Intel Sandy Brdige Q4 2010?

[ambam]

I've read that the release date for Intel's "Sandy Bridge" is Q4 of 2010, or within the next Four months. Has Intel announced the actual name of this CPU? Have they dropped the "Core" prefix yet? The release model will only be quad-core, but in 2011 and 2012 there will be many 6 and 8 core models.

Intel as well as other hardware makers usually tease PC enthusiasts with advertisements and technical information about unreleased products shortly before they are released. But I have not seen or heard anything "official" from Intel about the Sandy Bridge, yet.

 

[jimmysmitty]

^The article was that with 33% more cores (6 core vs 8 core) they got 50% better performance. That means that a 8 core BD based CPU is at 150% better than a 6 core Deneb based CPU, thus the theoretical 12.5% per core comment.

1. Intel currently has a nice IPC advantage in pure performance. They really don't need to clock their CPUs higher since at a lower speed they perform the same as or better than AMDs offering. It was the same with Athlon 64 X2 vs Pentium D. Pentium Ds were clocked well above a Athlon X2, but didn't perform as well.

2. I doubt Intel is scared of BD. They have been trying to break the 4GHz stock barrier since Netburst and their 32nm is damn close. Intels 22nm was showing 3.8GHz stock at less than 1v. We may see 4GHz stock with 5GHz air clocked CPUs soon.

3. Intel more than likley will not limit OCing. This was a rumor for LGA1156 based Nehalems as well which proved to be false. There is also a lot of info that goes against this. Turbo alone is one reason. Turbo is a OCing tool for the masses. Add to that the 'K' editions of their LGA1156 CPUs. The chep 'poor' mans Extreme Edition CPU. They are cheaper and have a unlocked multiplier.

Also, Intels Core 2 series gained a lot of rep for being very OCing friendly. Intel put higher than needed TDPs knowing their CPUs could clokc higher than that. The QX9650 is a great example. It had a 125w TDP with a stock clock speed of 3GHz. It was able to be OCed to 4GHz on air while still being under that 125w TDP.

So as far as I can tell, BD wont be the amazing boost AMD needs to take the performance crown back from Intel. Even in the server market. Current Opterons with 2x more cores perform at about the same level as a Xeon with half the cores.

SB should boost performance a bit and will be out this next quarter. Until there is 100% diffinitive proof from Intel and third party reviewers, we should expect SB to OC just like Westmere CPUs do.

until then, as I said before, its all speculation.

 

[ares1214]

Ironic actually how we are the most knowledgable and opinionated group in the whole buying market. And yet we are likely less than 1% of intels profit, so they could care less :lol: But if you think about it, if at these clock speeds, we are seeing lets say 15% overall performance increase (thats likely best sandy bridge vs worst i7). We already see up to 20% increase from overclocking! take that away, and whats there to buy? A new socket? Yeah...

 

[jimmysmitty]

Ironic actually how we are the most knowledgable and opinionated group in the whole buying market. And yet we are likely less than 1% of intels profit, so they could care less :lol: But if you think about it, if at these clock speeds, we are seeing lets say 15% overall performance increase (thats likely best sandy bridge vs worst i7). We already see up to 20% increase from overclocking! take that away, and whats there to buy? A new socket? Yeah...

Yep. We are the smallest part of Intel and AMDs profits. Server is their largest followed by OEM.

But lets say intels prediciton of a Conroe like leap..... That would mean SB would have to be at least 20% better than Nehalem and probably overclock just as well. I would not be against that. But then again I am waiting for 22nm for a upgrade anyways since that will probably be of the best benefit.

 

[ares1214]

Wasnt 22nm suppost to be Ivy Bridge into Haswell. Thats a bit too far in the future to guess if ya ask me lol. And i really highly doubt it will be 20% better than Intel. For marketing purposes, I think we can be sure they compared best SB to worst i7. Thats 2.66 i7 vs 3.4 GHz SB. That right there is a 27% increase in clock speed. If they are getting 20% increase in overall performance, which is even being generous, then we are looking at a possibly slightly weaker core. All speculation though.

