermal is the biggest reason for today's clock speeds. Look at what you have to do to get an x86 chip there today.

 

[halfcalf]

Actually I know a guy who spent well over a grand cooling his system... and it's just a Q9400! However, I can see what you mean about the RISC. It just seems disappointing that the max frequency on the highest announced Sandys is the same as it was on the old and unlamented Prescott (talking about blast furnaces...)

 

[fazers_on_stun]

http://www.design-reuse.com/news/4850/intel-building-blocks-10-ghz-processors.html

The original statement was made in 2000 IIRC, this was a follow-up at ISSC basically saying the building blocks were there to have 10GHz by the end of the decade (Dec 31, 2009).

If you just google "Intel 10GHz" you can find enough references to know what was said. I used to have the interview with the guy that made the statements, but it has been so long that I can no longer find it.

OK, thanks. Guess it goes to show that it's pretty hard to predict the future accurately. Take Jen-Sun Huang's "can o' whoop-ass" statement a couple years ago .