Haswell model specs leaked

Page 15 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
NTMBK

NTMBK

Lifer
Nov 14, 2011
10,480
5,897
136
Acanthus said:
What.

The performance difference in most apps between ddr3-1333 and ddr3-2133 is 0-4% save synthetics and 7zip.

The board TDP limit? it would be a smaller die?

Are you suggesting we haven't done this before? Or that Intel isn't doing it right now with the same technology and just overcharging?

http://www.newegg.com/Product/Produc...82E16819174635

130w

http://ark.intel.com/products/52576/Intel-Xeon-Processor-X5690-12M-Cache-3_46-GHz-6_40-GTs-Intel-QPI
Click to expand...

130W- compared to the 2600k's 95W, including on die graphics. Yeah, that's a big jump in TDP.

Those memory bottlenecks get worse with more cores. It's pretty obvious. There's a good reason why Sandy Bridge E has a quad-channel memory controller.
 
M

MightyMalus

Senior member
Jan 3, 2013
292
0
0
Considering that I been wanting to build my first desktop for awhile and I have been waiting for all sorts of info about Haswell, the more I know, the less I even want one.

If GT3 is all about perf/watt, does that make the GT2 the "enthusiast" iGPU? GT2>GT3? wtf Intel, WTF!

Not only that, no GT3 in desktop OR notebooks. Again, what the heck? They got a deal with Nvidia or something? I would not be surprised about that.

To sum it up, all I cared about Intel was its iGPU.
 
ShintaiDK

ShintaiDK

Lifer
Apr 22, 2012
20,378
146
106
MightyMalus said:
Considering that I been wanting to build my first desktop for awhile and I have been waiting for all sorts of info about Haswell, the more I know, the less I even want one.

If GT3 is all about perf/watt, does that make the GT2 the "enthusiast" iGPU? GT2>GT3? wtf Intel, WTF!

Not only that, no GT3 in desktop OR notebooks. Again, what the heck? They got a deal with Nvidia or something? I would not be surprised about that.

To sum it up, all I cared about Intel was its iGPU.
Click to expand...

There are 3 GT3 models.
Regular laptop quadcore+GT3.
Regular laptop dualcore+GT3.
ULV laptop dualcore+PCH+GT3.

GT3 is only about performance/watt in the ULV.
 
A

Acanthus

Lifer
Aug 28, 2001
19,915
2
76
ostif.org
NTMBK said:
130W- compared to the 2600k's 95W, including on die graphics. Yeah, that's a big jump in TDP.

Those memory bottlenecks get worse with more cores. It's pretty obvious. There's a good reason why Sandy Bridge E has a quad-channel memory controller.
Click to expand...

How much power do you think my CPU is drawing?

I think it might be a tad more than 95w.

http://www.legionhardware.com/articles_pages/sandy_bridge_e_memory_performance_guide,8.html

Go through that, and tell me again about how bandwidth starved 6 cores are.

Difference between DDR3-1333 and DDR3-2133
Winrar = 23%
Photoshop = 8%
x264 = 8%
Dirt3 = 0%
Just Cause = 0%
The Witcher = 0%
 
M

MightyMalus

Senior member
Jan 3, 2013
292
0
0
ShintaiDK said:
There are 3 GT3 models.
Regular laptop quadcore+GT3.
Regular laptop dualcore+GT3.
ULV laptop dualcore+PCH+GT3.

GT3 is only about performance/watt in the ULV.
Click to expand...

Where is the info about the GT3 in notebooks being higher performance parts than the ULV GT3? I have not seen that.


But, on the Desktop side of things, looking at the artic leaks here -> http://wccftech.com/arctic-cooling-...50-cpu-lineup-core-i73980x-confirmed-lga2011/

I wonder what a cheap(EDIT:low clocked) HSW cpu will cost? Mainly the Core i7-4765T.
What's comparable in price by IVB terms right now?

I might go haswell if the cpu+mobo combo doesn't cost an arm and a leg...
And I also hope the "1150 low power" and "1150 standard power" sockets don't mean anything other than the TDP's of the chip.
 
Last edited:
ShintaiDK

ShintaiDK

Lifer
Apr 22, 2012
20,378
146
106
Homeles said:
It'd be a miracle if it was representative of the final product? Can you imagine getting a flawless Haswell back from the fab?
Click to expand...

At this time they better be close, if not flawless. Mass production will soon start.
 
FlanK3r

FlanK3r

Senior member
Sep 15, 2009
321
84
101
So, I believe, someone must have some ES to show :)...I saw only one old benchmark of first ES at Chiphell, but is it long time ago..:(.
 
H

Homeles

Platinum Member
Dec 9, 2011
2,580
0
0
FlanK3r said:
So, I believe, someone must have some ES to show :)...I saw only one old benchmark of first ES at Chiphell, but is it long time ago..:(.
Click to expand...
Honestly, I don't want to know ahead of time for Haswell. I want it to be like opening a Christmas present.

But that isn't going to happen.
 
