Sign in with
Sign up | Sign in
Your question

Xeon Woodcrest Preys On Opteron

Tags:
  • Memory
  • Xeon
  • Opteron
Last response: in Memory
Share
June 26, 2006 11:57:04 AM

Intel must have craved this day: The Xeon 5100 at 3.0 GHz and 4 MB L2 cache launches today to decapitate the AMD Opteron. Our look under the hood explains what the Core 2 microarchitecture is capable of.

More about : xeon woodcrest preys opteron

June 26, 2006 1:01:01 PM

1 mistake:
In page 9 you wrote "Digital Media Boost... Oder Doch SSE4?" but shouldnt that be in English rather than German?
June 26, 2006 1:21:40 PM

I don't care for the new Table of Contents. I much preferred being able to scan all the page titles at a glance without a pull-down menu. I realize this may be off-topic but I wanted to get it out before all the fanboys on both sides crash the forum.
Related resources
June 26, 2006 1:32:19 PM

Quote:
I don't care for the new Table of Contents. I much preferred being able to scan all the page titles at a glance without a pull-down menu. I realize this may be off-topic but I wanted to get it out before all the fanboys on both sides crash the forum.


Agreed. Dropdown lists suck. Is my browser not loading the pages correctly? I see no benchmarks of an AMD system in that article at all.

Who wrote this crap?
June 26, 2006 1:36:46 PM

Quote:
Is my browser not loading the pages correctly? I see no benchmarks of an AMD system in that article at all.

Who wrote this crap?


Actually if you read the back page you will see that they are reserving the Intel vs. AMD benchies for another article, allegedly so that they can be more complete.

Also, when the article mentions the Advanced Memory Buffer, the writer refers to it as the "AMD" instead of "AMB". This could cause some confusion and should be addressed.

I do not want to be seen as nitpicking, but I have come to expect outstanding articles from the THG staff and I just feel that this article may not be up to their world-class standards.
June 26, 2006 1:40:57 PM

First, I am not a fanboy.

That being said, I do not understand why Intel is making such a fuss about being better than a 2 year old processor, unless they know AMD is about to crush them, and they are making the best of a bad situation.

Seems like whenever Intel talks about it's new processors, it also talks about AMD, isn't that a marketing no-no?

In the old days, Intel would market it's chips based on what they could do. This is the first time I recall them marketing their chips based on what AMD could do.
June 26, 2006 1:57:41 PM

Well, what did AMD say about beating a more-than-3-year-old chip at gaming?
June 26, 2006 1:59:02 PM

Quote:
First, I am not a fanboy.

That being said, I do not understand why Intel is making such a fuss about being better than a 2 year old processor, unless they know AMD is about to crush them, and they are making the best of a bad situation.

Seems like whenever Intel talks about it's new processors, it also talks about AMD, isn't that a marketing no-no?

In the old days, Intel would market it's chips based on what they could do. This is the first time I recall them marketing their chips based on what AMD could do.


The company that has the performance touts said performance..... it has nothing to do with "how old" a chip is, when purchasing managers go out to make a buy they will not look at the performance, say I want the best, then dismiss it simply because it is beating a 2 year old chip.Look at the 1987 Mustang GT. It was a phenominal success, yet was based on the fox platform. It had Pinto rear drum brakes FFS, but every other car on the road was a Mustang. I bought an '88 GT Loaded. Best bang for the buck, but old designs, mixed in with some new. Total POS though, quality wise.
June 26, 2006 2:02:57 PM

Quote:
Look at the 1987 Mustang GT. It was a phenominal success, yet was based on the fox platform. It had Pinto rear drum brakes FFS, but every other car on the road was a Mustang. I bought an '88 GT Loaded. Best bang for the buck, but old designs, mixed in with some new. Total POS though, quality wise.


