Intel to show off Sandy Bridge at IDF next week, AMD counters with Zacate demo nearby

[Mechanical Man]

i3s go in cheaper laptops or smaller laptops that aren't usually suitable for better discrete video cards. On the other hand the i5-430m with the HD 5650 does give a glimpse of what the i3 could do, if somebody paired it with a better video card.

Yes it could do, my point was entirely that amd offering is more balanced than intel one.

[ajaidev]

Nice to see the details emerging, i would like to point out that the 3mm^2 space allotted to video recoding engine as one of the highlights of the architecture. The amount of flexibility Sandy Bridge offers to mix and match is bound less. Ivy Bridge will improve on this and as i had stated earlier Q4 2011 is when u'll see the Ivy's.

Another thing that i wanted to put into perspective is that the llano video recoding capability maybe more flexible "Using the GPU cluster" but it would not be much faster than SNB that is even if it can match SNB's performance level.

Nvidia should open up CUDA for all that is the only way Nvidia can fight this war or else Intel will add specific function hardware for specific things and this is what flexibility is all about.



That looks minuscule lol the SNB i was given had a water cooler then again that was a 8c and 4-5 months ago.

[Hornet331]

Nvidia should open up CUDA for all that is the only way Nvidia can fight this war or else Intel will add specific function hardware for specific things and this is what flexibility is all about.

Which is ironically the oposite of what intels future vision of x86 is..

That looks minuscule lol the SNB i was given had a water cooler then again that was a 8c and 4-5 months ago.

SB-EN is still 12months away, enough time to get the power down.

[Mechanical Man]

SB is pretty impressive indeed.

[ajaidev]

What a beauty:

[FlanK3r]

nice photos

[savantu]

Those cores are so friggin huge compared to anything else. The equivalent of the muscle car.

[FlanK3r]

or for me to Bulldozer chip (like more AMD)

[Nintendork]

AMD current gen mobile platform:
x2 L625 (18W), 780G + SB710 (11.5W+4.5W)

AMD next gen mobile platform:
Ontario (9W), Hudson M1 / D1 (~6W)
Zacate (18W), Hudson M1 / D1 (~6W)

Intel Atom mobile platform:
Atom D51x - Pineview (13W), NM10 (2W)

Intel Core ULV platform:
i3-330UM - Clarsdale (18W), Ibex Peak (3.5W)

Error.

Turion II X2 Neo K665 1.7Ghz 15w 45nm (based in Athlon II) + 880G / HD4225 DX10.1

L625 is an ancient cpu based in k8 brisbane 65nm. Need an update in cpu's history.

[bamtan2]

Those cores are so friggin huge compared to anything else. The equivalent of the muscle car.

huge compared to anything else except the chips that are huger

[Hornet331]

Someone asked for the presentation vids:

Overview:
http://intelstudios.edgesuite.net/id..._SPCS001/f.htm
Details (SB itself, GPU, QPI)
http://intelstudios.edgesuite.net/id..._ARCS001/f.htm
http://intelstudios.edgesuite.net/id..._ARCS002/f.htm
http://intelstudios.edgesuite.net/id..._ARCS003/f.htm

AVX presentation is also somewhere tehre but can't find it right now... :p

[blindbox]

Makes me wonder, are we really going to see 8-cores with chips that big?

Unless.. MCM.
/speculates


OT: Could anyone tell me what that other chip on Westmere is? Anyone that cares enough to do a die-size estimate of SB?

Sandy Bridge is still smaller than Nehalem.

[[XC] gomeler]

Makes me wonder, are we really going to see 8-cores with chips that big?

Unless.. MCM.
/speculates


OT: Could anyone tell me what that other chip on Westmere is? Anyone that cares enough to do a die-size estimate of SB?

Sandy Bridge is still smaller than Nehalem.

That Westmere chip is a Clarkdale. The GPU and memory controller are one chip, the cores and cache are the other chip.

[kl0012]

Makes me wonder, are we really going to see 8-cores with chips that big?

Unless.. MCM.
/speculates


OT: Could anyone tell me what that other chip on Westmere is? Anyone that cares enough to do a die-size estimate of SB?

Sandy Bridge is still smaller than Nehalem.

Well, the SB die size was rumored to be 225 mm2 while each core is about 20 mm2. So looking at SB die shoot we my assume IGP & 8Mb L3 are ~50mm2 each (even without exact measurement). So 8xCore SB shoulde be roughly:
4xCore SB - IGP + 12M L3 + 4 cores = 225-50+75+80 = 320 mm2.

6 core westmere is 248 mm2, 4 core linnfield is 296 mm2

[blindbox]

Well, the SB die size was rumored to be 225 mm2 while each core is about 20 mm2. So looking at SB die shoot we my assume IGP & 8Mb L3 are ~50mm2 each (even without exact measurement). So 8xCore SB shoulde be roughly:
4xCore SB - IGP + 12M L3 + 4 cores = 225-50+75+80 = 320 mm2.

