Bulldozer screens/info (pcinlife)

[Tuna-Fish]

If these are samples, they are TESTING a range of frequencies to see if they are viable for retail.

Umm, no. They are model numbers for ES processors. The Hammer ES was 800MHz -- you can infer absolutely no useful knowledge of the retail clocks from the ES clockspeeds.

 

[Soleron]

Umm, no. They are model numbers for ES processors. The Hammer ES was 800MHz -- you can infer absolutely no useful knowledge of the retail clocks from the ES clockspeeds.

I know they are ESs. Early samples are far below retail clocks (like the 1.6GHz Barcelona sample). But when they get near retail, they put out a lot of samples at various speeds to see if they hold up on lifespan and power consumption. Bulldozer is in that phase. They wouldn't release more than ten finely grained SKUs and then release something slower than all of them unless something was very wrong.

 

[itsmydamnation]

and if you believe the doco there has been atleast one more silicon revision as well (its A1, there is B0).

 

[drizek]

I know they are ESs. Early samples are far below retail clocks (like the 1.6GHz Barcelona sample). But when they get near retail, they put out a lot of samples at various speeds to see if they hold up on lifespan and power consumption. Bulldozer is in that phase. They wouldn't release more than ten finely grained SKUs and then release something slower than all of them unless something was very wrong.

There were 3ghz barcelona samples at one point, right?

Also, we think that there will only be a max of 8 SKUs, right?

 

[MarkLuvsCS]

There were 3ghz barcelona samples at one point, right?

Also, we think that there will only be a max of 8 SKUs, right?

I kind of hope there is at least 3 of each 4,6,8 core CPUs. 4 cores could start 100-150, 6 cores 150-250, 8 cores 250-400. I think having a good variety gives consumers a lot of choices at a price point they want to reach. Although 8 isn't at far off the 9 I was thinking of, I guess it's not a big deal lol.

 

[nonameo]

I kind of hope there is at least 3 of each 4,6,8 core CPUs. 4 cores could start 100-150, 6 cores 150-250, 8 cores 250-400. I think having a good variety gives consumers a lot of choices at a price point they want to reach. Although 8 isn't at far off the 9 I was thinking of, I guess it's not a big deal lol.

Well, based on the speculation and AMD powerpoints(that I have seen on this forum) I have seen, I'd think that the best bulldozer should average out to be about the same as a 2600k or just a little better, and I don't think that AMD is going to try to charge a premium over intel's offerings, as they need some sort of advantage to lure people to their side. That said, I also don't think that AMD will heavily undercut the 2600k because well, 300$ for pretty much the top performing desktop CPU out there is nothing short of a steal. There just isn't a lot of room pricing, really.

So, at best I think we're looking at 200$ of flexibility here, and I don't honestly think that AMD will price any bulldozer core at 100$... That's phenom/athlon territory.

 

[piesquared]

Llano's CPU is die shrunk Athlon II(No L3).. Trinity is their next generation of APU.. based on BD cores.. instead of STARS.

Llano's cores codename are Husky.

 

[Topweasel]

Llano's cores codename are Husky.

Also known as die shrunk Stars, the ones used with the Athlon II's (because of the lack of L3 cache).

 

[piesquared]

Well, based on the speculation and AMD powerpoints(that I have seen on this forum) I have seen, I'd think that the best bulldozer should average out to be about the same as a 2600k or just a little better, and I don't think that AMD is going to try to charge a premium over intel's offerings, as they need some sort of advantage to lure people to their side. That said, I also don't think that AMD will heavily undercut the 2600k because well, 300$ for pretty much the top performing desktop CPU out there is nothing short of a steal. There just isn't a lot of room pricing, really.

So, at best I think we're looking at 200$ of flexibility here, and I don't honestly think that AMD will price any bulldozer core at 100$... That's phenom/athlon territory.

But, AMD has at least planned for a %100 32nm node transision by this time next year or ealier so that is probably Llano territory.

 

[piesquared]

Also known as die shrunk Stars, the ones used with the Athlon II's (because of the lack of L3 cache).

Yeah, so essentially a decendent of the Hounds core. I bet they will be impressive little things.

 

[nonameo]

But, AMD has at least planned for a %100 32nm node transision by this time next year or ealier so that is probably Llano territory.

Ah, good point. Although, maybe you could argue llano is phenom/athlon as it has k10 blood

 

[PreferLinux]

If these are samples, they are TESTING a range of frequencies to see if they are viable for retail. Likely they won't choose the top one listed (4.1GHz) as the actual top bin. I'm expecting about the middle of that list.

Still great for a new architecture on a new process.

Now we know Intel's cheapest 6-core SB is $500+, AMD could easily price an 8-core, 3.5GHz, greater-IPC-than-Phenom-II Bulldozer CPU at $400 and make a huge profit. It just has to be faster than quad-core SB, not hard.

From what? A leaked roadmap that they could change anytime they want?

Those are not consumer product names. those are the names for the engineering samples.

