2 different cores for Sempron?

[White Widow]

I think I missed some critical part of Sempron press release, but as I understand it, the Sempron will be produced for both Socket-A and S754. This must mean that one will have an on-die controller and one will not. Is that the only difference between the two? Don't S-A and S754 use completely different busses as well? Is it really that easy to rip the memory controller off the Hammer and slap it in a Socket-A platform, or will the Socket-A Sempron have an AthlonXP core, while the S754 Sempron has a Hammer core? Someone, please explain this to me! Thx.

 

[MDE]

The Socket A Sempron will most likely be based on the Athlon XP core and the 754/939 Semprons will be based on either the Hammer or Newcastle cores.

 

[White Widow]

That seems rather duplicitous. I realize the performance "rating" is adjusted to reflect the core differences, but to brand two fundamentally different pieces of hardware with the same name is somewhat misleading.

What's more frustrating on a practical level is that by icluding the AXP core into the "Sempron" family, AMD limits itself to the feature set they can offer with this brand. Intel will certainly give the Celeron 64-bit capability, not to mention enabling NX support. While these features might have a rather minimal effect performance wise, they offer a huge marketing advantage. This is an absurd position for AMD to put themselves in, considering the Sempron will be compteing directly against the Celeron, and considering Intel has an unfathomably large marketing budget; why would you want to just HAND Intel a marketing advantage for them to exploit?

AMD needs to be promoting the AMD64 architecture and all its benefits any way they can. Intel is gonna start telling people that 64-bit PC's are not just a privelege but a right, whether you buy a top-of-the-line compuyter or a budget machine. Consumers will turn their noses up at anything as dated and limited as a 32-bit system. AMD needs to look beyond the immediate future and commit themselves fully to their new architecture. What a pitty...

 

[Vee]

"The Socket A Sempron will most likely be based on the Athlon XP core and the 754/939 Semprons will be based on either the Hammer or Newcastle cores."

Sempron socket A will be based on "Thornton" core. Sempron 754/939 will be based upon "Paris" core.

 

[batmanuel]

I don't think that this is any worse than all of the various cores and cache combinations (PII, PIII, PIII-T, P4) that Intel has given the Celeron moniker to. The intended market for these processors won't know or care if the core is A64 or AXP derived, but the overclocking community to whom these differences DO matter to will know just by looking at the specs what they are getting.

Also, Intel really likes to keep the Celeron line distinct from the Pentium line by horribly crippling the chips, so I rather doubt that 64-bit or NX capability will be making its way into the Celeron line any time soon (especially since the Celerons don't get even get HT). The next Cleron upgrade, the one that will be competing with the Semprons this fall, is to a Prescott core. With the deep pipeline of the Prescott being bottlenecked by having only 256k of L2 and a 533Mhz bus speed, these Celerons at stock speeds ought to be outperformed by the Semprons (although the Celeron-P chips may overclock like nobody's business if the 2.4A P4s are any indication, which could give the new Celerons some fans in the enthusiast community).

 

[VirtualLarry]

Originally posted by: White Widow
That seems rather duplicitous. I realize the performance "rating" is adjusted to reflect the core differences, but to brand two fundamentally different pieces of hardware with the same name is somewhat misleading.

What's more frustrating on a practical level is that by icluding the AXP core into the "Sempron" family, AMD limits itself to the feature set they can offer with this brand. Intel will certainly give the Celeron 64-bit capability, not to mention enabling NX support. While these features might have a rather minimal effect performance wise, they offer a huge marketing advantage. This is an absurd position for AMD to put themselves in, considering the Sempron will be compteing directly against the Celeron, and considering Intel has an unfathomably large marketing budget; why would you want to just HAND Intel a marketing advantage for them to exploit?

AMD needs to be promoting the AMD64 architecture and all its benefits any way they can. Intel is gonna start telling people that 64-bit PC's are not just a privelege but a right, whether you buy a top-of-the-line compuyter or a budget machine. Consumers will turn their noses up at anything as dated and limited as a 32-bit system. AMD needs to look beyond the immediate future and commit themselves fully to their new architecture. What a pitty...

