Six-core 32nm Westmere Full Review @ HKEPC

Printable View

Show 100 post(s) from this thread on one page

08-11-2009, 12:31 PM
poke349

Quote:

Originally Posted by Helloworld_98

0.072v on a quad?

Voltage sensors on Harpertowns are all broken:

I have yet to see a dual-Harper rig with correct vcore readings... lol
http://www.numberworld.org/nagisa_runs/specs.jpg

It's defintiely not 1.5v at stock... That'll fry a chip within weeks of 100% cpu load...
08-11-2009, 12:39 PM
jcool

Naw.. CPU-Z still cannot to this day display Vcore correctly on Dual LGA771 systems.. is it that hard to beleive? :p:
Vcore on the E5420 was 1,14V load or something.

@poke: It's not the sensor that's broken, just CPU-Z. Everest for example displays the Vcore accurately on Harpertown-Systems.
08-11-2009, 12:42 PM
poke349

Quote:

Originally Posted by jcool

Naw.. CPU-Z still cannot to this day display Vcore correctly on Dual LGA771 systems.. is it that hard to beleive? :p:
Vcore on the E5420 was 1,14V load or something.

@poke: It's not the sensor that's broken, just CPU-Z. Everest for example displays the Vcore accurately on Harpertown-Systems.

I get the same wrong readings from CPUz, Speed Fan, and believe-it-or-not, the BIOS... 1.5v for cpu-0 and 1.1v for cpu-1...

Maybe they're all reading it incorrectly. :rolleyes:

EDIT: Even when I switch the two processors, same readings 1.5 for cpu-0 and 1.1 for cpu-1.
08-11-2009, 12:57 PM
jcool

For me, Everest has always displayed correct CPU voltages. Both on the E5420 and the X5450 (5400/Z7S) and the Supermicro X7DCA-L (i5100). Give it a try ;)

Edit: here you go, my Dual X5450 in "summer mode" (slightly undervolted, stock Vcore on all X series is 1,25V like the VID says, E series is 1,15-1,20)

http://database.he-computer.de/Bilde...erestvcore.jpg

But now back to topic, shall we? :wasntme:
08-11-2009, 01:04 PM
poke349

Quote:

Originally Posted by jcool

For me, Everest has always displayed correct CPU voltages. Both on the E5420 and the X5450 (5400/Z7S) and the Supermicro X7DCA-L (i5100). Give it a try ;)

Sure... When I get my mobo back that is... :rofl: Until then it's just gonna keep on collecting dust. :rolleyes:

*It couldn't take 8 months of prime95-equivalent, ram and HD-intensive torture... but it's still under that pretty Tyan 3-year warranty. :D
08-11-2009, 08:14 PM
kl0012

Quote:

Originally Posted by informal

If BD is not 2011 thing anymore(as you say),when is it supposed to launch then?!
BD cores are now already in 2nd major revision,first being "canceled" back in the 2007.Since then,it "gained" 256bit AVX support and SSE5 on top of that(XOP etc. in new renamed nomenclature).
So Chad is right,BD is in best case scenario late 2010,or in worst case Q2 2011.

I suppose it would be moved back well into Q4 of 2011 or Q1 fo 2012. Of cause it is only my opinion and AMD's official roadmap says 2011, but AMD was never good in observation of initial dates of new arch release (Hammer was delayed for an year and not to mention Barcelona intro disaster). Another point, as I said, is Sandy Bridge which was already taped out and was targeted to intro in the same timeline (Q4-2010/Q1-2011). Since there are no words about Bulldozer tape out, to me is hardly believable that Bulldozer will catch up with Sandy Bridge.

Quote:

Originally Posted by LOE

27% increase in cinebench through 50% increase of cores/threads is not very good scaling in my book, and generally rendering software scales almost perfectly. Imo cinebench suck and should be substituted with a real world application that can actually scale and are in use

Faster Excel - WOOT thats what the world needs

Vangate CPU test is as useless as sandra, no one plays vantage, or works with it, synthetic benches are useless IMO

Power figures on the other hand are nice, but six dual threaded cores are an overkill for desktops. A 32nm quad with higher clocks will be MUCH BETTER than a lower clocked six core part.