ShintaiDK

ShintaiDK

Lifer
Apr 22, 2012
20,378
146
106
Alittle more info perhaps:

y8olozde1.png
 
Last edited:
M

MightyMalus

Senior member
Jan 3, 2013
292
0
0
IntelUser2000 said:
Quote: Originally Posted by Haserath View Post Does anyone know the theoretical flops of Ivy's HD4k?
Click to expand...
IntelUser2000 said:
16EUs x 4 FMA per EU x 2 issue @ ~1.1GHz = 282GFlops. That's peak FLOPs. The practical FLOPs probably end up at ~220GFlops.
Click to expand...

"The raw performance of integrated GPU, in single-precision FLOPS, can be calculated as follows: EU * 4 [dual-issue x 2 SP] * 2 [multiply + accumulate] * clock speed.
For example, the HD Graphics 3000 is rated at 125 GFLOPS,[55] which is consistent with the formula (12 * 4 * 2 * 1,300 MHz). HD4K = 16EUs x 4 FMA per EU x 2 issue @ ~1.1GHz = 282GFlops."~from wikipedia

Can someone please dumb this down for me completely? I seem to not getting the HD4000 GFlops on my own. (16*4*2*1.150MHz = 147.2)
But its probably because HD3K is based on SNB and HD4K is based on IVB, meaning that IVB have different "FMA's" and "issues" or something...and making the info in wiki above, incorrect also?

Any help would be greatly apretiated!
 
Last edited:
IntelUser2000

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,787
136
MightyMalus said:
Can someone please dumb this down for me completely? I seem to not getting the HD4000 GFlops on my own. (16*4*2*1.150MHz = 147.2)
Click to expand...

You are missing a critical enhancement with Ivy Bridge.

Sandy Bridge has a math unit with every FMA vector pipeline. Ivy Bridge takes that math unit and adds the ability to execute lot of the common instructions the FMA pipeline can do. So it effectively becomes like having 2 FMA units.

So you gotta go and double your number again for Ivy Bridge.

Regular laptop quadcore+GT3.
Regular laptop dualcore+GT3.
ULV laptop dualcore+PCH+GT3.
Click to expand...

No, only the mobile quad core and ULTs are getting GT3. The regular dual cores don't.
 
M

MightyMalus

Senior member
Jan 3, 2013
292
0
0
IntelUser2000 said:
You are missing a critical enhancement with Ivy Bridge. Sandy Bridge has a math unit with every FMA vector pipeline. Ivy Bridge takes that math unit and adds the ability to execute lot of the common instructions the FMA pipeline can do. So it effectively becomes like having 2 FMA units. So you gotta go and double your number again for Ivy Bridge.
Click to expand...

I got it from the wiki article. But thanks. Now lets see...

So for IVB HD4K it would be 16*8*2*1.150MHz = 294.4GFlops?

And a theoretical HSW HD4600 is 20*8*2*1.200MHz = 384GFlops

So, about a 30% increase from IVB, I guess it does work.


Now for the GT3 is a bit theorical even more, unless I missed the frequency of them?
Ahh, lets just see.

40*8*2*500MHz = 320GFlops
Likely scenario -> 40*8*2*600MHz = 384GFlops (Higher than a 6620G, 355.2GFlops)
40*8*2*700MHz = 448GFlops (Higher than a 7560D, 389.1GFlops)
40*8*2*800MHz = 512GFlops
40*8*2*900MHz = 576GFlops

Its seems that no matter how you look at it, HSW iGPU won't be much competition against AMD's, what AMD already had for a year. Hopefully 4th gen Intel graphics on Broadwell will bring a bigger impact.

Kinda saddened really -.-'' Is there any info about actual architectural changes on the iGPU of Haswell? Not just rumors? Seems HSW is just IVB iGPU's+4 EU's.
 
IntelUser2000

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,787
136
MightyMalus said:
40*8*2*500MHz = 320GFlops
Likely scenario -> 40*8*2*600MHz = 384GFlops (Higher than a 6620G, 355.2GFlops)
40*8*2*700MHz = 448GFlops (Higher than a 7560D, 389.1GFlops)
40*8*2*800MHz = 512GFlops
40*8*2*900MHz = 576GFlops

Its seems that no matter how you look at it, HSW iGPU won't be much competition against AMD's, what AMD already had for a year. Hopefully 4th gen Intel graphics on Broadwell will bring a bigger impact.
Click to expand...

Search Anandtech for Haswell IDF coverage.

I'm thinking it'll end up at ~1GHz for the quad core GT3. With 640GFlops, that's approximately double what a Ivy Bridge GT2 running at 1.3GHz achieves. Since 20% gain in average will be enough to match Trinity in the mobile, they should be able to do it with GT2 alone. Of course they'll still be behind for Desktop.

Yes, there are cases where the HD 4000 is more than 20% behind but Haswell increases more than computational performance as well. Like doubling fixed function pipeline performance, and increase in texture performance in some scenarios up to 4x. Also adding the Resource Streamer to reduce driver bottlenecks. Those should help a lot to close up in the weak areas.
 
Last edited:
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom