Anand Sandy Bridge performance preview is up

[Idontcare]

that wasn't what I was getting at

 

[bryanW1995]

I'm not so sure about that.

The performance SB showed is enough for all non-gaming purposes (Intel says it can do 3D 1080p decoding), and who buys integrated graphics for gaming ? Llano will be better, we all know that, but it still won't be enough for serious gaming, so to some degree you have to wonder, what exactly is the point ?

it replaces the integrated chipsets and ensures that intel/amd get to keep all that money for themselves. also on laptops it will be huge, desktop mostly "meh" other than for the casual gamer at least for now.

 

[grimpr]

I seriously doubt they, ATI engineers, even cared about Sandys graphics capabilities.

you can bet some llano designers just breathed a big sigh of relief

 

[Cogman]

I seriously doubt they, ATI engineers, even cared about Sandys graphics capabilities.

The engineers might not have, but the management does. A strong sandy bridge graphics card will cut right into low cost and mobile market. The power savings and cost savings of including a graphics card on a CPU is not good news for AMD.

 

[BD2003]

I'm sure AMD cares. Up until this point, I was all about fusion. Now that I see SB has decent graphics, I'm probably going to stick with intel. They now have competition where it seemed they were gonna own the market.

Still, I hope AMD brings it, as the competition will benefit us all.

 

[ilkhan]

that wasn't what I was getting at

so...what were you getting at?

 

[BD2003]

I also really can't wait to hear more details on the video encoding. That's something that can potentially cut into high end sales. I'm really impressed with what intel is bringing to the table here.

 

[Idontcare]

so...what were you getting at?

http://forums.anandtech.com/showpost.php?p=30363094&postcount=46

 

[DrMrLordX]

Now that we've sorted out the whole BCLK thing, one thing I still want to know: are the uncore and QPI multis fully unlocked for K-class Sandy Bridge chips? And will there be any wiggle room with those multis on the non-K-class chips?

 

[ydnas7]

As this IGP was about 200% to 300% better than Clarkdale, I'ld guess it was 12EU, without turbo boost and with preliminary drivers. So getting close to 400% over Clarkdale is likely at launch.
I also guess that Anand had only a short amount of time to run the tests, with someone from intel on hand to not answer questions but to make sure anand didn't 'lose' the chip or especially the drivers. If so, then the tests (other than the drivers) were not AVX optimized as they would have been performed on an old SSD drive with the same software as back at clarkdale era.
It seemed to have DX 10 stable, what about DX10.1 or DX 11?
Reading about DX10 benchmarks running in a 3rd party reviewer would not have pleased AMD, They would have been hoping that the drivers were not up to that stage yet...

ATI 5450 has 292 million 40nm transistors that don't need to share thermal limits or bandwith limits with a CPU. SB IGP 12 EU and transcode unit probably has a similar transistor budget, superior transistor speed, but trades some 3D graphics ability for stronger video replay (inc 3D) (watch HD longer on battery power alone compared to gaming optimized graphics)

 

[ydnas7]

i'ld speculate that SB 6EU, is about half the performance of SB 12EU
that SB 12 EU approaches the performance limit that sharing DDR3 memory with a strong 4 core CPU brings and that Llano exceeds the the limit that sharing a DDR3 memory with a 'moderate' 4 core CPU brings.
Ie for lightly CPU intensive gaming, Llano wins solidly, but for average gaming, both are memory access bound and Llano's advantage is marginal.
Llano would however have a crossfire/SLI with a mid range ATI card for unbeatable bang per buck. Nvidia will find SB/Llano will eliminate discrete notebook graphics, eliminate low end PC discrete graphics, and llano with crossfire will make lots of Nvidia midrange graphics unprofitable.

 

[ydnas7]

SB 2400 and SB 2500 series are pretty similar, I doubt that SB i5-2500 will be less than 12 EU, but I could see SB i5-2400 being 6EU (so to differentiate it from SB-i5-2500), but so far Intel prefers not to shutdown cores (like amd's Phenom II X3 or X2) but to turn off features (ie no HT, reduce the cache, less corporate features). I'ld still expect SB graphics units that contain 12EU to run at 12EU (that includes i5-2400 series)

 

[Campy]

When SB gets released will we still be seeing a couple new releases for 1366 before LGA2011 or will there be no new CPUs for the 1366 socket?

 

[Nemesis 1]

When SB gets released will we still be seeing a couple new releases for 1366 before LGA2011 or will there be no new CPUs for the 1366 socket?

.
Haven't heard anything on the highend socket sandy . But 2011 looks to be it.
But the timing here is a bit confusing . By the time we see socket 2011 in 2011 22nm. Ivy bridge or what ever bridge is do . Looking at AT very incomplete results shows already that socket 1055 is very strong performance wise. The K series open changes my view abit on the highend. The reason being By the time we have highend sandy on 2011 low end 22nm is between 3 and 6 months away. So going to the 1055 socket has alot of merit over the highend . Looking at the results of intels present highend vz. ATs sandy. It looks like taking the middle road will assure higher overall performance for a longer period of time vs. the high end .

