Printable View
Upping the L3 from 6 to 8 MB will not have significant performance boost in real world apps. Unless one wants to rotate with approx ~12-16 *N MB dataset and is good at cache thrashing just to demonstrate the importance of cache. Then again, it is also possible to craft a piece of code in which SB's (possibly to be glorified later on by certain person)branch predictor fails 100 %, to demonstrate how stupid the BP actually is. :p:
CB 10 is just too damn old. they should have used CB 11.5 for single to multi speedup ratios.Quote:
Originally Posted by mAJORD
they also used 3dsmax 9 for the 3dsmax benchmark.
I earn my living with 3dsmax, they basically used a 4-5 generation old software.
You have after 3dsmax9, 3dsmax 2008, 2009, 2010 and 2011, all with improvements to multi=threading/rendering performance.
using 3dsmax 9 it's like the guy who invented the round wheel started to use a square one the next day.
i dont think SB did a very good job with their GPU honestly.Quote:
Originally Posted by Hornet331
looking at the IGP vs best gpu across time
we got 40SPs back when 800SPs was the max for ati cards, so 1/20 ratio.
this gpu is about the same as 80SP, and it will be double that with their 2C version so around 160SPs,
and when its due, we should see new ATI cards stronger then the current which is already at 1600SPs, which puts the SB gpu under 1/10th of max
basically what im saying is that while SB might beat out all IGPs, those havnt changed much in the last 3 years anyway, and this is going to be as weak as an IGP in 2011. SB will have no purpose being in a miniITX sized case, claiming to be a gaming PC, while Llano will have that ability.
People should remember that there will most probably be new products. 240 SP Llano and 2C SB tomorrow, 480 SP Bulldozer-based Llano and 4C SB day after tomorrow.
Llano's GPU will do fine. Ontario... That little guy will get bulldozed.
Well ontario is an Atom competitor in netbook segment,it won't compete with anything that fast in GPU subsystem anyway.
I wonder who claimed that.. mini itx gaming rig... even Llano will be hard on the line for gaming @ 1280x1024 if it performs like a 5570.Quote:
Originally Posted by Manicdan
Sure we raise the baseline, but as it stand igp still and will suck for gaming. Especial since the avarage resolution is increasing. Imho in a view years 1920x1080 will be the new mainstream resolution... even if i hate 16:9 for pcs. ;)
Well Ontario's GPU (on paper) should keep up with even a 2c SB GPU.
Thank you.Quote:
Originally Posted by IntelUser
Casual gamers.. Kids, etc, don't need to run native Res to play games.Quote:
Originally Posted by Hornet331
I can tell you first hand they don't care.. I've set up BFBC2 on a HD5570 @ 1368 Res on a 1080 23" Monitor.. The non native Res softens everything a bit of course, but in a game it's not a bad effect.. And quite a playable Framerate on this particular game.
..80 SP? 5450 with 80 SPs can't beat 1C SB GPU, let alone 2C model which should be some 85-100 % faster. I'd guess, if one can compare like this, that Intels 1C is equivalent to approx 80-100 SP, based on these early numbers. So 2C model could almost touch 240 SP Llano.Quote:
Originally Posted by mAJORD
However, I'm sure ATI engineers know their stuff better than Intel engineers, they have a slight advantage there, but they also have disadvantages like mfg process.
monitors are generally one of the last things upgraded for average joe. they dont realize how quickly the next resolution has come down in price, because they dont think it does, nor do they even look as long as their current monitor works.Quote:
Originally Posted by Hornet331
1680x1050 is still the current trend, and considering people would be buying a tiny pc, they might not want a super big monitor anyway. they would be happy enough if games looked like their consoles, which a 5570/Llano should do, but a 2C SB still cannot. SB with 2C is ~160 SPs, Llano is 480, how can they even be remotely close in gpu comparisons.
Thats not a big deal to copy-paste additional gpu cores on the die, but the target segment for such solution is small to none. It would be costly (because of the large die size), it would suck power like a mad (forget notebooks segment...), it would be slower then similar discrete solution (because of memory lmitation) e.t.c. Current SB IGP looks like good trade-off. Let see what AMD will bring to the table...Quote:
Originally Posted by Manicdan
And after all SB is a first cpu which fully shares mem subsystem (including L3) with IGP. If Intel decide to implement OpenCL for its igp then cpu core <-> igp data interchange would be superfast. Also I still hope Intel will implement better support for software shader/vertex processing in its drivers. After all AVX has 200 SP GFLOP/s of computing power (twice of ATI HD 5450) and shared L3 can help a lot in this case.
