Yonah Preview at Anandtech!

Printable View

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

11-30-2005, 02:20 PM
zbogorgon

if i would conclude based upon this preview, i would say that i am kinda disapointed, I was expecting more, clock to clock should beat X2, but i am not, cause the only real thing is when some of HC guys over here (like macci :D) get this baby in his hands and show us the real potential. Right know we have no idea about real overclocking potential of any 64nm Intel proc, expecially yonahs, if they clock easily to 3.5ghz, we have a winner, none of the upcomming or existing AMD proc could beat that. If it clocks to about same level as dothans, that no real deal here :(.
11-30-2005, 02:44 PM
DilTech

Quote:

Originally Posted by kl0012

I do not understand why many so are disappointed.
Doom3:
Dothan(2,13GHz): 95.5 fps
Yonah (2GHz): 93.8 fps
Dothan has 7% higher frequency and only 2% faster than Yonah in Doom3. It points that Yonah will beat Dothan in games.
It also drive less power than Dothan:
http://www.tomshardware.com/cpu/2005...ntium4-19.html
http://techreport.com/reviews/2005q3...9/index.x?pg=7

They're comparing a 133 mhz fsb dothan to a 166mhz fsb yonah. Considering the 133mhz dothan was pretty limited by it's fsb, It's a pretty bad sign.
11-30-2005, 03:11 PM
B5I8

I think it's too early to tell how Yonah will be. I mean we're talking engineering sample processors on sample boards with bare beta drivers. Don't be so quick to judge this new processor.
11-30-2005, 05:22 PM
saaya

Quote:

Originally Posted by Bar81

Because it's another in a long line of retarded "professional" reviews. You can tell right from the open that the reviewer has never actually used a Dothan. When it comes to gaming Dothan is still king clock for clock and an absolute knockout when it comes to performance per watt.

I'm really disgusted at Intel once again, for no reason, switching pinouts on these chips forcing everyone to use new chipsets and killing any upgrade path for P-M users.

Either way, I don't see any reason to get yonah over Dothan if you already have a Dothan.

its not only a few pins they changed, yonah needs a new chipset. i think it works with the 945/955/975 chipsets but yeah, they should have come up with a way to make it compatible with existing plattforms :/

Quote:

Originally Posted by kl0012

I do not understand why many so are disappointed.
Doom3:
Dothan(2,13GHz): 95.5 fps
Yonah (2GHz): 93.8 fps
Dothan has 7% higher frequency and only 2% faster than Yonah in Doom3. It points that Yonah will beat Dothan in games.
It also drive less power than Dothan:
http://www.tomshardware.com/cpu/2005...ntium4-19.html
http://techreport.com/reviews/2005q3...9/index.x?pg=7

you cant compare those numbers, those are different systems, and no, yonah doesnt consume less power than dothan! it consumes around 2x as much under load.

and the dothan is clocked 7% higher in that article but has a 25% slower fsb!

Quote:

Originally Posted by B5I8

I think it's too early to tell how Yonah will be. I mean we're talking engineering sample processors on sample boards with bare beta drivers. Don't be so quick to judge this new processor.

this is xs :D
we see beta boards and engineering samples all day, and there is usually no real difference between those first samples that get sent to reviewers and the final products.

what im confused about i wesleys article on anandtech is this:

http://www.anandtech.com/cpuchipsets...spx?i=2627&p=7
http://img220.imageshack.us/img220/2365/100767am.jpg

http://www.anandtech.com/cpuchipsets...spx?i=2627&p=3
http://img220.imageshack.us/img220/9190/huh4de.jpg

so does yonah score 95fps or 93fps? whats the difference between the two numbers?

