TechReport on Barcelona

Printable View

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

09-10-2007, 01:54 AM
Nedjo

Quote:

Originally Posted by metro.cl

Someguy form Anand i believe said those numbers.

About OC on Anand they have a cpu overclocked to 2.5 and they need 1.5v for that.

They've received 2.5 GHz! How do you think they could have O/C this CPU on the server mobo?
BTW CPU-Z can't detect good voltage of Barcelona...
09-10-2007, 02:02 AM
freeloader

I was expecting more. However, AMD did make a great showing in the power consumption department.
09-10-2007, 02:07 AM
LordEC911

Quote:

Originally Posted by gallag

?is this even true? what info did you base this on?

2.5 Barc has a 120w TDP, 100w APC.
Stock QX6700/6850 have a 130w TDP.
A stock phase supposedly can't handle 200w of heat, which a decently +3.5ghz quad puts out.
09-10-2007, 02:24 AM
Hornet331

Quote:

Originally Posted by LordEC911

2.5 Barc has a 120w TDP, 100w APC.
Stock QX6700/6850 have a 130w TDP.
A stock phase supposedly can't handle 200w of heat, which a decently +3.5ghz quad puts out.

intel quads might have 130W TDP but far away form actually consuming it. They are more in the range of 90W.
09-10-2007, 02:29 AM
Shintai

http://processorfinder.intel.com/det...px?sSpec=SLACQ

G0 Q6700 is 95W aswell. And I dont think 1 stepping changed that with 35W ;)
09-10-2007, 02:45 AM
zakelwe

The server benches look good, the desktop less so. It will be interesting to see how well it improves with motherboards and memory speeds designed for the Phenom in Q4.

I guess the happiest people will be the drop in replacement server guys who have a cheap way of boosting their current K8 based systems with little hassle. It's a no brainer for them. For the home user it still remains to be seen whether it is a good upgrade or not. It's not a disaster but it's not the second coming either.

For overclockers things look decidedly iffy at this, admittedly early, stage. It's going to have to compete against something that is likely to overclock to 4GHz+ come November and have the potential to be pretty cheap.

I hope this is not another HD2900XT, ie a good product but not up to everyones expectations cultured over the extended developement phase.

Regards

Andy
09-10-2007, 02:54 AM
R101

The only question is why would someone buy 1st revision chips NOW when in 2-3 months we'll see the good ones coming out?
09-10-2007, 03:49 AM
duploxxx

Quote:

Originally Posted by R101

The only question is why would someone buy 1st revision chips NOW when in 2-3 months we'll see the good ones coming out?

that's the whole point that many people don't understand... in serverland there is always a rule to buy cpu n-x you never buy the best performer.

in other words for AMD it is enough to have 2.0-2.3-2.5 server parts, the amount of sold highest cpu's is minimal. not like the desktop parts where there are plenty of idiots buying the top bin.

even if a new stepping will arrive it won't be much faster clock/clock. it will have other main improvements like power consumption and improved clock speed.

Quote:

Originally Posted by Shintai

Reality bites again. No miracle chip. How...unexpected...

@shintai when are you going to stop all those idiots comments each 5 minutes. Yes it is due to g0 that there is a huge drop in powerconsumption against B3. What the hell is a power rating of a desktop chip doing on a newsforum of an AMD server chip, stay happy and post only on the INTEL fanboy subforum.

and for the chip itself, it is more than decent and a nice competitor for current offerings when the speedbump is there in a few weeks, the only problem is that it is a few months late, let's see what you're favorite vendor will bring after this 45nm die shrink. Barcelona has a lot of room to improve. clock/cache/memory/shrink/split powerplain not to forget the HTT3 that is very interesting on a +2 socket system.
09-10-2007, 03:53 AM
accord99

Quote:

Originally Posted by duploxxx

that's the whole point that many people don't understand... in serverland there is always a rule to buy cpu n-x you never buy the best performer.