6 core westmere is 248 mm2, 4 core linnfield is 296 mm2

I was thinking of photoshop when I said an estimate.

We need jaredpace here.

[DragoonXX]

<image removed>

That looks minuscule lol the SNB i was given had a water cooler then again that was a 8c and 4-5 months ago.

I can't see that cooler being too terribly effective, even at stock clocks (assuming that the heatsink on the right is the same size as an e8400 stock heatsink)

[kl0012]

Found it on amdzone.com:

I've just played some Batman AA on the Mobility 5470 for a while (same 1024x768, high as Anand) I found it ran at a solid 30fps, dropping only to 23fps and sometimes hitting 38fps.

This was with the 5470 512MB GDDR3 and Phenom II X3 1.8GHz

Do these results say anything about the number of SP's in Zacate?

Obviously, this system has an advantage with the mobile Phenom II X3 and dedicated GPU memory (GDDR3 1600MHz effective), so taking that into account does it look like Zacate is 80SP?

Off topic, it's nice to see that this low end 5470 can do some damage! 30fps avg on high, not bad!

EDIT:
I just ran Batman AA with the same settings on the ATI 4250. The results this time we more like Zacate, hovering around 16fps dropping to 12fps. I'm sure you know, but ATI 4250 is 40SP

http://www.amdzone.com/phpbb3/viewto...189875#p189877

I know that zacate has less mem BW and lower clock but wasn't it worth to reduce number of SPs (save die area) and increase clock speed and mem BW (which would help to CPU as well)? After all (if i remember correctly) 790GX has 15W TDP and adding two "sub 1W" cores would bring it to ~18W.

[Manicdan]

80 low clocked SPs is more efficient than 40 high clocked SPs
also 790gx is based on 55nm, so why even compare tdp

[Florinmocanu]

you cannot compare a discrete GPU with dedicated memory to an on-die gpu which uses the system ram. also probably the frequency of the gpu from Zacate is lower than the freq of the retail 5470.

So the results probably are comparable with an underclocked 5470, with low freq DDR2 , but such a card probably does not exist.

[kl0012]

80 low clocked SPs is more efficient than 40 high clocked SPs

You're wrong. The power consumption delta between let say 700Mhz and 500 MHz (assuming the same voltage) is much less then power consumption of additional 40SP (here you should add a leakage into equation)

also 790gx is based on 55nm, so why even compare tdp

Dosn't it contradicts your statement? Potentially 40nm 790GX would have even lower TDP.

[madcho]

/off topic

A new hero is born in france :
http://www.lalsace.fr/fr/region/mulh.....-a-velo.html

Just awesome !!!.

//

For moment this idf is really boring ...

[Hornet331]

/off topic

A new hero is born in france :
http://www.lalsace.fr/fr/region/mulh.....-a-velo.html

Just awesome !!!.

//

For moment this idf is really boring ...

IDF is already over... way to spam a thread.. .

[Manicdan]

You're wrong. The power consumption delta between let say 700Mhz and 500 MHz (assuming the same voltage) is much less then power consumption of additional 40SP (here you should add a leakage into equation)

if you look a FPS/Watt your wrong, which is what efficiency is. why do all laptop gpus run at lower clocks (and lower voltages)? cause its more efficient.

Dosn't it contradicts your statement? Potentially 40nm 790GX would have even lower TDP.

contradict what statement i said? i never made any statements about tdp. your comparing a desktop part based on a 3 year old process, vs something thats coming out shortly on a new process. you saw what AMD did with Deneb to Thuban on per core efficiency, i imagine they got better at making 40nm chips too, especially with idle power draw.

if your argument is then trying to figure out how much of the TDP is used by the gpu vs the cpu, then good luck, cause theres no way to even know what the actual power draw is right now since u only have a tdp, and keep in mind that ati cards are known for insane heat pulls in furmark and occt, and they might have taken that into consideration instead of going with ideal gaming power draw.

i really suggest just waiting for benchmarks where hopefully you can tell the power cost for the cores vs gpu with a much better guess.

[madcho]

IDF is already over... way to spam a thread.. .

Sry i didn't notice was over, but all info that gave intel was boring, zacate was a more interesting stuff.

[Hornet331]

Sry i didn't notice was over, but all info that gave intel was boring, zacate was a more interesting stuff.

Yeah one partially failed PR event with 2 benchmarks is more interesting then all the background information we got on SB (µarch improvments, infrastructure improvments etc.), first upcoming P67 boards, Lightpeak, display over USB 3.0, date on SB release, Tunnel Creek ...