Quote:

Intel still needs to improve their video encoding and gaming performance, but it looks like we may have to wait for Conroe and Merom for that.

wha? :confused:
its performing almost as good as as the x2 dual cores, now show me a laptop with an x2! :D
theres one from alienware, but thats it, and its not really a laptop, x2 cpus run too hot to be used i laptops...

for a laptop cpu thie performence is really nice!
but then again the power consumption of almost double of dothan is not good... :/

for a desktop cpu the performence isnt good enough and below what i expected actually... it looks like a single core yonah is slower than a dothan at the same clockspeed and fsb... and dual core doesnt really help that much yet.

if you ask me yonah is a good mobile chip and a good desktop chip... but its neither a great cpu for either segment of the market...
its not really a mobile chip, nor is ait a desktop chip...

they wanted a new mobile chip but they focussed more on performence than on power consumption, while dothan was all about power consumption...
this made perfect sense imo. you dont realy need a lot of cpu power in your laptop, what people really want is long battery life times.

in that situation yonah is a step back... but we havent see any yonah laptop reviews yet, who knows how its improved speedstep helps to lower power consumption...

so assuming yonahs speed step works well and the battery life stays the same as with dothan if the cpu is not under heavy load, then id say its a great laptop chip. if your not using a lot of cpu power once core shuts down and you basically have a dothan with slightly improved architecture, and if you need a lot of performence theres a second core to power on and give you a nice boost in multithreadded apps :)

but then again, the biggest advantage of dual core is having two cores all the time to work faster in windows, but if both cores run then yonah will consume more power than dothan = less battery life

speedstep will make aor break yonah.
if its unbalanced and the second core is active too mch and sucks extra power without really giving a big performence boost, a boost that isnt really needed on a laptop most of the time anyways, then it will make it a worse chip than dothan. if its implemented nicely and you get dothan power consumption or less with light system usage on youir laptop and the second core jumps o when you really need it, then its a great update for the pentium m family :D
11-30-2005, 05:31 PM
saaya

damn i wrote a lot :D
sorry :bows:
12-01-2005, 12:54 AM
Carfax

Quote:

Originally Posted by DilTech

They're comparing a 133 mhz fsb dothan to a 166mhz fsb yonah. Considering the 133mhz dothan was pretty limited by it's fsb, It's a pretty bad sign.

Yes, but Yonah's FSB is being shared between TWO cores, unlike Dothan's one.

So all else being equal, Yonah has even less bandwidth than Dothan.

*Edit* I just saw over at Aces that someone posted that Yonah was benched using a 533 FSB, because it is a test chip, and not a launch chip.

Launch versions will have 667 FSB.
12-01-2005, 02:57 AM
R101

Quote:

Originally Posted by saaya

for a laptop cpu thie performence is really nice!
but then again the power consumption of almost double of dothan is not good... :/

I just don't share your view - if you look at Anand's power consumption test, it shows only 16W difference between idle and full load. No way it could use 30+W while idling :nono: .
12-01-2005, 03:05 AM
skane

it could be that it consumes that, i dont remember where, but i read that eist isnt working in that test.
12-01-2005, 03:23 AM
kl0012

skane

Quote:

it could be that it consumes that, i dont remember where, but i read that eist isnt working in that test.

Here is the same difference for Dothan (16-18W) when EIST and power saving wasn't used:
http://techreport.com/reviews/2005q3...9/index.x?pg=7
http://www.gamepc.com/labs/view_cont...ookie%5Ftest=1

It means that Dothan use about 10W power on idle which is typical power usage
in C1-state with max. Vcc attached according to the intel manual (30218907.pdf)
Now you mean that Yonah use 30W+ in C1 state? Impossible...
12-01-2005, 03:38 AM
skane

still, you cant compare anands data with that of others, different systems, different methods.
12-01-2005, 05:01 AM
saaya

Quote:

Originally Posted by Carfax

Yes, but Yonah's FSB is being shared between TWO cores, unlike Dothan's one.

So all else being equal, Yonah has even less bandwidth than Dothan.

*Edit* I just saw over at Aces that someone posted that Yonah was benched using a 533 FSB, because it is a test chip, and not a launch chip.

Launch versions will have 667 FSB.

the engineering samples are 166fsb as well, believe me ;)
unles he has a really really old one... and i dont think so :D
if wesles benched 133fsb then only because his board couldt do more or didnt support a higher setting at that time.

Quote:

Originally Posted by R101

I just don't share your view - if you look at Anand's power consumption test, it shows only 16W difference between idle and full load. No way it could use 30+W while idling :nono: .

actually this could very well be the case if eist isnt working, and this is an engineering sample desktop board, so i wouldnt be surprised at all if eist wasnt working yet.

