Bulldozer may not provide dramatic performance increase

[khon]

But Bulldozer has to come up roses for AMD. They can't afford for this cpu to be mediocre.

They'll hang in there somehow.

We all need Bulldozer to be good though, because without AMD there to push them, Intel will have no insentive to develop their own products as fast as they might, which means we all get slower computers.

 

[IntelUser2000]

Xbit is a solid site. One of the better ones.

Xbitlabs is a excellent review site. Not so much on the news side lately.

Computer Bottleneck: I think it has 2MB L2 shared per module. Don't know about L3 cache sizes.

 

[Tsavo]

Frankly, at this point, I'd be happy with a melodramatic increase in performance, just so long as it isn't Scarlett and Rhett not giving a damn about each other over the soundtrack to Days of Our Lives.

 

[Schmide]

Isn't all this kind of a misnomer?

I mean when they're talking a 16 core package. It's actually 8 modules consisting of 16 integer units, 8 floating point units, 8 fetches, 8 L2 caches, etc.

Not to mention this is probably a dual package of 4 modules each connected by a HT link.

So there is really no apples to apples comparison.

 

[JFAMD]

That was a very misleading article. I posted this on another site:

Oh, and just so everyone is clear, Hans de Vries pointed out a very interesting tidbit in comparing the Nehalem to Westmere transition.

Nehalem: Int 267, FP 207
Westmere: Int 382, FP 257

So that means 43% better integer, 24% better FP. If you average the two together, what do you get? 33.5% better performance with 50% more cores.

In a single generational change, Intel gives you 50% more cores with 33.5% greater performance. AMD gives you 33% more cores with 50% better performance.

If our performance gain is not dramatic, then what is theirs?

It was clear that posting an article saying "performance won't be dramatic" when in reality a 50% performance increase (after this year's 80-120% increase) is really missing the mark.

 

[zsdersw]

A really lackluster predecessor usually does make what comes next look, when compared to each other, like a big improvement.

That's good for marketing departments (of which you're an obvious part, JFAMD), but what matters more to the rest of us is how it stacks up to the competitor(s) it will be facing.

 

[IntelUser2000]

There's simpler way to explain all this. Theoretical boost with 2x cores = 2x. Theoretical boost for the new multi-threading tech on BD is 1.8x. 33% more cores = 33 x 0.8 ~ 25% more performance for 33% more cores. Does a 50% boost for 33% more "cores" sound better now?

 

[Idontcare]

I can't be the only one who is thinking to themselves "who cares?".

Can I buy BD today? No. So whether the performance is 10% better than today's product or 500% better it is pretty much irrelevant unless I am an stock investor/speculator or if I happen to be in the position of deferring IT capital expenditures for another year on the hopes of getting ever more bang for my budget buck come next year.

But all this hoopla over rumors and speculations of performance being xyz and then dissecting the grammatical nuances of those rumors to the nth degree just seems like a colossal waste of time and effort since the nut of it you can't buy it regardless the performance.

 

[Edrick]

I do not know what the big deal is.

According to what I have read, Core for Core, SB will beat BD. That is even IF BD makes it out before Ivy Bridge is released.

Now if you take a 16 core BD and compare it to a 8 core SB, then sure, the BD chip will outperform the SB chip IF all cores are taken into account. (Again, according to speculatiuon)

Now the real question is what application(s) on the market today is going to saturate 16 cores? (Other than benchmarks and folding apps). Games sure as hell do not. Most productivity apps do not. And the few apps that do, also are making the move to GPU tech, so having a GTX480 will still boost performance even better.

Personally, I would rather have 8 higher performing cores than 16 lesser performing cores. At least until software catches up to the hardware. I am not even taking into account the wattage/heat difference.

Everyone thinks BD is going to be the next big thing and de-throne Intel. I do not think that is the case at all except on a few benchmark graphs.

 

[jvroig]

Now the real question is what application(s) on the market today is going to saturate 16 cores?

Personally, I would rather have 8 higher performing cores than 16 lesser performing cores.

You'll actually get what you are wishing for. For desktop variants, Bulldozer only has up to 8 cores. The 16-core variants are for servers, so all those 16 cores would not go to waste (and certainly will not go to your desktop).