Interesting observation, especially given that AMD is the one that introduced x86-64, but most probably true, as well. Intel is a master at exploiting "useless" technological improvements for their marketing gain. (MMX, HT, anyone?) In this case, x86-64 isn't actually useless, but AMD isn't using it as a marketing/features "lever" to gain marketshare like they could and should. It's sad, really. AMD is sitting on some really good tech here. I would hate for Intel to turn things around from a marketing POV and use that very tech as a selling point against AMD.

 

[yhelothar]

From what I understand, Sempron is using the AthlonXP core on both the socket-A and Socket 754.
AT's article specifically stated that it will not include x86-64 functionality.
I believe the 754 version is the AthlonXP core with the integrated memory bus built in.

 

[darkswordsman17]

I think the Sempron for A64 boards is A64 based with just the 64 bit functionality diabled (and maybe a lower cache or clockspeed)

As for Intel enabling the 64 bit functionality of their future Celerons, I think I'd doubt they would so that it would be even more cost-effective for them. I think it will be similar to the Semprons on A64 systems where they are chips that didn't operate right will full 64 bit functionality so they disable it. At least thats what AMD had planned to do with their budget line for A64 chipsets, but might have changed with Sempron.

After the way the P4 based Celeron performs, I'd doubt they would let it run 64 bit until all chips move to 64 bit.

Its definitely possible though that Intel could do it for marketing reasons.

Thinking about how the Prescott still isn't performing noticeably better than the Northwoods (even on the 775 boards), I can't imagine how poorly the Prescott based Celerons are gonna perform.

Oh and I think HyperThreading was actually a non useless technology. It didn't really offer benchmark performance gains, but it does help when multi-threading.

 

[Gagabiji]

Arn't the Semprons worse than the Athlon XPs? I know they are next gen and all, but...😕

 

[MDE]

Originally posted by: SirDude
Arn't the Semprons worse than the Athlon XPs? I know they are next gen and all, but...😕

Nobody really knows, but they are Duron replacements...

 

[White Widow]

I'm pretty sure Intel is going to enable64-bit for the next-gen Celerons. Unlike the K8 architecture, the P4 core relies heavily on cache for performance. Because the pipeline is so long, a cache miss really hurts performnce. So, the easist way for Intel to differentiate their processors is the to add ro reuce cacahe. Again, for a P4, cacje makes a huge diference in performance. This is NOT the case with the K8 core. The K8 core has a much shorter pipeline and because of this (and a few other reasons) cache size is not nearly so important. As a result, AMD cannot simply cut the L2 cache in half for their budget CPU and leave everything else intact. Doing so would create a "budget" CPU that performed almost as well as the mainstream part. Instead of reducing cache, AMD has decided to take away 64-bit functionality. The resaoning being that if they can't cripple performance, they might as ell cripple features. This kind of thinking is why AMD continues to struggle. Intel, on the other hand, can throw every feature possible into the Celeron because it will ALWAYS perform tangibly worse than the mainstream P4 becase if the reduced cache. Unfortunately, by choosing this approach to segment their product line, AMD leaves themselves open to exploitation by Intel, which is certainly a huge - and OBVIOUS - mistake. When will AMD learn?

On another note - the reason AMD is using the AXP core is that

1) they have a lot of manufacturing equipment set up and running very efficiently making these cores

2) All the MB manufacturers and OEM's have large production lines, mature engineering, etc. based around Socket-A

The more I think about it, the less unreasonable this decision is by AMD. Of course, only the Socket-A Sempron's will be AXP-based. The S754 Semprons will have a K8 core of soe kind with the x86-64 registers disabled. This *might* be because the core is a POS and they needed to "downgrade" it, but likely this will NOT be the case. AMD has already claimed their yields for K8 are as good as for K7. Unlike the Celeron, where even with all the bells and whistels it still underperforms due to low cache, the Sempron will probably perform very well without the bells and whistles. Amd is going to transition the A64 to it's 90um process, and theave their 130um A64 process intact to make Semprons (and leave their AXP equipment intact to make S-A Semprons).

For those of us who don't need 64-bit capabilities, I am pretty sure the S754 Sempron will be a great performer. Only probelem? The world+dog is going to think the Celeron is a better choice because Intel will enable the 64-bit registeres in there and tell the world how great it is. How about adding some cache instead...