Saturday, March 17, 2007

AMD dual core frags Intel's quad core

2x dual core Opteron (2.8GHZ), total 4 cores, score 44.5

2x Clovertown, total 8 cores, score 41.3

2x Woodcrest (3GHZ), total 4 cores, score 41.4

33 Comments:

Blogger enumae said...

Sharikou

You may want to correct your post, or your link.

The AMD system your linking to is an "AMD Opteron 8220, 8 cores, 4 chips, 2 cores/chip."

---------------------------

Nice cherry picking...

Try this (58.9) for a Cloverton system.

... trying to be Intel huh?

7:00 PM, March 17, 2007  
Blogger Core2Dude said...


Nice cherry picking...
Try this (58.9) for a Cloverton system.
... trying to be Intel huh?

Wow! A 2P Intel system beating the cr@p out of a 4P AMD system? No wonder AMD is in big trouble!!

Sharikou, are you blind? You should have at least read the specs before posting the benchmarks... Pitting an 8-core 2.8 GHz Opteron system that is probably 5x more expensive agiainst an 8-core 2.33 GHz 2P Intel system? How desperate can you get?

7:10 PM, March 17, 2007  
Blogger Core2Dude said...

Shakira,

The title should have been, Intel 2P frags AMD 4P.

7:26 PM, March 17, 2007  
Blogger zeppelinrox said...

8 opteron cores
http://www.spec.org/cpu2006/results/res2007q1/cpu2006-20070108-00213.html
82.5
a bit higher than 58.9 linked above

8:52 PM, March 17, 2007  
Blogger Core2Dude said...

Sharikou, you are sick!! You are a shameless b@$tard that will sell his mom for a dime!!!

Funny how you do not mention the frequency on the Clovertown now. What are you afraid of? Are you afraid of the fact that people will see the hypocracy of pitting a 2.8 GHz processor against a 1.6 GHz one?

And why endit the original post? That was really good. 2.8 GHz 4P AMD Opteron with 32 GB of RAM giving only 54 base rate.

Enjoy while it lasts. With Penryn, AMD won't even be able to claim lead on specfp_rate.

@$$h01es like you should be banned from posting a blog.

11:19 PM, March 17, 2007  
Blogger Core2Dude said...

For the benefit of everyone else.

http://www.spec.org/cpu2006/results/res2007q1/cpu2006-20070216-00514.html

This is the system Shakira had first linked to, without realizing that this was a 4P system. But when Enumae and myself pointed out that a 2P Clovertown annihilates this system completely, he changed the original post. However, sadly for him, he could not find a single 2P AMD system that had a better base rate than the original Clovertown system at 2.3 GhZ. So now, he is comparing a 1.6 GHz Clovertown system with a 2.8 GHz Opteron system. The funny thing is, the 1.6 GHz processor will have a TDP of 50W.

No matter how you configure it, Clovertown always ends up on the top. That, in fact, is the moral of the story.

11:26 PM, March 17, 2007  
Blogger pezal said...

u r the stupid one core2dude.. No matter it is 2 or 4 cores just accept that the amd opteron is faster thn ur intel stupidtown chips... wakakakakka..

12:22 AM, March 18, 2007  
Blogger Giant said...

Core2dude's got it perfectly right. We saw a 2.33Ghz Clovertown 2P system beat out a 4P 2.8Ghz Opteron system. What's even more damning is that Intel has a 2.66Ghz version of Clovertown as well!

All the proof is there:- Intel's 2P platform frags AMD's 2P platform.

In many cases the 2P Bensley platform matches performance or even frags AMD's 4P solution. It wouldn't suprise me if 2P Bensley ends up faster than a 4P Tulsa system as well. I expect this would continue until Intel's new 4P server platform comes along sometime in Q3'07. (Tigerton processor with the Canelake platform)

Tigerton will be the final nail in the coffin for AMD's server division.

AMD BK Q2'08.

12:52 AM, March 18, 2007  
Blogger LeeCooper said...

Hm :).

Look at this

http://www.syndrome-oc.net/articles.php?article=94&lang=en

They are coming!

8:55 AM, March 18, 2007  
Blogger Bubba said...

Hey Doctor, people don't run synthetic benchmarks, they run applications.

Here you go, a real application that companies pay real money for.
Oh, and an AMD system didn't even make it into the top 50.

http://www.sap.com/solutions/benchmark/index.epx

AMD is going to need that 40%

10:45 AM, March 18, 2007  
Blogger tech4life said...

AMD x2 6000+ vs C2D E6700

With AMD's 4-5 year old architecture matching up pretty well to Intel's high end E6700 processor, doesn't this worry Intel? If Barcelona is even reasonably better than K8 it will be set to leapfrog Intel again for the performance and performance per watt crown. Can't wait to see those Barcelona benchmarks!

4:05 PM, March 18, 2007  
Blogger Giant said...

AMD x2 6000+ vs C2D E6700

With AMD's 4-5 year old architecture matching up pretty well to Intel's high end E6700 processor, doesn't this worry Intel? If Barcelona is even reasonably better than K8 it will be set to leapfrog Intel again for the performance and performance per watt crown. Can't wait to see those Barcelona benchmarks!


Sorry. The 6000+ is slower than the E6700. It matches the E6600. It also costs a lot more than the E6600 and uses double the power. It doesn't overclock nearly as well. No competition there. Intel has faster 65nm parts coming in Q3 and 45nm parts on a revised architecture this year. Then next year they do it all again with the new Nehalem architecture.

The only thing AMD has innovated lately is the 4x4 spaceheater! Two sockets on a motherboard? This is meant to be new and awesome? I vaguely remember doing this back in 1999 with the old Abit BP6 motherboard.

Didn't Sharikou also proclaim that 4x4 would "permanently pin Intel at 50% of AMD's performance."?

AMD BK Q2'08.

5:23 PM, March 18, 2007  
Blogger tech4life said...

Sorry. The 6000+ is slower than the E6700. It matches the E6600. It also costs a lot more than the E6600 and uses double the power. It doesn't overclock nearly as well. No competition there.

According to the benchmarks it compares very well to the E6700 running 64bit software (Vista). And this is 4-5 year old tech! It costs less than the E6700 (-$63) and I'm not an overclocker so that is a non-issue for me (and probably most people). Barcelona will solve the power issue with 65nm so what does Intel have left? Higher clocked C2Ds? Scary times ahead for Intel...

5:45 PM, March 18, 2007  
Blogger Azary Omega said...

Sorry. The 6000+ is slower than the E6700. It matches the E6600. It also costs a lot more than the E6600 and uses double the power. It doesn't overclock nearly as well. No competition there

Sorry but your wrong. According to this review X2 6000+ can do very good at gaming, exactly the place where Core 2 Duo had the advantage in the past: in Raindow Six Vegas Athlon X2 6000+ was .55% faster than Core 2 Duo E6700, Core 2 Duo E6700 took the lead in Oblivion being 6.18% faster than X2 6000+.

In rendering Athlon X2 6000+ took the lead all the way, beating Core 2 Duo E6700
in Cinebench by 7.51%
in POV-Ray rendering (chess2.pov) by 19.04%
in in POV-Ray rendering (benchmark.pov) by 2.94%

It looks to me that Athlon X2 6000+ IS better than Code 2 Duo E6700. End of story.

8:09 PM, March 18, 2007  
Blogger Ho Ho said...

tech4life
"With AMD's 4-5 year old architecture matching up pretty well to Intel's high end E6700 processor, doesn't this worry Intel"

Hey, but Sharikou said C2D is based on Pentium3, that means it is way older than K8.


tech4life
"Higher clocked C2Ds?"

At least Intel can clock its CPU's higher when needed. 3GHz quadcore will be here this year, also 65W TDP 3GHz dualcore. That 6000+ is 125W TDP. It is interesting to note that the 6000+ takes 250% more power than 1.86GHz LV quadcore.


azary omega
"X2 6000+ can do very good at gaming"

x2 is 6.6% slower in Oblivion. In HL2 particle benchmark it is around 53% slower. That technology will be used in upcoming Episode 2. In 3dmark game tests it is 6.8 and 4.5% slower. So on the average it seems like x2 is still considerably slower.


azary omega
"It looks to me that Athlon X2 6000+ IS better than Code 2 Duo E6700. End of story."

First, what makes it so much better?

After you ansver that question consider this. When all that Intel had was Netburst everyone were saying how bad it was since it took so much power. Now that AMD is in the same situation nobody seems to care for some reason. Intel takes 63% less power to do the job and that is together with the powerhungry northbridge.

1:24 AM, March 19, 2007  
Blogger Giant said...

The easiest way to compare CPU performance is by using the TomsHardware CPU charts. http://www23.tomshardware.com/cpu.htm

You'll get a clear picture. The 6000+ is similar in performance to an E6600, and is a bit slower than the E6700.

Unless someone already has an AM2 board and wants to upgrade there's no comparison. The E6600 is $313 vs. $459 for the 6000+. Over $100 cheaper for the E6600 for similar performance. It also uses half the power.

The comment about the 6000+ using 250% more power than a quad core 1.86Ghz LV Xeon is quite amusing too!

2:12 AM, March 19, 2007  
Blogger Ho Ho said...

giant
"The comment about the 6000+ using 250% more power than a quad core 1.86Ghz LV Xeon is quite amusing too!"

Oops, I made a little mistake there. It is actually only around 150% more or 2.5 times more. Just a bit bad wording.

2:28 AM, March 19, 2007  
Blogger TedsPlace said...

Intel is up to it again!!! Read this article.

http://content.zdnet.com/2346-10741_22-59008.html

7:15 AM, March 19, 2007  
Blogger Ho Ho said...

tedsplace, didn't you read what Sharikou wrote in his previous post?

7:31 AM, March 19, 2007  
Blogger tech4life said...

giant
The easiest way to compare CPU performance is by using the TomsHardware CPU charts. http://www23.tomshardware.com/cpu.htm


Some (most) of these benchmarks from Toms were run under 32bit. The latest benchmark I referenced here uses 64bit Vista with 64bit benchmarks. It would seem that AMD makes up a little ground going from 32bit to 64bit.

ho ho
x2 is 6.6% slower in Oblivion. In HL2 particle benchmark it is around 53% slower. That technology will be used in upcoming Episode 2. In 3dmark game tests it is 6.8 and 4.5% slower. So on the average it seems like x2 is still considerably slower.


I wouldn't call 4-6% considerably faster. In gaming performance if the E6700 musters 60fps and the x2 6000 scores 6% lower then that equals 56.4fps. Is the E6700 considerably faster? No. Its barely faster.

ho ho
After you ansver that question consider this. When all that Intel had was Netburst everyone were saying how bad it was since it took so much power. Now that AMD is in the same situation nobody seems to care for some reason. Intel takes 63% less power to do the job and that is together with the powerhungry northbridge.


Less power consumption is always a good thing no doubt. My original point though was that AMD's old tech could compete on performance with Intel's new tech. With Barcelona on 65nm the power consumption issue will be solved for AMD. Until then good job Intel.

12:49 PM, March 19, 2007  
Blogger Scientia from AMDZone said...

Obviously Sharikou made a mistake. It is well established that C2D is roughly twice the speed of K8 on SSE at the same clock. You'll have to wait for K10 to have a fight in SSE.

However, some of the other comments are a bit off. For example, there is no new cpu architecture coming in 2007 from Intel. In fact, I'm not aware of any new architecture coming in 2008 either. The 45nm chips are just a straight shrink of the 65nm C2D chips unless you are counting the additional L2 cache. The actual new architecture which won't work with socket 775 and has IMC and CSI won't be out until early 2009. There have been many claims made about Intel's 45nm process but Intel has not yet published roadmaps showing any higher clock speeds. So, I guess we'll have to see.

As far as Tom's Hardware Guide being a good way to truly compare Intel and AMD I'm sorry but that notion is absurd. THG has routinely slanted its "testing" in Intel's favor. These include:

Using the Intel Compiler instead of the neutral Portland Group compiler. Obviously, the Intel Compiler does not produce code optimal for AMD processors.

Using very fast DIMMs with high latency. The faster speed of the DIMMs adds nothing since the memory speed for AMD chips is always within a normal range however the increased latency does hurt performance.

Loading the tests with single threaded apps on multi-core processors. This obviously benefits Intel since a single core can use all of the shared cache.

Comparing overclocked Intel processors with stock speed AMD processors.

Using a non-NUMA OS for mult-threaded testing.

1:04 PM, March 19, 2007  
Blogger Scientia from AMDZone said...

On the issue of power consumption there is no doubt that AMD 90nm chips use more power than Intel 65nm chips. Why would anyone expect otherwise? However, the only true comparison that I am aware of was done at Tech Reports were nearly identical AMD and Intel systems were used. There were a handful of tests run at Anandtech on two similar systems.

Claiming that AMD draws 150% more power is rather silly. AMD like Intel makes chips in 120, 90, 65, and 45 watt ratings. AMD's ratings improved with 65nm the same as Intel's did. There is no significant difference currently in power draw between AMD and Intel on 65nm. However, I suppose you could make the case that the quad core Kentsfield uses less power than a dual socket AMD system. That's about as close as you can get to showing any advantage for Intel. And, of course, that advantage will be gone in a few months.

1:13 PM, March 19, 2007  
Blogger SAB said...

enumae said...

The AMD system your linking to is an "AMD Opteron 8220, 8 cores, 4 chips, 2 cores/chip."

You are wrong Enumae
IBM Bladecenter LS 21 has a max of 2 CPU.. Next time use Google..It applies to Core 2 Dude

2:05 PM, March 19, 2007  
Blogger enumae said...

This comment has been removed by the author.

2:48 PM, March 19, 2007  
Blogger enumae said...

This comment has been removed by the author.

2:50 PM, March 19, 2007  
Blogger enumae said...

Sab if you had read the previously deleted comments I appologize.

You should try reading some other post, his original link was incorrect, and since Sharikou doesn't mark that he has made changes it may look like Core2Dude and my self are incorrect, while we are actually not.

Here is his original link.

5:49 PM, March 19, 2007  
Blogger Giant said...

Scentia, you might want to check out the following:- http://www.reghardware.co.uk/2006/09/26/idf_otellini_keynote_process_tech/

CSI and Nehalem are still on track for next year.

Penryn is just a shrink of the current Core 2 CPUs with higher clocks, more cache and SSE4.

6:31 PM, March 19, 2007  
Blogger Core2Dude said...


Penryn is just a shrink of the current Core 2 CPUs with higher clocks, more cache and SSE4.


And improved FP. Expect FP performance on par with K8 (maybe K10, but not sure).

8:43 PM, March 19, 2007  
Blogger Scientia from AMDZone said...

giant

No, CSI will be released first on Itanium in 2008, not Xeon. The Itanium chip will use the same socket and CSI as Xeon so it is easy to get these confused.

Intel will release Itanium first to give Itanium some breathing room. Also, Intel is not in a hurry to toss out the FSB. When this happens, IBM's custom northbridge will be useless and any preference for Intel will vanish.

Also, Intel is not stupid enough to launch its own quad FSB northbridge chip and then have it made obsolete in less than a year.

core2dude

Are you talking about scalar or vector FP?

1:43 AM, March 20, 2007  
Blogger pointer said...

Scientia from AMDZone said...
As far as Tom's Hardware Guide being a good way to truly compare Intel and AMD I'm sorry but that notion is absurd. THG has routinely slanted its "testing" in Intel's favor. These include:

Using the Intel Compiler instead of the neutral Portland Group compiler. Obviously, the Intel Compiler does not produce code optimal for AMD processors.


Typical AMD fanboy style of accuse and excuse. Give the comment details enough to accuse, not enough to reveal the truth. First, may I know which test they do that? I don't believe they have the game source code for the compilation right? So, game scores are not affected. And same for other commercialized software too. And for those open source code, what's wrong of using Intel compiler for compilation? I would agree it is wrong to run Intel compiled code in AMD system though. Again, please point out which test they did that and how they did that. Software and hardware combine producing a platform. Just because you do not have a good compiler, it doesn't means Intel test has to be compiled with other compiler. The best way to do it (for open source code) is to have Intel system compiled with Intel compiler and AMD system compiled with whatever other compiler that you think it is best.


Using very fast DIMMs with high latency. The faster speed of the DIMMs adds nothing since the memory speed for AMD chips is always within a normal range however the increased latency does hurt performance.


so if they use the 'slow' DIMM for AMD (best case according to you) and fast DIMM for Intel (best case) would be fail enough? You think that you would say that it is not fair to use slower DIMM in AMD's test?

Loading the tests with single threaded apps on multi-core processors. This obviously benefits Intel since a single core can use all of the shared cache.

So this test should be run? it is just one of the tests in the benchmarks. This reflect part of the real world apps too, as sometime one apps would really occupy most of your CPU cycles? Don't let the fanboism fooling yourself till you can't see this. One of the benchmark objective is to let user judge a CPU under certain workloads, and single thread is certainly a possible workload, even for dual core or quad core.

Comparing overclocked Intel processors with stock speed AMD processors.
They did tel you it is overclocked right? Just because AMD's processors aren't that overclockable, then you prevent other to test it? I also saw them overclock AMD CPU too, just that AMD's can't be overclocked too much.

Using a non-NUMA OS for mult-threaded testing.

May i lknow what is the best NUMA OS for their testing at the time then?

6:59 AM, March 20, 2007  
Blogger Core2Dude said...


core2dude
Are you talking about scalar or vector FP?


I am talking about scalar. In vector, I believe Core 2 already leads K8 due to the 128-bt SSE, right? I am not sure if the FP improvements will also improve the vector--but I might be able to find that out. Of course, K10 will rectify AMD's vector performance. I have no idea how Penryn will compare against K10 in SpecFP or SpecFP_rate. But I would be extremely surprised if K10 blows it out of the water. And with a decent clock-speed advantage, I expect Penryn to pull ahead at least slightly.

Intel has realized that FP has been their weak point for a long time, and with Penryn, they are rectifying it.

10:40 AM, March 20, 2007  
Blogger Core2Dude said...


No, CSI will be released first on Itanium in 2008, not Xeon. The Itanium chip will use the same socket and CSI as Xeon so it is easy to get these confused.


Scientia, there is no confusion. CSI indeed is on track for Xeon in 2008. Of course, no one has seen the future, and hence, there is always a possiblity of a slip. But looking at the recent flawless execution, I would put that chance as extrememly remote.

If recent history is any indicator, Intel has consistently been beating its own roadmaps by a quarter or two. I think, after the Prescott fiasco, they have learnt the lesson, and are purposely over-estimating the time it would take to deliver the product. Take Penryn for example. On your blog, you have a huge post about how it is an 08 product, and not 07. But Intel has now officially confirmed that Penryn will be launched in Q3 07. Again, when Intel specifies a quarter, it typically means beginning of the quarter. So, Intel is beating the earlier roadmaps by about 6 months.

Both Itanium and Xeon will get CSI in 2008. On desktop, however, the CSI could be used as FSB, implying that it will probably connect the CPU to the northbridge that in turn will have the memory controller. I haven't figured out the logic behind this--probably Intel doesn't want to lose the Northbridge revenue? Another possibility is that it will give more memory choice on the desktop front. Third possibility is, it will allow Intel to sell processors whose IMC is defective as desktop products.

10:53 AM, March 20, 2007  
Blogger MÃ¥rten said...

Hey guys! Why don't you stop for a while and look at costs? What will give you the best performance for the money? I'm just about to buy two machines, one for real-time image processing (memory-heavy) and one for off-line computations using Lapack (CPU-heavy) and I've (more or less) decided for a 2220 (or 285) and a dual 5150 (or 5140)? For the chips that might come into question the current prices are as follows: $1172 (X5355), $690 (E5320, E5150), $698 (2220), $523 (2218, 285).

5:50 AM, March 23, 2007  

Post a Comment

Links to this post:

Create a Link

<< Home