You're right, sorry. 80SP indeed would be around SB 1C performance. not sure what I was thinking.Quote:
Originally Posted by Calmatory
As informal said though ontario is a part aimed at very low power notebook/netbooks, so an 80SP part will still be quite decent
Umm, 200 GFLOPS based on what, at which clocks and how many cores? AVX itself isn't anything more than just instructions slapped on silicon, like MMX, 3DNow! and SSE, which makes me wonder all the glorifying.Quote:
Originally Posted by kl0012
(8x 32bit add + 8x 32 bit mul) x 4 cores x 3.1 GHz = 198.4Quote:
Originally Posted by Calmatory
Your underestimating how fast 1920x1080 is adopted, in steam its already on 3rd place, if the trend continues in a few months 1920x1080 will even surpass 1280x1024 as dominant resolution.Quote:
Originally Posted by Manicdan
Everyone I know who still had a 19" 4:3 monitor switched to a 1920x1080 16:9 monitor... which is no wonder, since they basically cost the same as a 1680x1050 16:10 display.
No. According to Anand, scores was done without Turbo :Quote:
Originally Posted by MAS
Quote:
The other major difference between what’s out today and what’s coming in Q1 is turbo. Early Sandy Bridge samples, ours included, do not have turbo enabled. The CPU simply runs at 3.1GHz all the time, regardless of workload. The final retail 2400 will be able to run at up to 3.4GHz.
Actually, Anand used his historical data from prior reviews in his charts, all those CPUs, including the Phenom, were ran with Turbo on where as SB ran without turbo because the engineering sample had it disabled (re-read the Test page) ... so the clock for clock comparison is obfuscated by this anomoly.Quote:
Originally Posted by MAS
Whatever clock for clock % increase you see in that data, add another 6 or 7%.
In terms of what went into SB -- nothing has been disclosed, just a few random leaks, we don't know what changed in SB to any great extent.
10-15%, on average, where have I seen that estimate before???Quote:
Originally Posted by saaya
you missed the part where the 5450 uses 10 watts not including board power. the 5450 is 5gflops/W and the 5970 is 15gflops/W. big difference.Quote:
Originally Posted by kl0012
but comparing flops to flops as a performance metric is probably about as inaccurate as you can get. we have already seen the 5870 achieve 2.2tflops in sgemm. an ideal case for sandy bridge is 20x slower.
In Anands test the diff between idle to load for SB is ~60W means that we have 3.3gflops/W which is not far from 5450 and also OOO/branch prediction/prefetchers/ which uses most of the power allows to utilize these flop/s in more effective way. BTW nvidia gt310/320 has even lower gflops/w ratio.Quote:
Originally Posted by Chumbucket843
~12x slower you mean?Quote:
but comparing flops to flops as a performance metric is probably about as inaccurate as you can get. we have already seen the 5870 achieve 2.2tflops in sgemm. an ideal case for sandy bridge is 20x slower.
http://software.intel.com/en-us/arti...mkl-v103-beta/
* AVX DGEMM (M, N, K=8Kx4Kx128) performs 1.8x over NHM. AVX DGEMM/SGEMM achieves 88-90% machine peak.
* The AVX/NHM speedup is 1.8x for radix-2 1D CFFTs with N=1024
* The Intel® Optimized LINPACK benchmark, using Intel AVX optimizations, performs over 1.86x (or over 80% overall efficiency) on 4 cores with N=20000.
60 W power usage delta for 80 SP core? What and where!? :confused:Quote:
Originally Posted by kl0012
true, but if you can choose, for about the same price, between native res and lower res, dont tell me people still dont care or even go for the lower res ;)Quote:
Originally Posted by mAJORD
that doesnt necessarily mean it was done without turbo... :DQuote:
Originally Posted by Olivon
just cause cpuz or other apps dont detect that the cpu turbos up, doesnt mean it doesnt turbo up... :D
it does sound odd... if the 880 was really running at 3.33-3.73, which is 7.5%-20.5% faster, and the 3.1sb part was still 10-15% faster, then that would mean an ipc boost of 17.5-35.5%... and that DOES sound too good to be true... i mean really...
IF its true, then wow... sb will really be a big step ahead... but i find that hard to believe...
in an old post of mine where i posted that imo even a 10-15% ipc boost of sb wouldnt be enough to make an upgrade worth it for most people? :DQuote:
Originally Posted by JumpingJack
I trust Anand when he said that's done without Turbo. He knows his job. ;)Quote:
Originally Posted by saaya