It wasn't about reall/not reall world test. It was about your words "It's lacking performance everywhere except Sandra" There are no many desktop RealWorld well multithreaded apps in this test. There are no many such apps at all. And, of cause, there are no too many users which needs such apps (at least until games will get good multithreading support). Thats why Intel positioned it only for entusiast. But I've talked about its potential which can be seen not only from "sandra".
08-13-2009, 11:07 AM
saaya

wait, they got a couple more fps in games with gulftown at the same clocks and settings as bloomfield? 0_o
that doesnt sound right somehow... its minor gains, so id say fluctuations, a few fps up and down is normal... but gulftown has around 3-5fps more in all games they tested? hmmm

im very surprised about the power, same tdp with 2 more cores, just like amd... impressive!
if only they would focus on less cores with the same tdp but higher PERFORMANCE!!!! :P
08-13-2009, 11:36 AM
NBF

Quote:

Originally Posted by saaya

wait, they got a couple more fps in games with gulftown at the same clocks and settings as bloomfield? 0_o
that doesnt sound right somehow... its minor gains, so id say fluctuations, a few fps up and down is normal... but gulftown has around 3-5fps more in all games they tested? hmmm

im very surprised about the power, same tdp with 2 more cores, just like amd... impressive!
if only they would focus on less cores with the same tdp but higher PERFORMANCE!!!! :P

I think it could be the 12Mb L3 on Gulftown performing better than the 8Mb on Bloomfield.
08-13-2009, 12:02 PM
zalbard

Quote:

Originally Posted by saaya

if only they would focus on less cores with the same tdp but higher PERFORMANCE!!!! :P

Every Intel CPU developer must be a cruncher, haha. Too bad for most of us...
08-13-2009, 12:48 PM
Hornet331

Quote:

Originally Posted by NBF

I think it could be the 12Mb L3 on Gulftown performing better than the 8Mb on Bloomfield.

Dont forget we don't know how fast the cache is, if intel tweak it a bit its not unusual to see a bit gains there.
08-13-2009, 02:47 PM
Chad Boga

Quote:

Originally Posted by saaya

if only they would focus on less cores with the same tdp but higher PERFORMANCE!!!! :P

It is funny how when I was saying the same thing in recent weeks, I had people getting all sooky with me, when what you have just said is nothing more than common sense.
08-13-2009, 03:03 PM
poke349

Quote:

Originally Posted by saaya

if only they would focus on less cores with the same tdp but higher PERFORMANCE!!!! :P

There's a reason for that.
Current technology is nearing the limit for linear speed.

So the only way to increase performance is to add more cores.
08-13-2009, 03:51 PM
Hornet331

Quote:

Originally Posted by poke349

There's a reason for that.
Current technology is nearing the limit for linear speed.

So the only way to increase performance is to add more cores.

Its possible to make singlecore more efficent, but therefore x86 and all its itterations need to be gone...
08-13-2009, 05:49 PM
damtachoa

Intel just goes to the end of cores. I don't see any big performance on those cores. Waste of money.
08-13-2009, 07:22 PM
Chumbucket843

Quote:

Originally Posted by Hornet331

Its possible to make singlecore more efficent, but therefore x86 and all its itterations need to be gone...

true, but not by scaling with clockspeeds and internal parallelism can only scale so far.
08-13-2009, 10:16 PM
poke349

Quote:

Originally Posted by Chumbucket843

true, but not by scaling with clockspeeds and internal parallelism can only scale so far.

Exactly. :D

Think of work as having to move a bunch of school kids from the school to the museum. The faster you do it, the faster the work gets done.

Traditionally, the method was to make the school bus drive faster and faster.
That worked up until Prescott. Then they realized that they couldn't make the bus go any faster as it would be difficult to control and manage.

So Intel's idea was to use instruction-level parallelism (Itanium... SSE)... Basically making your school bus bigger... which helps but it also has limits.
At some point your bus becomes too long to manage any turns - not all programs/tasks can vectorize very well.

Intel, now out of options... has to resort to using multiple buses:rofl: Welcome to the world of multi-core computing.

And it's screwing over programmers... :rolleyes:

It's not gonna end there... I can see more problems in the future that will require drastic changes in hardware that will once again screw over programmers... I'm 90% sure they will happen. But just in case I'm wrong. I'll save myself the embarrassment by not writing it here. :D
08-13-2009, 10:41 PM
Chad Boga

Quote:

Originally Posted by poke349

It's not gonna end there... I can see more problems in the future that will require drastic changes in hardware that will once again screw over programmers...

Since when was the purpose of advances in hardware meant to be about making life easier for programmers?

Show 100 post(s) from this thread on one page