Quote:

Originally Posted by kl0012

skane

Here is the same difference for Dothan (16-18W) when EIST and power saving wasn't used:
http://techreport.com/reviews/2005q3...9/index.x?pg=7
http://www.gamepc.com/labs/view_cont...ookie%5Ftest=1

It means that Dothan use about 10W power on idle which is typical power usage
in C1-state with max. Vcc attached according to the intel manual (30218907.pdf)
Now you mean that Yonah use 30W+ in C1 state? Impossible...

unlikely, true, but not impossible...

1. its a different system, 2nd keep in mind that dothan is 70% cache and only 30% logic, and cache consumes very little to no power in c1.
yonah is 2 cores and its 70% logic i think, so there are a LOT more transistors and if eist isnt working corerectly yet the second core is not shut down in idle, wich meansyou have all the logic enabled and consuming power.

and for those who still dont believe yonah wil consume up to 50W and even more:
http://www.computerbase.de/news/hard...-schema_yonah/

and here they report how the 1.5ghz yonah system at idf consumes around 31W under load:
http://www.computerbase.de/news/hard...ossaman_yonah/

and about ocing, according to sam at x86-secret who has always been 100% reliable in my expirience, the first yonah chips only overclock to around 2.5ghz
http://www.x86-secret.com/?option=newsd&nid=907

more proof for 50W power consumption:
http://www.hkepc.com/bbs/viewthread.php?tid=488561

i was wrong about the cache latency thing btw, its one large block of cache, so the cache latency wil be the same for single and dual core yonahs...
and even yonah celerons with only 1mb cache will have the same increases cache latency :/

so from dothan to yonah we get:
around the same per core performence
double the cores, but they are limited by bad multithreadding implementations and a low fsb, only 83mhz fsb per core!
2x the power consumption under ful load
same power consumption in idle and light load (?)

and again, the last thing will make or break yonah, if the power consumption in idle and light load is the same as dothan or even less, then it will be a great mobile cpu.

but again, in my opinion theres a bigger demand for cpus of the same speed as todays cpus but with lower power consumption for laptops than cpus with the same or higher power consumption and higher performence...
i mean what would you prefer? a laptop you can ripp dvds with quickly but the battery only lasts 4 hours or a laptop wich isnt that fast but has a battery life of6 hours or more?

extra cpu power is always good, but i think what the mobile market wants is longer battery life... if you can increase the cpu resources without shortening the battery lifes or even increasing them, perfect, but a compromise for more pwerformence that isnt really needed at the cost of shorter batter life is a bad decision imo.

thats what yonah was originally supposed to be all about afaik, less power consumption at the same or increased performence. the yonah we are seeing now is a long distance runner on anabolica on its way to become a heavy weight lifter :D
12-01-2005, 09:54 AM
skane

Quote:

Originally Posted by saaya

thats what yonah was originally supposed to be all about afaik, less power consumption at the same or increased performence. the yonah we are seeing now is a long distance runner on anabolica on its way to become a heavy weight lifter :D

:confused: what is that supposed to mean, i cant figure out what the weight lifter has got to do with it.
12-01-2005, 09:59 AM
saaya

well weight lifters like fast cpus and always have the latest hardwae, plus they eat a lot wich means they have a lot of money to spend for a quad core yonah laptop :D
12-01-2005, 10:08 AM
Fred_Pohl

I fail to understand how anyone can be disappointed with Yonah's performance in this very early review. IMO Yonah did very well considering that it was crippled by a low FSB (533 or 667) that can barely feed one core, let alone two cores. The low FSB and subsequent lack of sufficient memory bandwidth is what kills Yonah's performance. Especially in the encoding tests. Also, the increased L2 latency, while hurting performance a little, should permit much higher clock speeds than Dothan can run with it's lower latency L2.