 

[Edrick]

.
But the timing here is a bit confusing . By the time we see socket 2011 in 2011 22nm. Ivy bridge or what ever bridge is do . Looking at AT very incomplete results shows already that socket 1055 is very strong performance wise. The K series open changes my view abit on the highend. The reason being By the time we have highend sandy on 2011 low end 22nm is between 3 and 6 months away. So going to the 1055 socket has alot of merit over the highend . Looking at the results of intels present highend vz. ATs sandy. It looks like taking the middle road will assure higher overall performance for a longer period of time vs. the high end .

Well, if we speculate the roadmap out even further, the high end could see higher overall performance for a longer period. It just depends on when the "period" starts.

Q1 2011 - LGA1155 SB (32nm)
Q2/Q3 2011 - LGA2011 SB (32nm)
Q1 2012 - LGA1155 IB (22nm)

Now this is where is can get interesting:
Q2/Q3 2012 - LGA2011 IB (22nm)
Or Intel could just skip that (almost similar to Westmere) and go to this:
Q4 2012 / Q1 2013 - LGA2011? Haswell (22nm)

 

[epidemis]

Ouch <-- Anand's credibility taking a hit.

 

[bryanW1995]

Well, if we speculate the roadmap out even further, the high end could see higher overall performance for a longer period. It just depends on when the "period" starts.

Q1 2011 - LGA1155 SB (32nm)
Q2/Q3 2011 - LGA2011 SB (32nm)
Q1 2012 - LGA1155 IB (22nm)

Now this is where is can get interesting:
Q2/Q3 2012 - LGA2011 IB (22nm)
Or Intel could just skip that (almost similar to Westmere) and go to this:
Q4 2012 / Q1 2013 - LGA2011? Haswell (22nm)

that would be great if they consistently released this way. however, last round they came out with the high end first so we have no idea what ivy bridge will be like.

 

[Nemesis 1]

Well, if we speculate the roadmap out even further, the high end could see higher overall performance for a longer period. It just depends on when the "period" starts.

Q1 2011 - LGA1155 SB (32nm)
Q2/Q3 2011 - LGA2011 SB (32nm)
Q1 2012 - LGA1155 IB (22nm)

Now this is where is can get interesting:
Q2/Q3 2012 - LGA2011 IB (22nm)
Or Intel could just skip that (almost similar to Westmere) and go to this:
Q4 2012 / Q1 2013 - LGA2011? Haswell (22nm)

AH yes haswell. Thats the whole point. 1155 is here for 2 years than what . 2011 is for a cpu without IGP . and haswell is suppose to bring alot. So does Haswell highend include Intel gpcpu ? Just as IDC stated about AM3+ Thats for BD. Without graphics.
What happens when these highend modles get graphics ? New Sockets also? Haswell weill be the last Bridge and the end to the socket game . As intel works towards a universal socket for its products. That was the plan.

 

[Nemesis 1]

that would be great if they consistently released this way. however, last round they came out with the high end first so we have no idea what ivy bridge will be like.

Well we can guess. Its either that the highend sandy released in third qt. is on 22nm. or the 1st.qt 2012 is Ivy released on 1155. IVY because of timing should be 1155 socket. Wouldn't ya think?

 

[khon]

Well we can guess. Its either that the highend sandy released in third qt. is on 22nm. or the 1st.qt 2012 is Ivy released on 1155. IVY because of timing should be 1155 socket. Wouldn't ya think?

I don't think they can have 22nm Ivy ready for Q3 2011, and I doubt they'll want AMD to release an octocore BD, while their latest line is still only dual/quad cores.

 

[IntelUser2000]

The performance SB showed is enough for all non-gaming purposes (Intel says it can do 3D 1080p decoding),

Arrandale/Clarkdale, will do 3D 1080p decoding when the drivers are released sometime this year. They promised summer release, but it seems they have a bit of a setback.

Haswell, will be similar to Sandy Bridge for release products. Mainstream first, high end later. It's to keep 1 year product cycle.

 

[Idontcare]

I don't think they can have 22nm Ivy ready for Q3 2011, and I doubt they'll want AMD to release an octocore BD, while their latest line is still only dual/quad cores.

I agree, Ivy will be a Q4 2011 at best, Q1 2012 more likely. Regardless what AMD does with BD.

 

[Nemesis 1]

I don't think they can have 22nm Ivy ready for Q3 2011, and I doubt they'll want AMD to release an octocore BD, while their latest line is still only dual/quad cores.

Thats the point. It wouldn't make sense either to skip 32nm high end. Haswell should be the final solution. Ivy should be 1st qt 2012 with enchantments on the lowend/middle. Than Ivy does its 2 years on highend socket 2011 Than haswell on the universal socket giving 22nm 3+ years befor 16nm.

 

[IntelUser2000]

enchantments on the lowend/middle. Than Ivy does its 2 years on highend socket 2011 Than haswell on the universal socket giving 22nm 3+ years befor 16nm.

Enchantments, somebody's been playing too much RPG's.

Are you suggesting delay on 16nm? Can you not do that if you are?


GT1 and GT2, interesting...
http://www.anandtech.com/show/3885/sandy-bridge-graphics-update

 

[KingstonU]

First time I hear of 16nm plans, and wow we are already talking Ivy Bridge and Haswell. In a weird way I'm actually happy that BullDozer and Sandy Bridge are still 6 months - 1 year away, cause my 2 year old C2Q is still more than enough for my needs, and even in late 2011 I'll probably only upgrade because I can and the new tech is exciting, not because I need to.