True. So why would they bother posting them then?

 

[Topweasel]

Well, based on the speculation and AMD powerpoints(that I have seen on this forum) I have seen, I'd think that the best bulldozer should average out to be about the same as a 2600k or just a little better, and I don't think that AMD is going to try to charge a premium over intel's offerings, as they need some sort of advantage to lure people to their side. That said, I also don't think that AMD will heavily undercut the 2600k because well, 300$ for pretty much the top performing desktop CPU out there is nothing short of a steal. There just isn't a lot of room pricing, really.

So, at best I think we're looking at 200$ of flexibility here, and I don't honestly think that AMD will price any bulldozer core at 100$... That's phenom/athlon territory.

Keep in mind that Zambezi (BD on Am3+) will be available in 3 Flavors, FX-8K, FX-6K, and FX-4K. What i think is we will see an intermixing, but it will be FX-4k $100-$200, FX-6K $150-$300, FX-8K $250-$400.

 

[drizek]

True. So why would they bother posting them then?

So we can know the clocks/caches and core counts in future leaks from ES.

 

[HW2050Plus]

Frequencies are pretty low on these engineering samples, but that is possibly because of usual ramping issues. The processors which emerge in June are currently in production (apparently B0 or B1) and it is unknown in what speed bins those can be divided.

The information gain from that is very little, so let's hope that there will be some more serious benchmark leaks as there should be many ES flying around at this time.

 

[flexcore]

I am hoping tis is the beginning of BD performance leaks. I actually think these are doing well for preproduction ES's.

 

[Soleron]

Frequencies are pretty low on these engineering samples, but that is possibly because of usual ramping issues. The processors which emerge in June are currently in production (apparently B0 or B1) and it is unknown in what speed bins those can be divided.

The information gain from that is very little, so let's hope that there will be some more serious benchmark leaks as there should be many ES flying around at this time.

Low. Seriously?

 

[Topweasel]

Low. Seriously?

I would recommend ignoring types of comments that make no sense. I don't know if some people believe that Engineering samples run higher then one sold, most of the time its the opposite and the stuff used for testing and parts building tend to be binned low to give manufacturers head room. Apparently someone forgot the lead up to the original Opteron release.

That or maybe he has taken it into account that these are either at or below shipping speeds and thinks its not fast enough. I am pretty sure this is the same poster that thought that after crunching his numbers insistently that AMD made a CPU that's cores were larger, with more transistors, and yet slower then Stars. Maybe with that CPU design theory it would require a core running at or near 5GHz at default and 6GHz turbo to be competitive.

 

[StrangerGuy]

I would recommend ignoring types of comments that make no sense. I don't know if some people believe that Engineering samples run higher then one sold, most of the time its the opposite and the stuff used for testing and parts building tend to be binned low to give manufacturers head room. Apparently someone forgot the lead up to the original Opteron release.

That or maybe he has taken it into account that these are either at or below shipping speeds and thinks its not fast enough. I am pretty sure this is the same poster that thought that after crunching his numbers insistently that AMD made a CPU that's cores were larger, with more transistors, and yet slower then Stars. Maybe with that CPU design theory it would require a core running at or near 5GHz at default and 6GHz turbo to be competitive.

I'm not sure how can some people can even claim BD has lower IPC than Phenom II when every successive architecture from AMD has higher IPC since K6.

 

[SickBeast]

I'm not sure how can some people can even claim BD has lower IPC than Phenom II when every successive architecture from AMD has higher IPC since K6.

It depends on how you look at it. Seeing as the FPU units are shared between two "cores", it's possible that on a per-clock and per-core basis, Bulldozer will be weaker than Phenom II in terms of raw FPU power. That said, AMD has probably made the shared FPUs extremely powerful.

 

[Idontcare]

I'm not sure how can some people can even claim BD has lower IPC than Phenom II when every successive architecture from AMD has higher IPC since K6.

Even if the opposite were true, that is to say that even if AMD had always decreased IPC instead of increasing it, the fact that the one person who has officially talked a lot about Bulldozer (John Fruehe) has said time and again that the IPC of bulldozer has gone up is all we need to know until proven otherwise.

I have no reason to doubt John's assertions. He has credibility, no history of leading people astray or misleading them at all.

 

[podspi]

It depends on how you look at it. Seeing as the FPU units are shared between two "cores", it's possible that on a per-clock and per-core basis, Bulldozer will be weaker than Phenom II in terms of raw FPU power. That said, AMD has probably made the shared FPUs extremely powerful.

As I understand it, the FlexFP is actually stupid clever. I'm really excited to see it in action, hoping it is as good as it sounds

 

[drizek]

Does "IPC" take core count into consideration?

I think you can make a reasonable argument that an Athlon III X12 based on Stars would be superior to an 8-core bulldozer in the same die space for certain applications

Of course, I have no idea how big an Athlon III X12 would be and what kind of clocks it would run at, but it is interesting to think about.