 

[yannifb]

These are the kind of threads i like, the speculation ones. They're just interesting imo. But remember that BD will use AVX as well as several new instruction sets that may really help performance. This was in the 1.25 bil settlement, it said something along the lines that AMD is now allowed t use these and future instruction sets that intel uses. It was near the part describing how AMD wouldn't have to pay quarterly x86 license fee's.

 

[ares1214]

Yeah, i remember something like that. Too be honest, clock for clock, i dont expect much from SB. If it cant OC, which the motherboards might fix, its entirely dead to me. Although who knows, BD might not be much better. Lets not forget last time AMD did a full arch change...

 

[jimmysmitty]

Wasnt 22nm suppost to be Ivy Bridge into Haswell. Thats a bit too far in the future to guess if ya ask me lol. And i really highly doubt it will be 20% better than Intel. For marketing purposes, I think we can be sure they compared best SB to worst i7. Thats 2.66 i7 vs 3.4 GHz SB. That right there is a 27% increase in clock speed. If they are getting 20% increase in overall performance, which is even being generous, then we are looking at a possibly slightly weaker core. All speculation though.

Actually it was a 2.5GHz SB chip. It says so in the article.

I was basing the 20% off of Prescott -> Conroe that Intel said SB is supposed to be compared to Nehalem. According to those preliminaries from Intel, they are pretty close and they are not even on final products.

Yeah, i remember something like that. Too be honest, clock for clock, i dont expect much from SB. If it cant OC, which the motherboards might fix, its entirely dead to me. Although who knows, BD might not be much better. Lets not forget last time AMD did a full arch change...

K8 was the last major arch change. It was good. K10 was the last jump from the same arch. That was bad. Hopefully Bulldozer will be decent.

 

[ambam]

Actually it was a 2.5GHz SB chip. It says so in the article.

I was basing the 20% off of Prescott -> Conroe that Intel said SB is supposed to be compared to Nehalem. According to those preliminaries from Intel, they are pretty close and they are not even on final products.



K8 was the last major arch change. It was good. K10 was the last jump from the same arch. That was bad. Hopefully Bulldozer will be decent.

I heard that the bulldozers are supposed to be promising. Probably at par with, or even slightly faster than the Nehalems. Probably within the $300-400 range for the flagship model. Instead of $1,000+++ like Intel's flagship CPU.

 

[jimmysmitty]

I heard that the bulldozers are supposed to be promising. Probably at par with, or even slightly faster than the Nehalems. Probably within the $300-400 range for the flagship model. Instead of $1,000+++ like Intel's flagship CPU.

AMDs pricing is only based on their performance compared to Intels current CPU equivalent. The reason why AMDs 6 core CPUs were $300 or less was because the fastest 6 core Thuban barley kept up with or beat Intels $300 quad.

Since BD is coming out after SB, they will price BD based on BDs performance compared to SB not Nehalem. Is BD is as good as Nehalem, then they will price it pretty low. If its better than Nehalem, they will price it a bit higher than Nehalem. If its better than SB then they will price their flagship CPU to $1K+.

Its in their history, look at K8. K8 was their best CPU. It pounded Intel and got them the recognition they needed. So their flagship FX based CPUs went for $1K+ and are still expensive.

 

[jonpaul37]

This we do not know yet. Even if they drop a pin or two it can still be put into a older socket. AM3 has less pins than AM2+ but they work.

It has been confirmed that the new core line from intel will not work with 1156 or 1366

 

[simon12]

Well heres what I think. 1st I hope BD is better than SB as I think it will do the whole market alot of good if Intel has to play catchup again. As for SB overclocking my understanding was everything was linked to the same clock speed so unless board manufacturers can unlink them only chips with unlocked multipliers will be able to overclock and who knows how many there will be.

 

[ares1214]

Thats exactly what I heard. See, the same was said for nehalem and lynnfield, but it was never really demonstrated. People demonstrated how it could only get a stable OC of less than 100 MHz, as the USB and SATA crashed. If motherboards dont do anything, im fairly convinced overclocking will be limited, or none at all, besides on the K edition. And if it wasnt, why would there be a K edition? So while im just speculating, i do definitely think SB overclocking will be, well nothing, except on the K edition, when even then it might just be uping the multiplier, not the bus.

 

[jimmysmitty]

It has been confirmed that the new core line from intel will not work with 1156 or 1366

You got a link?