:lol:  :lol:  :lol:  LOL
June 26, 2006 2:04:12 PM

Quote:
I don't care for the new Table of Contents. I much preferred being able to scan all the page titles at a glance without a pull-down menu. I realize this may be off-topic but I wanted to get it out before all the fanboys on both sides crash the forum.


Agreed. Dropdown lists suck. Is my browser not loading the pages correctly? I see no benchmarks of an AMD system in that article at all.

Who wrote this crap?

The only line in this article you need to read is the conclusion which says:

However, we do not see Woodcrest knocking out AMD, since the HyperTransport architecture remains the best choice for 4P solutions thanks to its point-to-point layout and dedicated memory controller per processor.
June 26, 2006 2:15:53 PM

To clarify, Intel is setting up Woodcrest as THE Opteron beater. They are making sure everyone gets that message.

What happens to the message when Opteron is no longer around to beat?

To answer Ycon, AMD beating Intel for any length of time was news, as it had never happened before. Intel proclaiming itself as the AMD beater for whatever window they have until AMD releases a new chip (I believe the window should be counted in months) tells us that AMD did indeed whoop on Intel, in case there were any doubters, and also points out the tragedy of how far Intel fell.

In short, there is a big difference in how the #2 chip maker can and should market, and how the #1 chip maker can and should market.
June 26, 2006 3:14:12 PM

Excuse me? QUAD Channel DDR and it only can give 5700 MB/s??
This has something to do with buffering?? :?
June 26, 2006 3:23:26 PM

Quote:
The only line in this article you need to read is the conclusion which says:

However, we do not see Woodcrest knocking out AMD, since the HyperTransport architecture remains the best choice for 4P solutions thanks to its point-to-point layout and dedicated memory controller per processor.


That certainly caught my eye. More than anything, I think of '06 as a time of huge flux in the small computer market and it may take a while before this all shakes out. The roadmaps are so aggressive, I'll be amazed to see the waypoints tagged within 6 months of schedule for 45nm, for example.
June 26, 2006 3:27:58 PM

Quote:
Excuse me? QUAD Channel DDR and it only can give 5700 MB/s??
This has something to do with buffering?? :?


It's serial. Regular DDR is parallel.

I'm interested to see the benches. Architecture can be argued all day, but I'd like to see how Woodcrest holds up under load.

_________________________________________________

TO THG: I know it's not a gaming system, but I'd rather have a woodcrest than conroe, even if I have to take a hit in gaming. I would like to know what kind of hit I'd take, though.

When you run the benches, can you please run some gaming benches? I just want to know that if I get a woodcrest, the FB-DIMM latency or something else won't murder my chance of decent gaming. I'm not looking for a Woodcrest/Opteron gaming-system shootout, but just some F.E.A.R. and 3dMark on an XTX or Dual GPU SLI card would be nice for reference against a Conroe.

Thanks.
_________________________________________________
June 26, 2006 3:33:02 PM

Every company says theyre #1...
Plus, Woodcrest does beat Opteron at everything, thats the differnce.

However, just like you stated before, this is OT.
June 26, 2006 3:33:24 PM

Quote:
Excuse me? QUAD Channel DDR and it only can give 5700 MB/s??
This has something to do with buffering?? :?



Quote:


FB-DIMM technology offers scalability of 192 gigabytes — 6 channels, 8 DIMMs/channel, 2 ranks/DIMM, 1 gigabyte DRAMs and offers bandwidth of 6.7 gigabytes per second (GBps) sustained data throughput per channel.
Source
June 26, 2006 3:39:45 PM

1 mistake on page 1:
...has had to suffer from tremendous criticizm over the last 18 months.

2 mistakes on page 4:

...conventional DDR SDRAM due to repidly degrading signal strength.
...administrator to deploy four memory modules at once to polulate all four channels

1 mistake on page 5:
...featured PCI-X 133 bus that poweres two 64 bit slots.