As for Yonah consuming more power than Dothan, should this surprise anyone? Did anyone expect two cores with double the cache of Dothan not to use more power? Instead of comparing a bandwidth-starved dual-core Yonah with a single-core Dothan, we should compare it to the X2 and note that even crippled as it is, it performs just as well clock for clock and consumes far less power. Keep in mind that Yonah is a mobile CPU that matches the best desktop CPU available, clock for clock. Just imagine how Yonah would perform if it wasn't bandwidth starved.
12-01-2005, 10:27 AM
skane

im not really disappointed by the power consumption, the thing i am disappointed at is that i expected it to do better than an x2 at the same clock.
12-01-2005, 10:42 AM
saaya

Quote:

Originally Posted by Fred_Pohl

I fail to understand how anyone can be disappointed with Yonah's performance in this very early review. IMO Yonah did very well considering that it was crippled by a low FSB (533 or 667) that can barely feed one core, let alone two cores. The low FSB and subsequent lack of sufficient memory bandwidth is what kills Yonah's performance. Especially in the encoding tests. Also, the increased L2 latency, while hurting performance a little, should permit much higher clock speeds than Dothan can run with it's lower latency L2.

As for Yonah consuming more power than Dothan, should this surprise anyone? Did anyone expect two cores with double the cache of Dothan not to use more power? Instead of comparing a bandwidth-starved dual-core Yonah with a single-core Dothan, we should compare it to the X2 and note that even crippled as it is, it performs just as well clock for clock and consumes far less power. Keep in mind that Yonah is a mobile CPU that matches the best desktop CPU available, clock for clock. Just imagine how Yonah would perform if it wasn't bandwidth starved.

its not 2times the cache of dothan, its half the cache per core compared to dothan.
and 166fsb quad pumped isnt really THAT much of a limiting factor i think...
but im not sure about this one...
and the cahce latency was not increase on purpose to allow higher clockspeeds, that wasnt the target to begin with, yonah was supposed to clock at the same speeds as dothan.
its a higher latency because the cache is shared i think...
id assume that both cores are hooked up to some kind of a switch/request interface wich controls and coordinates the accesses to both cache and to the system, thats where the increased cache latency comes from i think.

and comparing yonah to an x2 is interesting as it lets us guess how well conroe and merom will perform, but yonah is a mobile cpu, so we should compare it to other mobile cpus imo.

Quote:

Originally Posted by skane

im not really disappointed by the power consumption, the thing i am disappointed at is that i expected it to do better than an x2 at the same clock.

maybe it does with a higher fsb?
12-01-2005, 11:31 AM
railer

From Intel dev's

Quote:

A single core Dothan would have required 4 clock cycles to perform a read-out when hitting a L2 cache line. With Yonah's Bandwidth Adaptation Buffer, it is possible to perform it in 2 clock cycles. As such, the available bandwidth stays the same, but access time is shortened. Yonah will also feature optimized pre-fetchers and a deeper write output buffer to further increase cache performance

I realy dont know how this feature was utilized in that review. Lets wait until we get something more official . I'm realy intrested how high it will clock and what damage will it do with 900+fsb :D
12-01-2005, 11:38 AM
saaya

i think the yonah design was changed, it was initially supposed to be a dothan successor but was then changed to perform better as an alternative for netburst on the desktop market... so maybe the cache access was changed to allow more bandwidth at a higher latency now and not at the same bandwidth as dothan with lowe rlöatency as planned originally?
12-02-2005, 02:36 AM
TL1000S

Quote:

Originally Posted by railer

From Intel dev's

I realy dont know how this feature was utilized in that review. Lets wait until we get something more official . I'm realy intrested how high it will clock and what damage will it do with 900+fsb :D

Second that..
If "yonah" shows the same behaviour as "dothans" they will most likely easily take 200+ FSB at default VCore.
Then I think they will "destroy" the competition..
Looking forward to the time some of the "extreme" members here gets their hands on a "Yonah".
I think records will fall.. :p:
12-02-2005, 05:38 AM
saaya

Quote:

Originally Posted by TL1000S

Second that..
If "yonah" shows the same behaviour as "dothans" they will most likely easily take 200+ FSB at default VCore.

if...
did prescott show the same behaviour as northwood?
did the a64 show the same behaviour as an athlonxp?
all we can do is guess :D

Quote:

Originally Posted by TL1000S

Looking forward to the time some of the "extreme" members here gets their hands on a "Yonah".
I think records will fall.. :p:

;)
12-04-2005, 02:41 AM
cirthix

saaya, check this out
http://www.theinquirer.net/?article=26904
12-04-2005, 10:03 AM
saaya

thats old news, there have been reports about 50W+ of the extreme edition based on yonah for months :P
check xbitlabs, they reported about this 2 or 3 times during the last months :)

if you ask me intel is running through the door with their head again instead of opening it :D

you have two cores, right? with the latest drivers a dual core cpu performs 10% better than a single core cpu in games, at best!
but having two cores means 2x the power consumption.
so instead of running both cores at a higher speed wouldnt it make 1000 times more sense to shut down the second core and increase the clockspeed of the other core?
that way you would get the same or more performence for less watts... and for laptops watt is king and performence comes second...

well lets wait and see, maybe intel actually does something like this.
12-04-2005, 04:39 PM
accord99

Quote:

Originally Posted by saaya

but having two cores means 2x the power consumption.

Only if they're the same core. Yonah is also on 65nm and the difference between idle power and full load in the anandtech test was only 16w. As long as the leakage isn't bad (and its reputed to be even better than Dothan) than you are looking at 20w chip for the 2GHz Yonah. Which would match the reports of dual, dualcore 2GHz Sossaman system power usage of 80-90w.
12-05-2005, 01:04 AM
cirthix

Quote:

Originally Posted by saaya

thats old news, there have been reports about 50W+ of the extreme edition based on yonah for months :P
check xbitlabs, they reported about this 2 or 3 times during the last months :)

if you ask me intel is running through the door with their head again instead of opening it :D

you have two cores, right? with the latest drivers a dual core cpu performs 10% better than a single core cpu in games, at best!
but having two cores means 2x the power consumption.
so instead of running both cores at a higher speed wouldnt it make 1000 times more sense to shut down the second core and increase the clockspeed of the other core?
that way you would get the same or more performence for less watts... and for laptops watt is king and performence comes second...

well lets wait and see, maybe intel actually does something like this.

From what i know, it does shut down one of the two cores for power savings. i dont remember fully, but anandtech's idf coverage comes to mind as the source.
12-05-2005, 03:12 AM
saaya

Quote:

Originally Posted by accord99

Only if they're the same core. Yonah is also on 65nm and the difference between idle power and full load in the anandtech test was only 16w. As long as the leakage isn't bad (and its reputed to be even better than Dothan) than you are looking at 20w chip for the 2GHz Yonah. Which would match the reports of dual, dualcore 2GHz Sossaman system power usage of 80-90w.

what do you mean with only if they are the same core?
the difference between idle and load was low because the desktop board never went as low in idle as a laptop will go ;)
20w for the 2ghz yonah?
where did you read that?
did you check the links i posted?
the 2ghz yonah was oprigianlly called t1500 and is now called t2500, the first number is now probably going to indicate the number of cores.
its a T chip, there is no yonah chip at 2ghz in thwe L group.
T= up to 50W and L is up to 25W

now where did a dual dual core sossaman system consume only 80W under load?
the only thing i saw was this:
http://img220.imageshack.us/img220/9666/108hy.jpg

and thats a dual dual core sossaman at 1.5ghz for each core!
and its consuming 95W! so thats 47W for each sosaman dual core chip running at only 1.5ghz!
a 4 way dothan system, if possible, would allow you to clock each core to around 2.4ghz if 95W is the limit as an example.
so we have almost 10ghz dothan power vs 6ghz yonah power for the same power consumption. and the preview at anandtech has shown us that yonah is actually quite a bit slower than dothan if you think of the per core performence! so 1 dothan ghz is actually worth more than a yonah ghz.
now where is sossaman/yonah a better performence per watt cpu than dothan? we still have to see final results of final hardware though... but as for now it doesnt look like the uber chip i thougt it would be...

if you look at the graphs on computerbase you can see that you probably need 4! yonah cores (2 dual core chips) clocked to ~1800mhz each to beat a single (dual core) 2.4ghz X2 4800+ ... so power consumption doesnt look that nice (for a mobile chip, for a desktop chip its very nice!) and performence doesnt look that nice either, more than you need on a laptop and not enough on the desktop to beat the X2s...

Quote:

Originally Posted by cirthix

From what i know, it does shut down one of the two cores for power savings. i dont remember fully, but anandtech's idf coverage comes to mind as the source.

i never said it wouldnt.
what i said was it should shut down one core in games and other single threadded apps! not in idle. :)

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