As for performance, who knows, too early. But at least right now you already know you won't get 16 cores in your desktop, just 8, like half of what you wished for (hopefully, the other half, "high performance", also gets granted, we'll see at the end of 2011)

Everyone thinks BD is going to be the next big thing and de-throne Intel.

I don't (and you don't either), and a lot of Intel users don't think so either. A lot of sensible persons don't think so, too. After all, we don't even have any data at all about performance. Those making too much noise about it now are just doing their thing "rooting for the underdog", a common behavior. I personally find those posts entertaining to read, for one thing. And it makes the waiting all the more exciting.

 

[Martimus]

Core for Core, SB will beat BD. That is even IF BD makes it out before Ivy Bridge is released.

Why make a statement about two unreleased products that you know nothing about, declaring how they will perform? If the purpose is to make people doubt everything you write after that statement, then it will likely succeed.

In the future, if you want any credibility in your musings, you probably should avoid making declarative statements about things that you cannot know, or at the very least wait until the end of your argument so that people have a chance to digest what you are trying to get across before completely writing off your opinion.

Everyone thinks BD is going to be the next big thing and de-throne Intel. I do not think that is the case at all except on a few benchmark graphs.

Who is this "everyone" you speak of? I don't know anyone who believes that. From what little I have seen, it looks more like BD is designed more to run cool and efficient, and be cheap to manufacture than it is to have huge performance increases. Then again, I am no expert on the BD architecture, and my opinion is no better than the next guy.

 

[aldamon]

I might have my Q6600 forever.

 

[bryanW1995]

I do not know what the big deal is.

Core for Core, SB will beat BD. That is even IF BD makes it out before Ivy Bridge is released.

Now if you take a 16 core BD and compare it to a 8 core SB, then sure, the BD chip will outperform the SB chip IF all cores are taken into account.

Now the real question is what application(s) on the market today is going to saturate 16 cores? (Other than benchmarks and folding apps). Games sure as hell do not. Most productivity apps do not. And the few apps that do, also are making the move to GPU tech, so having a GTX480 will still boost performance even better.

Personally, I would rather have 8 higher performing cores than 16 lesser performing cores. At least until software catches up to the hardware. I am not even taking into account the wattage/heat difference.

Everyone thinks BD is going to be the next big thing and de-throne Intel. I do not think that is the case at all except on a few benchmark graphs.

keep in mind that BD is really going to be 8 cores + 8 half cores. sb will be 4-8 cores with another 4-8 ht cores. if bd is the same size as an 8 core sb and has better performance, then who cares how they do it?

 

[OCGuy]

AMD's six core gets smoked by current Intel quad-cores. Why on earth would anyone assume that the new software on the market in the next 12 months would make 8 core processors worth anything to the average user buying a Dell or Best Buy Special?

Adding more cores is not what is needed at this point. We need faster clock-for-clock, from both companies.

 

[Edrick]

Why make a statement about two unreleased products that you know nothing about, declaring how they will perform? If the purpose is to make people doubt everything you write after that statement, then it will likely succeed.

Every new thread on this forum based on unreleased products (and there are many) all are full of posts by many people who make statements based on rumors or unconfirmed reports. Half the reason I come here is to speculate and hear other people's opinions on these topics.

So what you are saying is that every single word ever written on a product pre-release is not even worth reading. Well, thats your opinion and you are entitled. And I am quite certain that not once in your 2,000+ posts here that you ever made a statement on an un-released product.

So for everyone on this forum who enjoys this type of discussion, we apologize to you high and mighty forum master of truth!

PS. How do you know there is no way I could or could not know certain things? Just saying I could work for AMD for all you know.

 

[IntelUser2000]

AMD's execution is more important than technical features. If they don't pull a FX5800/Prescott/Fermi they'll be fine.

So for everyone on this forum who enjoys this type of discussion, we apologize to you high and mighty forum master of truth!

Difference is Edric, we are really speculating while you are making claims.

 

[Edrick]

AMD's six core gets smoked by current Intel quad-cores. Why on earth would anyone assume that the new software on the market in the next 12 months would make 8 core processors worth anything to the average user buying a Dell or Best Buy Special?

Adding more cores is not what is needed at this point. We need faster clock-for-clock, from both companies.

That is what I was trying to say in a more direct to the point way. 😀

 

[Edrick]

Difference is Edric, we are really speculating while you are making claims.

Not my intention to make claims. I am just basing what I say off everything I have read on the subject (including these forums.) Perhaps if I said "According to what I have read" at the beginning, it would not come across that way. Noted and will make an effort to make my speculations seem less like claims.

 

[endervalentine]

I do not know what the big deal is.

According to what I have read, Core for Core, SB will beat BD. That is even IF BD makes it out before Ivy Bridge is released.

Now if you take a 16 core BD and compare it to a 8 core SB, then sure, the BD chip will outperform the SB chip IF all cores are taken into account. (Again, according to speculatiuon)

Now the real question is what application(s) on the market today is going to saturate 16 cores? (Other than benchmarks and folding apps). Games sure as hell do not. Most productivity apps do not. And the few apps that do, also are making the move to GPU tech, so having a GTX480 will still boost performance even better.

Personally, I would rather have 8 higher performing cores than 16 lesser performing cores. At least until software catches up to the hardware. I am not even taking into account the wattage/heat difference.

Everyone thinks BD is going to be the next big thing and de-throne Intel. I do not think that is the case at all except on a few benchmark graphs.

To play devils advocate for a bit ... is there any software or applications out there for the regular joe that will require 8 cores? I'm all pushing the technology envelope, but if either company want to sell a lot of their silicon, wouldn't they tailor their chips to the regular joe versus trying to out best each other and get the performance crown?

Or is the the goal not the regular joe, but large corporations, where power and performance does matter when they do their pitch? If it's this latter case, then, I guess it does matter ...

 

[IntelUser2000]

That's why I keep saying the Socket 1366/2011 offering 6 and 8 cores are not important. The benefits come from a new architecture. If 6 cores are brought to the mainstream socket it might be little better.

The larger shared caches offered by more core CPUs are more relevant to most than the actual number of cores. Of course even that has detrimental effects like higher latency as well.

Same is true for AMD. But when they say they have 8 cores, they also mean there are 8 threads. Hyperthreading is one thing that makes 4+ core CPU less important.

 

[Martimus]

So for everyone on this forum who enjoys this type of discussion, we apologize to you high and mighty forum master of truth!

I apologize for upsetting you with my post. I should have worded it more delicately, or not written it at all. You just happened to catch me at a bad time while doing one of my pet peeves: stating an opinion as fact, without any backup data to prove your point.

 

[Edrick]

I apologize for upsetting you with my post. I should have worded it more delicately, or not written it at all. You just happened to catch me at a bad time while doing one of my pet peeves: stating an opinion as fact, without any backup data to prove your point.

Thank you for the apology. And I too apologize for getting a little "snippy" in my reply. You were right however, my post was coming across as claims when I did not intend it to. I edited my post to show that it was speculation. I guess if your post was a little nicer and constructive, I would have been much more receptive to it. I will admit my mistakes, and I have no issue people pointing them out to me (in a nice way).

 

[Mr. Pedantic]

But, if clocks remain about the same as for the current Thubans, then a 13% increase in speed per core should put clock-for-clock performance about par with Nehalem, shouldn't it?

 

[bryanW1995]

AMD's execution is more important than technical features. If they don't pull a FX5800/Prescott/Fermi they'll be fine.



Difference is Edric, we are really speculating while you are making claims.

how could you leave out R600? that was worse than any of the others!

edit: @ mr pedantic: keep in mind that if BD is on clock/clock parity with nehalem that is a win for amd because 1/2 their cores or only, um, 1/2 cores. I think I might be bugging JF for some of his server 16 core parts...

 

[VirtualLarry]

Now the real question is what application(s) on the market today is going to saturate 16 cores? (Other than benchmarks and folding apps). Games sure as hell do not. Most productivity apps do not. And the few apps that do, also are making the move to GPU tech, so having a GTX480 will still boost performance even better.

Personally, I would rather have 8 higher performing cores than 16 lesser performing cores. At least until software catches up to the hardware. I am not even taking into account the wattage/heat difference.

Everyone thinks BD is going to be the next big thing and de-throne Intel. I do not think that is the case at all except on a few benchmark graphs.

Think server virtualization. More cores = better virtualization. At least, as long as your not saturating the I/O. But even that has CPU overhead in a VM.