Thats exactly what I heard. See, the same was said for nehalem and lynnfield, but it was never really demonstrated. People demonstrated how it could only get a stable OC of less than 100 MHz, as the USB and SATA crashed. If motherboards dont do anything, im fairly convinced overclocking will be limited, or none at all, besides on the K edition. And if it wasnt, why would there be a K edition? So while im just speculating, i do definitely think SB overclocking will be, well nothing, except on the K edition, when even then it might just be uping the multiplier, not the bus.

I am still awaiting officia word. If there is a limit to any OCing I would understand it for the lower end ones with the IGP thats going to be integrated on the CPU core. But as for the CPU only chips, I still doubt Intel would limit OCing and kill themselves like that.

 

[ares1214]

Thats the thing, it wouldnt be killing themselves, its helping themselves. Like i said, we are but a sliver of a percent of the market, intel could care less what we think, as does AMD. If intel limits OC'ing on everything but the K edition, they stop annoying 11 year olds from messing up computers trying to OC, and they make motherboards cheaper. Then they make us pay extra (knowing intel, a lot extra) to get OC ability, since we are the only ones that need it. It might seem dumb to us, but its actually a fairly smart move. All the same, if they do limit it, and K editions are expensive as i think they will be (the 655K is 17% more expensive than the equivalent clock speed non unlocked. those are just unlocked, im sure it will be more expensive for flat out overclocking.) then AMD is looking pretty nice right now. Doubly so considering I have AM3 mobo. Gotta give em credit, might not be the fastest, but atleast things are compatible.

 

[gamerk316]

"FPU?" Floating point unit?

There aren't any games which make use of a quad core processor, let alone 6 or 8. It is very unlikely a PC game will exist which fully utilizes 6-8 cores for many, many years. Even Crysis only uses 2 cores. The ONLY PC game that uses a quad core is GTA4, due to a poorly coded port from consoles to PC.

Maybe Crysis 2 will use 3, hopefully 4 cores.

A lot of games are starting to use 4 cores; Dragon Age uses up 90% of my QX9650, and I know a few other games that operate in the 80-85% range across all my cores. Some games are already being coded to support up to 8 cores...

Again, its not hard to scale to multiple cores; threading is easy enough (and already being done), so provided there are no bottlenecks to consider between different threads, the only other thing to do is ensure each thread is dispatched to a unique CPU. [Of course, dispatching a thread to a CPU already under work is worse then not doing so in some cases, so that needs to be accounted for as well...]

 

[jimmysmitty]

^Multicore is not the hard part. The hard part is optimizing it to work well with that extra performance advantage.

 

[gamerk316]

^^ Agreed. In my opinion, Windows is not a good OS for massivly multithreading anyway (due to all those dynamic DLL's and the bottleneck conditions they bring into play...), but it can be done, and be done well with a bit of work.

In any case, I'm not really expecting more performance like we see on the GPU side, just smaller and less power consuming. 10% gains would be impressive to me.

 

[ares1214]

I wonder how an Android OS for desktop would do performance wise. There was that article saying Nvidia might enter the market with an Android OS in mind.

 

[fazers_on_stun]

A lot of games are starting to use 4 cores; Dragon Age uses up 90% of my QX9650, and I know a few other games that operate in the 80-85% range across all my cores. Some games are already being coded to support up to 8 cores...

Again, its not hard to scale to multiple cores; threading is easy enough (and already being done), so provided there are no bottlenecks to consider between different threads, the only other thing to do is ensure each thread is dispatched to a unique CPU. [Of course, dispatching a thread to a CPU already under work is worse then not doing so in some cases, so that needs to be accounted for as well...]

Ahh, Dragon Age ... I'm so looking forward to DA2 next year!

Anyway, according to http://www.techpowerup.com/129392/AMD_Details_Bulldozer_Processor_Architecture.html it looks like BD won't be backwards compatible with any AM3 or AM2+ sockets, but require a new one. So AMD fans are in the same new-mobo boat as Intel enthusiasts, it seems.

 

[bobdozer]

I'm glad because I kept my AM2+ socket with my last cpu and it will be nice to buy a new motherboard this time around. I do hope AMD put better integrated graphics into the new AM3+ mobos though.

 

[ares1214]

Its too early. AMD said they would, people say they wont, who knows really.