Sign in with
Sign up | Sign in
Your question

Data on Barcelona performance

Last response: in CPUs
Share
July 26, 2007 6:22:45 PM

Finally some data on barcelona performance (http://www.itweek.co.uk/vnunet/news/2195086/amd-silence...). Not much but still. Combined with this article - http://www.pcpro.co.uk/news/121139/amd-makes-quadcore-p... e-performance-figures.html it is clear that Barcelona is not going to beat even current Intel Xeons.

July 26, 2007 6:56:54 PM

nilo said:
Finally some data on barcelona performance (http://www.itweek.co.uk/vnunet/news/2195086/amd-silence...). Not much but still. Combined with this article - http://www.pcpro.co.uk/news/121139/amd-makes-quadcore-p... e-performance-figures.html it is clear that Barcelona is not going to beat even current Intel Xeons.

WTF are you talking about? There's nothing of that sort mentioned. They say that their Launch speed starts @ 2Ghz and is able to compete against a 2.33Ghz Intel, could mean that they could still upscale and compete against a 2.4Ghz Core Processor without any problems. They never said anything about their higher speeds, which every one is sure is imminent soon as Barcelonas are launched.

I suggest you delete this post. It's nothing new and most things that your typing is pretty much not related to what is in the link, save yourself the B/S accreditation.
July 26, 2007 7:22:08 PM

sorry cant find any benches there
Related resources
July 26, 2007 7:31:07 PM

YO_KID37 said:
I suggest you delete this post. .


Thank you for suggestion. :) 

YO_KID37 said:
It's nothing new and most things that your typing is pretty much not related to what is in the link, save yourself the B/S accreditation.


Both articles are new. In fact they were published today after AMD Analyst Meeting.

Here is a new piece of information: "At AMD's Technology Analyst Day at its Silicon Valley headquarters, the chipmaker boasted that its processor outperformed the an Intel Xeon X5355 quad core processor by 25 per cent in the SPECfp 2006 benchmark, and claimed a 30 per cent advantage on the performance per Watt metric."

Unlike AMD's previouos claims that were based on "simulated" Barcelona (2.66) this is comparison for real chips. Too bad they do not say what was Barcelona speed but we could guess it was for 2.0 (what they are going to release soon).

The second news is:

Claims that the absolute performance of the new CPUs is going to outstrip Intel's highest-frequency parts were, however, conspicuous by their absence. Worldwide channel market development manager Felipe Payet was at pains to stress that "you need to compare apples to apples" and performance should be assessed in the context of cost and power consumption.

If you read carefully, you'd understand that AMD admits that 2.0 Barcelona will not beat in performance even currenly available Xeon cheips (not to mention Penryn).


The third item worth mentioning is this:

Opteron product manager Steve Demski deflected talk that 2GHz seemed low in the light of Intel's higher-frequency parts: "We'll put our 2GHz part up against their 2.33 and be happy with that all day long," he said.

What they ar esaying is that Barcelona 2.0GHz has performance similar to Xeon 2.33GHz which means that by the end of the year when they promise to have 2.3 Barcelona it will still be underperforming currently available Xeons.






July 26, 2007 7:43:52 PM

More details from (http://blog.pcmag.com/blogs/miller/archive/2007/07/26/2...):

A MD showed a comparison between its 2.0GHz quad-core Barcelona chip and an existing intel Xeon 5345 “Clovertown” chip running at 2.33 GHz. Technically the AMD chip is a 95watt chip, while the Intel chip is an 80 watt chip, but they were really are both in the same energy framework, since the two firms measure energy use differently. In AMD’s demo running the SPEC floating point benchmark, the Barcelona chip was scoring 69.5, while the Intel chip was at 54. So AMD claims a 25 percent performance advantage; and when looking at actual power draw; 30 percent better at performance per watt.
a c 124 à CPUs
a b å Intel
July 26, 2007 7:45:30 PM

All speculation, but it looks weasely to me. Yes, companies are becoming concerned about power consumption in the data center, but performance still comes first; performance per watt won't count for squat if the performance is low.

Wait and see...
July 26, 2007 7:46:03 PM

Quote:
Too bad they do not say what was Barcelona speed but we could guess it was for 2.0 (what they are going to release soon).


No, that cannot possibly be the case. They said the 2.0Ghz Barcelona is good against a 2.33Ghz Intel CPU; how could it be that it's 25% faster than the Xeon 5355 (2.66Ghz) then?...

Nothing really new there...
July 26, 2007 8:14:35 PM

Mephistopheles said:
No, that cannot possibly be the case. They said the 2.0Ghz Barcelona is good against a 2.33Ghz Intel CPU; how could it be that it's 25% faster than the Xeon 5355 (2.66Ghz) then?...


Another article confirms that it was 2.0 Barcelona against 2.33 Xeon (Xeon 5345) Link: http://blog.pcmag.com/blogs/miller/archive/2007/07/26/2...

kelceylehrich said:
Nothing really new there...


While the numbers are not new, there is more specifics here. More importantly previous reports were about "simulated" performance or pre-production samples. This time AMD tells the numbers as they will look when Barcelona is actually released. No more speculations: good FP performance (but nothing special), integer performance and most other meaningful benchmarks will come below Xeon 5355.
July 26, 2007 8:36:49 PM

Onus said:
All speculation, but it looks weasely to me. Yes, companies are becoming concerned about power consumption in the data center, but performance still comes first; performance per watt won't count for squat if the performance is low.

Wait and see...

The FB-DIMMS and INTEL chip sets use a lot more power then ECC ram and amd / nforce pro chip sets.
July 26, 2007 9:49:21 PM

Joe_The_Dragon said:
The FB-DIMMS and INTEL chip sets use a lot more power then ECC ram and amd / nforce pro chip sets.

Memory capacity is much greater than what AMD can offer.
July 26, 2007 11:32:50 PM

This statement seems very important to me:

Quote:
Opteron product manager Steve Demski deflected talk that 2GHz seemed low in the light of Intel's higher-frequency parts: "We'll put our 2GHz part up against their 2.33 and be happy with that all day long," he said.


Yes, you can be pretty sure a 2.0GHz Barcy won't destroy a current gen 2.66 GHz Xeon quad, or the comparison would be to that.

Incidentally, Penryn launch frequencies are reported to be:

X5460 3.16 GHz 120W 12MB $1,172
E5450 3.00 GHz 80W 12MB $851
E5440 2.83 GHz 80W 12MB $690
E5430 2.66 GHz 80W 12MB $455
E5420 2.50 GHz 80W 12MB $316
E5410 2.33 GHz 80W 12MB $256
E5405 2.xx GHz 80W 12MB $209

These chips will be available some time in 4Q07. Depends on how Barcy does, but most sources expect an early launch. Note the pricing. If Barcy at 2.0GHz is between E5410 and E5420 in terms of performance, then in order to remain competitively priced it'll have cost less than $300. For a top end part? That's awful.

The truth is, AMD is gonna have to get Barcy up to speed very quickly, or they are going to get bulldozed by cheap, higher clocked midrange Xeons.

That is, unless AMD can convince corporations that they should pay premium prices for energy efficient chips.
July 27, 2007 2:07:48 AM

Back when A64s were the top dog, AMD used to argue against Intel by not only having better performance but also a much lower power consumption.

I guess they must have gotten used to it: right now, they're still singing to the same "low power" song, but they don't have the performance leadership anymore. Now it's just plain silly.

Also, I'm still not really convinced that it will be an easy win for Agena when looking at efficiency, i.e. performance per watt. They might get it anyway by using low-power specially-binned CPUs without us ever knowing that they did so. They would then have horrible yields, but they would be able to claim efficiency leadership.

We all have to bear in mind that Intel isn't making a real effort for those C2Ds to dissipate only 65W, while AMD doesn't seem to be having good yields at those speeds as far as we know.

It's all speculation anyway though...
July 27, 2007 2:26:11 AM

CEO's care about performance per watt. They and other people outside of the IT departments have alot of say and sway when it comes to purchasing. Just from reading what Ive read here, a 2.0 Barcy is more likely to fit in between a 5420 and a 5410. That leaves it close to 300 USD. And IF the 3.0 mark is real, and it scales evenly as speeds increase, a 3.0 Barcelona will be better than a 3.5 xeon. That sounds competitive at the very least
July 27, 2007 2:46:50 AM

JAYDEEJOHN said:
And IF the 3.0 mark is real...


There is absolutely no indication of that. AMD only promises 2.3Ghz by year end. They did show 3.0Ghz PC today but it was Phenom not Barcelona and they did stress that it was technology demonstration and not a product introduction.
July 27, 2007 2:49:28 AM

Quote:
And IF the 3.0 mark is real, and it scales evenly as speeds increase, a 3.0 Barcelona will be better than a 3.5 xeon. That sounds competitive at the very least


...And that's a big if: if the 3.0Ghz Barcelona is out in volume, that would be competitive. We're not seeing indication of that happening in 2007, while a >3Ghz Xeon is there. So at least in the next 6 months, Barcelona will probably get kicked around very nicely by Yorkfield...
July 27, 2007 2:55:31 AM

Mephistopheles said:
Quote:
And IF the 3.0 mark is real, and it scales evenly as speeds increase, a 3.0 Barcelona will be better than a 3.5 xeon. That sounds competitive at the very least


...And that's a big if: if the 3.0Ghz Barcelona is out in volume, that would be competitive. We're not seeing indication of that happening in 2007, while a >3Ghz Xeon is there. So at least in the next 6 months, Barcelona will probably get kicked around very nicely by Yorkfield...

Just plain DELETED
a c 99 à CPUs
July 27, 2007 4:03:40 AM

JPack said:
Memory capacity is much greater than what AMD can offer.


FB-DIMM can have 8 modules strung up per channel, so the 4-channel 5000 series chipset can support 32 modules. That works out to be 128 GB using the largest DDR2 DIMMS available (4 GB.) Opterons can support up to four modules per processor, so AMD needs eight sockets to get 32 modules. The other advantages to FB-DIMMs are that there are fewer traces to route per channel and that the AMB handles the data, making whether DDR2 or DDR3 memory is used irrelevant.

However, the tradeoff only wins if there needs to be a ton of memory for a few CPUs. Otherwise, FB-DIMMs are hotter, more expensive, and have much higher latency and lower performance than normal ECC buffered DDR2 that's used in the Opterons. Generally Xeon DP boards have 4, 6, or 8 FB-DIMM slots, so the advantage is largely lost.
a c 99 à CPUs
July 27, 2007 4:19:08 AM

Mephistopheles said:

We all have to bear in mind that Intel isn't making a real effort for those C2Ds to dissipate only 65W, while AMD doesn't seem to be having good yields at those speeds as far as we know.

It's all speculation anyway though...


Many C2Ds probably could be undervolted a bit and some of the lower-speed ones certainly don't dissipate 65 watts. But then neither do the lower-clocked Brisbanes, and the non-BE chips aren't specifically aimed at a certain power consumption figure.

The one thing that I think is glaringly absent in AMD's low-power mantra is a low-power notebook chip. Intel has a full range of standard, low, and ultra-low-voltage dual-core CPUs but AMD just has the standard-voltage Athlon 64 X2s. AMD used to have the Turion 64 MT series, and while it wasn't really a low-TDP chip at 25 watts, it at least was somewhat lower than the ML was and was lower than any Turion 64 X2. It is true that only a small portion of the notebook market uses anything but standard chips, but that is one area that is completely owned by Intel.
July 27, 2007 7:33:17 AM

MU_Engineer said:
The one thing that I think is glaringly absent in AMD's low-power mantra is a low-power notebook chip.


I just find it ironic that AMD can emphasize performance per watt at Analyst Day, while at the same time demoing a machine sporting three (count em, 3) Radeon HD 2900 XTs. :lol: 

Those video cards prolly draw 1000 watts (if not more) between them. It's ludicrous to wear the mantle of environmental responsibility while at the same time pushing stuff like trip HD 2900 XTs, let me tell you. How they are able to do it with a straight face is beyond me.
July 27, 2007 1:00:45 PM

Quote:
How they are able to do it with a straight face is beyond me.


Never underestimate their powers of misdirection. They'll confuse you until you think they've got something!
July 27, 2007 3:06:28 PM

easyg said:
I just find it ironic that AMD can emphasize performance per watt at Analyst Day, while at the same time demoing a machine sporting three (count em, 3) Radeon HD 2900 XTs. :lol: 

Those video cards prolly draw 1000 watts (if not more) between them. It's ludicrous to wear the mantle of environmental responsibility while at the same time pushing stuff like trip HD 2900 XTs, let me tell you. How they are able to do it with a straight face is beyond me.

Its ironic indeed, but do you understand these products are for different markets? They do have low power desktop/server chips, and more are coming, but it doesnt mean enthusiasts are very concerned about it also, thus 3 x 2900XT looks sweet :)  AMD have tech for single system to have 4 video cards or even more, lets see if they can make mature drivers for that. Lets not forget new high-end revision on 65nm or 55nm will consume much less power, same expected from G92 on 65nm, since 8800 GTX/Ultra's arent very power friendly as well ;) 
July 27, 2007 3:18:33 PM

Quote:
Lets not forget new high-end revision on 65nm or 55nm will consume much less power


Considering AMD's execution lately, nVidia will probably get their next-gen GPUs out to market before AMD. These will likely be on a smaller process and therefore consume much less power. I think they're expected in Q4 and will greatly widen the already significant lead over R600.
!