AMD's problem is Intel's cpu n-x is faster than AMD's best performer.
09-10-2007, 04:03 AM
naokaji

Quote:

Originally Posted by accord99

AMD's problem is Intel's cpu n-x is faster than AMD's best performer.

amd's advantage is the price....

comapny x has x number of k8 opterons... and its x $ cheaper to buy k10's and to put them into the socket they removed the k8 from compared to buying intel cpus + mobo's + ram
09-10-2007, 04:04 AM
accord99

Quote:

Originally Posted by naokaji

comapny x has x number of k8 opterons... and its x $ cheaper to buy k10's and to put them into the socket they removed the k8 from compared to buying intel cpus + mobo's + ram

Unless the K8 opterons are from the S940 era. Socket F sales didn't do too well vs Woodcrest and especially Clovertown.
09-10-2007, 04:13 AM
Jacky

Quote:

Originally Posted by naokaji

amd's DISadvantage is the price....

I corrected the little mistake.. AMD *really* needs higher ASPs
It seems Intel will rule the desktop/Notebook sector for quite a while (barc clocks slower than conroe, only 15% ipc improvements over K8, penryn around the corner) and it seems that they will put up a real fight with fsb 1600mhz penryn against barc. but yeah barc is quite strong when it comes to server applications.
09-10-2007, 04:17 AM
LowRun

Quote:

Originally Posted by metro.cl

Someguy form Anand i believe said those numbers.

About OC on Anand they have a cpu overclocked to 2.5 and they need 1.5v for that.

My C2D E6850 does 4.0GHz at same volts so overclock doesnt seem good for AMD.

What's the point in comparing a dual core and a quad?
09-10-2007, 04:39 AM
BeardyMan

SmartAss

face it K8 at 3Ghz wanst going to happen neither, LOL instead they nailed guys like you with releasing a 3.2ghz 90nm cpu at stock..

in other words why all teh guessing and hating?:rolleyes:
09-10-2007, 04:47 AM
mAJORD

I almost mistook you for shintai then..

MR Smartarse.. some valid points, but a lot of opinion

Forget DDR, when comparing the same RAM tech, K8 was very dependent on bandwidth for certain work
http://www.anandtech.com/memory/showdoc.aspx?i=2800&p=7

You can be pessimistic about seeing a higher clocked K10 before the years end, but it seems things are looking not so bad as it is.. The impressive power consumption when compared to K8 90nm say it all.
09-10-2007, 04:50 AM
red

http://www.techreport.com/r.x/optero...power-peak.gif
Damn, 25% faster clocked 2360 peaks 62W more!
09-10-2007, 04:51 AM
Sparky

3GHz Barch not here yet? Oh well so what, it'll come eventually. I mean did everyone whine and complain when C2D was released below 3GHz? No, they just OC'd them to get there ;)

Now this is just the server arena 1st so we won't be seeing much of an OC yet to tell if it is the same case or not. Once Phenom is released though then we'll be able to clock these chips.
09-10-2007, 05:04 AM
mAJORD

on a slightly different note...

Another review:

http://www.tecchannel.de/index.cfm?p...pk=1729224&p=1

I Can't read german so that entire review just makes my head hurt bigtime.. (plus it's bed time). I can read the graphs though, Anyone (or many of you) who CAN read german care to take a looksie, The results of 2ghz barc vs 2.33ghz Xeon look more impressive than the other 2 reports, interested in why it may be.

goodnight
09-10-2007, 05:13 AM
mAJORD

Quote:

Originally Posted by MR_SmartAss

This is true for the high clocked K8's and for certain work. But if you test K8 with same relative(in ns, not in cycles) RAM latency, you will find out that in most cases the bandwidth doesn't improve the performance at all. I was experimenting with my 4200+ and with DDR2-667 CL4 4-4-12 and DDR2-800 CL5 5-5-15. I got almost the same 3D Mark, PC Mark and Super Pi 32M results.

BTW, check this article:
http://www.anandtech.com/cpuchipsets...spx?i=2741&p=5
Why the K8 with DDR-400 2-2-2 is outperforming the same clocked K8 with DDR2-667 3-3-3 in most cases?

I am not insisting latency is not important either, They both have an impact, but bandwidth is definatly more important for more workloads in my experience.. you're comparing 2 different mem controllers there, it's just invalid.

Super pi is no exception I've found. Read the article agian you'll see the latency settings are slightly higher at the higher frequency, yet performance is still scaling in a big way compared to core.. that's undeniable, and that's what the point is. If core 2 also scaled with faster memory in the same way, a suggestion that phenom will perform better with A grade ram would be flawed on that basis alone.

Now the impact of the L3 on K10 on the other hand, we'll have to see..

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