2 mistakes on page 6:

...we are still sceptical concerning the efficiency...
...numerous FB-DIM modules...
June 26, 2006 4:26:29 PM

Why do you always post these in the Memory section?
June 26, 2006 4:38:12 PM

This is beside the topic.
But why does everyone call Americans "Yanks"? (just want to know)
Thanks
June 26, 2006 4:40:01 PM

Ditto. If the Woodcrest really did kill the Opteron, prove it. All I see is information about the woodcrest and test of its quad-channel FB-DIMM controller.

And about quad-channel-I know that the FB-DIMMS are serially linked, but I can see this moving to the desktop once quad-core chips start showing up unless RAM gets a LOT faster in a hurry. I think it's time to move to 128-bit-wide memory modules, personally.
June 26, 2006 5:05:59 PM

Thanks for the reply. I checked the dictionary and Yankee really is a native inhabitant of the U.S. So the word Yank really is "technically" correct English. I personally think very highly of the UK and Ireland. Notice my avatar is Tom Baker? I think it also doesn't hurt to have better rock-bands, better actors, and a richer culture than the U.S.
Ok, I said enough. I didn't want this to turn into a thread.
June 26, 2006 5:24:39 PM

Quote:
This is beside the topic.
But why does everyone call Americans "Yanks"? (just want to know)
Thanks


Yanks you around! LOL :p  :lol:  :lol: 
June 26, 2006 5:42:49 PM

This is completely off topic but I'm going to agree with you completely. American culture as we know it is just a mix of everything that's been brought over. The only true culture we have is that of the native americans but how many of them are left? Even though I am an native-born american I don't always think fondly of our culture.
June 26, 2006 5:47:48 PM

Good, we all agree. Now we can stop talking about this.
BTW - Yank is not an offence, I just wanted to know what that was exactly.
June 26, 2006 6:51:29 PM

Quote:
Why do you always post these in the Memory section?


Yes... Why is that?

This should go into Hardware---->CPU---->Intel-AMD Flame Wars---->Xeon Woodcrest Preys On Opteron
June 26, 2006 8:17:03 PM

lol...my point was simply that at my work we have editors
June 26, 2006 8:50:44 PM

I wonder why Intel feels they need higher clock speed, and double cache. Didnt we see this in the Pentium D?
If the chip was so superior they would not need to double the cache to compete. Maybe they have an inferiority complex.
June 26, 2006 9:28:40 PM

am I the only one who is shocked that a NEW product line from the industry leader is still offering a 1.6ghz processor? a 1ghz processor was the best price/performance option what, 5 years ago? how can the budget line of a BRAND NEW line of processors have such slow clock speeds, when a cheap older dual core processor probably being discontinued soon can be pushed over 4ghz?

yes yes I know, ghz is not everything in determining speed, but that sounds like a marketing line rather than admitting that processor speed technology is stagnating, so rather than address that, let's pile a bunch of processors together (ala voodoo/sli with graphics), which offers less than a 50% gain in speed per additional core (or graphics card).

secondly, I have to agree, that spelling and grammar DOES matter on a professional website. After all, Tom's has a California office, I'm pretty sure people there speak english? I think Tom's should hire me as an editor, because I could very easily do better. The translator should not also be the editor, as seems to be the case here. :roll:
June 26, 2006 9:39:11 PM

Quote:
The only line in this article you need to read is the conclusion which says:

However, we do not see Woodcrest knocking out AMD, since the HyperTransport architecture remains the best choice for 4P solutions thanks to its point-to-point layout and dedicated memory controller per processor.


That certainly caught my eye. More than anything, I think of '06 as a time of huge flux in the small computer market and it may take a while before this all shakes out. The roadmaps are so aggressive, I'll be amazed to see the waypoints tagged within 6 months of schedule for 45nm, for example.

This definetely is an interesting time.... Hopefully we the customers will win, because of AMD, and Intel pushing each other.... And I think the IBM 'cell cpu' is going to have a big effect on this battle when it hits the scene....

I am hopeing AMD is quiet, and short on stock right now because they decided to change their plans at the last minute to beat conroe.... Could they be cranking out quad cores as we speak - with optimum ddr2 performance?
June 26, 2006 9:45:57 PM

America does generally have better rock bands than the Europeans do (and a heck of a lot more too), but the Canadians give us a run for our money. However, a lot of the electronica and dance-club bands are European. That kind of meshes with the fact that those kinds of clubs are really popular over there and we tend to have more rap and R&B played in the clubs over here (which explains why I don't go to them :D  )

Better actors? I have seen good and bad ones from both here and over there. If you are talking comedians however, the U.S. wins by a landslide, at least in my opinion. British humor seems to me to be just slightly funny as the comedians are much more predictable- you can see their jokes' punch lines a mile away or so it seems, so it isn't as funny when they do deliver. They are also much more highbrow than the American comics, and most of the good comedians are not too sophisticated (Adam Sandler, the Blue Collar TV guys, etc. etc.) and tell jokes that are a bit more crass.

America does in fact have a very strong culture, although it is not that apparent as it is more manifested in the populace than in physical works. Our culture is that of trying to "beat the Joneses" so to speak, so we tend to spend a lot of effort at our jobs to make as much money as possible. That consumes all of our energy and most of our time, so we tend to just veg out after work and not want to do anything else that requires effort that's not a job. That is why many Americans love entertainment so much as it is completely passive. And also why we are rather overweight as a whole and not too willing to do any work that is not required of us, and are WAY in over our heads in debt.
June 26, 2006 9:49:11 PM

Quote:
1 mistake:
In page 9 you wrote "Digital Media Boost... Oder Doch SSE4?" but shouldnt that be in English rather than German?


"repedily" is a word ? somewhere on page 3-5 . . .
June 26, 2006 9:53:15 PM

Quote:
i didn't know a yank was an insult. its just another word for an american. no offence meant. what does annoy me is people stating "facts" about how things are spelt and also doing the whole THG bashing. you can call me anything you want frankly but some members of your nation let you down.

also better rock bands is debatable and subjective and better actors is just laughable. it depends on what you define as acting. oh and culture, you guys only have what cultures have been brought over from other places and have none to call your own. sorry but i aint much a fan of you yanks when you act the way you do. let yourself down totally.

hey look what you've done hojacked a thread, congrats.


Well atleast americans dont go around planting flags in other countries, and absorbing thier culture eh ?

Anyhow what is this ? "America preys on Iraq", or "Xeon Woodcrest Preys On Opteron"

Really . . .
June 26, 2006 9:54:50 PM

Quote:
I wonder why Intel feels they need higher clock speed, and double cache. Didnt we see this in the Pentium D?
If the chip was so superior they would not need to double the cache to compete. Maybe they have an inferiority complex.


They do it for a few reasons:
1. They have a frontside bus arrangement that has a higher latency penalty for accesses to RAM than AMD's integrated memory controller. Intel increased the cache size on their chips to counteract that latency as with a larger cache, fewer reaches into the RAM must be made.

2. Intel has a very large fab capacity and is at a smaller node than AMD. So it doesn't cost Intel that much more to make a 4MB instead of a 1MB or 2MB cache, but performance increases with the larger sizes for the reason in #1.

3. They use the shared L2 cache to communicate between cores in the Core 2. I bet that the larger the cache, the better the cores can "talk" by puts and takes into and from the L2 and less FSB traffic needs to be generated for the cores to talk to one another.

4. Cache thrashing decreases with a larger cache as more data can fit in the cache and there's a lower chance that data will be written, overwritten, and then rewritten.
June 26, 2006 10:13:39 PM

Quote:
I wonder why Intel feels they need higher clock speed, and double cache. Didnt we see this in the Pentium D?
If the chip was so superior they would not need to double the cache to compete. Maybe they have an inferiority complex.


Ok, have you ever disabled the L2 cache on a system, and then booted an OS on it ? In case you havent, let me tell you how fast the system would perform. ON a PIII 833 with 512 KB's L2, it would take anywhere from 30-60 minutes to boot to logon from power on in WinXP (with this L2 cache disabled of course). I've done this recently, to make a 'newer system' work with a program written to be run on a 286. I quoted 'newer system' because obviously its not new, but it IS far newer than the program . . .

Anyhow while you probably wont see too much speed by doubling a L2 it WILL perform better. Last time I read about the MHZ race, it seemed Intel was having problems crossing the 4 GHZ barrier (dont know if this still stands), but Intel, and AMD are both approaching the limit of how much faster they can make thier respective cores. So without changing circut design RIGHT NOW, thier options are semi limited as to how they can increase thier CPUs performance. I dont claim to be an Electronics engineer, but the option I do know of right now are doubling cores, or increasing L2 cache sizes.

I also read semi recently that Intel (dont know if AMD can, or will follow suit) is considering a new transistor 'design?'. Transistor 3D or something simular to that. This will supposedly decrease power leakage / increase performance a certain % at the same time, which COULD mean they could finally break a barier or two. This however is still in experimentation, and may never even come to pass. . .
June 27, 2006 12:29:39 AM

My only comments are that it's too bad that you are using an older stepping 4 sample of Woodcrest. I believe the Engineering Samples that GamePC used in their Woodcrest Preview were Stepping 5s. This should probably be noted in the review. Also, you were only using DDR2 533 FB-DIMMs which means that you are in the ironic situation of the FSBs actually being underutilized instead of being the bottleneck. DDR2 667s should be used, and again the discrepancy should be noted. I think it should also be mentioned that there are concerns that going from DDR2 533 to DDR2 667 may cause a noticeable jump in power consumption so it would be appreciated if you could prove of disprove this notion. It's good that you are using 4GB of memory though since some people have felt that that is the minimum amount needed to start seeing the power impact of FB-DIMMs.

I look forward to seeing the full benchmark results and needless to say I hope it's comprehensive enough to be worth the wait.
June 27, 2006 1:12:23 AM

Quote:
i didn't know a yank was an insult. its just another word for an american. no offence meant. what does annoy me is people stating "facts" about how things are spelt and also doing the whole THG bashing. you can call me anything you want frankly but some members of your nation let you down.

also better rock bands is debatable and subjective and better actors is just laughable. it depends on what you define as acting. oh and culture, you guys only have what cultures have been brought over from other places and have none to call your own. sorry but i aint much a fan of you yanks when you act the way you do. let yourself down totally.

hey look what you've done hojacked a thread, congrats.


Why do you always seem to take personal offense to criticism of THG? He wasn't bashing THG: He was just noting some spelling errors. When you're as big as THG, people expect a touch more from the quality.

If you want bashing: it only takes about 30 seconds to copy/past an HTML article into MS word and pick out the spelling errors ;)  You'd think with all the time the authors spend on these articles, they'd take a minute to do a spell check. 8)
June 27, 2006 2:09:15 AM

Get your Intel stock now before the desktop version hits and the stock hits $25+ a share :p  hehehehe (got mine at 17.12 and now its at 18.42 so im happy already)
June 27, 2006 2:24:21 AM

Quote:
My only comments are that it's too bad that you are using an older stepping 4 sample of Woodcrest. I believe the Engineering Samples that GamePC used in their Woodcrest Preview were Stepping 5s. This should probably be noted in the review. Also, you were only using DDR2 533 FB-DIMMs which means that you are in the ironic situation of the FSBs actually being underutilized instead of being the bottleneck. DDR2 667s should be used, and again the discrepancy should be noted. I think it should also be mentioned that there are concerns that going from DDR2 533 to DDR2 667 may cause a noticeable jump in power consumption so it would be appreciated if you could prove of disprove this notion. It's good that you are using 4GB of memory though since some people have felt that that is the minimum amount needed to start seeing the power impact of FB-DIMMs.

I look forward to seeing the full benchmark results and needless to say I hope it's comprehensive enough to be worth the wait.


Excellent analysis. I wondered about the DDR 533 but am sufficiently ignorant as to have thought that Core server busses ran slower RAM. The power consumption aspect is interesting and I look forward to seeing the outcome of a thorough review.
June 27, 2006 5:43:24 AM

Quote:

I look forward to seeing the full benchmark results and needless to say I hope it's comprehensive enough to be worth the wait.

Quote:

Excellent analysis.


You've won the most useful post of the day :trophy: :) 
Thanks..
June 27, 2006 5:50:07 AM

I get 9600 MB/s in STREAM using dual Opteron 246HE (not exactly the
latest technology, since I'm using HDAMA motherboards
that are > 3 years old and PC-2700 memory modules) and
the Pathscale compiler.

If this Woodcrest benchmark is even close, there's no reasonable
expectation that Intel can even remotely compete with AMD, which
is a bad deal for consumers. The Woodcrest dual channel results
are as pathetic as my dual 3.2 MHz Xeon Dell 670 which runs
at about 3600 MB/s in STREAM in scalar mode but only 3000 MB/s
in dual mode due to bus contention.

The current dual-core Oterons don't have sufficient memory bandwidth
to feed both processors, so a dual socket system is faster than a
single dual core by a substantial amount. Socket F looks to reduce
or eliminate that problem and if AMD really goes to beyond Socket F
in early 2007 as indicated and Intel has to wait until 2009 to get
a real memory bus, there's no competition in the HPC sector
for at least another 2 years. Unbelievable.
June 27, 2006 6:26:37 AM

Remember kids, meth is bad for you.
June 27, 2006 7:37:09 AM

Irregardless of which camp you are in, we are living in interesting times.

(Pay attention, class!)
June 27, 2006 11:43:06 AM

Quote:
Intel must have craved this day: The Xeon 5100 at 3.0 GHz and 4 MB L2 cache launches today to decapitate the AMD Opteron. Our look under the hood explains what the Core 2 microarchitecture is capable of.




Ummm... I think I missed the benchmark page.
June 27, 2006 12:25:30 PM

Meth is bad for humans, slow memory buses are bad for computers ;) 

On a more serious note, the STREAM benchmark has been
the most reliable predictor of real-world performance on most
of my HPC codes. Check it out:

http://www.cs.virginia.edu/stream/
June 27, 2006 12:41:57 PM

Quote:
Meth is bad for humans, slow memory buses are bad for computers ;) 

On a more serious note, the STREAM benchmark has been
the most reliable predictor of real-world performance on most
of my HPC codes. Check it out:

http://www.cs.virginia.edu/stream/




Indeed :-D

I couldn't agree more!

$STREAM=kewl();
June 27, 2006 1:48:34 PM

Quote:
Meth is bad for humans, slow memory buses are bad for computers ;) 

On a more serious note, the STREAM benchmark has been
the most reliable predictor of real-world performance on most
of my HPC codes. Check it out:

http://www.cs.virginia.edu/stream/




Indeed :-D

I couldn't agree more!

$STREAM=kewl();Meth isn't kewl. It's production wastes way too many cold medications. :wink:
July 2, 2006 5:23:31 AM

So where are the actual benchmarks? Article said they would be added soon but its been almost a week.
July 3, 2006 11:11:17 PM

What I would really like to see is a benchmark comparison between the conroe and woodcrest in a comparable single processor setup. Primarily showing heavy multitasking and database error checking, but also a small test for rendering speed or gaming framerates..

I would like to know the viability of the desktop version for server operations and also the use of server hardware for gaming.
      • 1 / 2
      • 2
      • Newest
!