backtop

Print E-mail del.icio.us 62 comment(s) - last by masher2.. on Mar 8 at 1:52 PM

System benchmarks

Woodcrest using FB-DIMM
Intel claims that it can outperform AMD by up to 33% on performance-per-watt

Intel is making a big push into performance-per-watt this year and Apple was touting Core Duo's performance in that realm when it announced its new Intel based Macs earlier this year. During Intel's keynote this morning, it demonstrated one of its next-generation cores, codenamed Woodcrest, against an AMD-based Sun server and claimed that core for core, Woodcrest outperformed AMD's offering by up to 33%.

Interestingly enough, the Woodcrest platform was using FB-DIMM technology while the AMD platform was running on existing DDR1 technology, which leaves a significant disadvantage for AMD.

Comments     Threshold


This article is over a month old, voting and posting comments is disabled
fair comparison?
By Homerboy on 3/7/2006 3:24:43 PM , Rating: 1
so Intel's next-gen versus AMD's current gen?
Am I wrong?




RE: fair comparison?
By GrammatonJP on 3/7/2006 3:28:07 PM , Rating: 5
they're trying to catch up to the current gen :)


RE: fair comparison?
By Questar on 3/7/2006 3:32:20 PM , Rating: 2
It's a comparison of what you will be able to buy when Woodcrest ships. AMD will not have another generation out at that time.


RE: fair comparison?
By JazzMang on 3/7/2006 3:34:36 PM , Rating: 2
Call me crazy, but I think DDR2 will be used by AMD at that point, no?


RE: fair comparison?
By Robertk20 on 3/7/2006 3:36:46 PM , Rating: 2
DDR2 doesnt even help amds performance as shown in the AM2 reviews.


RE: fair comparison?
By Griswold on 3/7/2006 3:45:52 PM , Rating: 3
AM2 reviews? Where? Oh you meant the pre-production sample previews.


RE: fair comparison?
By the eric conspiracy on 3/7/2006 4:05:41 PM , Rating: 2
Yeah, single channel DDR2-667 with 4-4-4-12 CAS didn't help over dual channel DDR400 2-2-2-5 CAS on an early rev CPU.



RE: fair comparison?
By DarthPierce on 3/7/2006 4:32:14 PM , Rating: 2
please help me find 2-2-2-5 CAS DDR2-667.
I can't seem to find really low latency DDR2.....


RE: fair comparison?
By robg1701 on 3/7/2006 3:41:54 PM , Rating: 2
And the Opteron 285 is already out now...


RE: fair comparison?
By JackPack on 3/7/2006 3:48:22 PM , Rating: 2
Which was paper launched yesterday....


RE: fair comparison?
By Doormat on 3/7/2006 4:02:44 PM , Rating: 2
http://www.monarchcomputer.com/Merchant2/merchant....

Paper launched? That looks like it says in stock to me.


RE: fair comparison?
By JackPack on 3/7/2006 3:46:04 PM , Rating: 2
So somehow, Intel is supposed to procure an NDA Socket F system and reveal it publicly?

I think the slower performance of FBD and the fact that it's running only at 667 MHz make up for DDR400.


RE: fair comparison?
By gamara on 3/7/2006 6:28:41 PM , Rating: 2
ECC DDR 400? ECC memory is notoriously slow as well.


RE: fair comparison?
By masher2 (blog) on 3/7/2006 11:27:32 PM , Rating: 2
ECC is only a couple percent slower. And, since FB-Dimms are purportedly far more reliable than even ECC RAM, I think its a more apt choice for a comparison than non-ECC RAM.



RE: fair comparison?
By Griswold on 3/7/2006 3:48:39 PM , Rating: 2
So what makes you think AMD wont have anything else to offer than that when Intel can ship an FD-DIMM platform? Nothing, you're just yapping.


RE: fair comparison?
By Questar on 3/7/2006 3:51:12 PM , Rating: 2
Really?

Please go ahead and tell me K9 (or K10) is shipping this year.

Please, say it and make a bigger fool of yourself.


RE: fair comparison?
By Duwelon on 3/7/2006 4:25:40 PM , Rating: 2
Gee, why would you bring this up?


performance-per-watt
By kattanna on 3/7/2006 4:50:55 PM , Rating: 2
all that i have seen claimed in this article is that PER WATT intel will have UPTO 33% more performance

theres nothing there about it actually performing faster then the opteron though





RE: performance-per-watt
By masher2 (blog) on 3/7/2006 4:57:22 PM , Rating: 2
According to TheInquirer today, Intel is claiming an 80% performance boost and 35% drop in power consumption for Woodcrest.


RE: performance-per-watt
By Questar on 3/7/2006 4:58:47 PM , Rating: 2
Please, don't quote that trash.


RE: performance-per-watt
By masher2 (blog) on 3/7/2006 5:02:17 PM , Rating: 2
> "Please, don't quote that trash..."

Hey, they got a fact right once just last year..give them a break.


RE: performance-per-watt
By Spoelie on 3/7/2006 6:01:22 PM , Rating: 2
Do you guys have your heads up your asses? The story was a direct quote of the intel guys at IDF. Go look on the slides that anand posted earlier, it says the exact same thing; and it is relative to a 2.8ghz xeon, no AMD Opteron.


RE: performance-per-watt
By masher2 (blog) on 3/7/2006 6:32:17 PM , Rating: 2
Read THIS article, not Anands:

During Intel's keynote this morning, it demonstrated one of its next-generation cores, codenamed Woodcrest, against an AMD-based Sun server and claimed that core for core, Woodcrest outperformed AMD's offering by up to 33%.

It's pretty clear whose head is up their own backside.


RE: performance-per-watt
By JumpingJack on 3/8/2006 1:42:36 AM , Rating: 2
From TGDaily:
http://www.tgdaily.com/2006/03/07/woodcrest_server...

The author reports the comparison was to a Sun Fire X4200 running at 2.4 GHz.


RE: performance-per-watt
By Spoelie on 3/8/2006 11:54:07 AM , Rating: 2
> Please, don't quote that trash...
> Hey, they got a fact right once just last year..give them a break.

So if those comments were about this article and not the inquirer one you are saying that this site is trash and this site only got one fact right last year?

Nice going


RE: performance-per-watt
By kattanna on 3/7/2006 5:00:31 PM , Rating: 2
aye..i saw that...but compared to what is my question?



RE: performance-per-watt
By DeathByDuke on 3/7/2006 5:55:47 PM , Rating: 2
a 486? lmao


DDR 1400??
By MIKEMIKE on 3/7/2006 3:46:38 PM , Rating: 2
wtf is DDR 1400




RE: DDR 1400??
By DarthPierce on 3/7/2006 3:56:35 PM , Rating: 2
If you're referring to the pic, it says DDR1 400.
not DDR 1400.
Big difference


RE: DDR 1400??
By gooser on 3/7/2006 4:24:11 PM , Rating: 2
ok, then WTF is "FB-DIMMs"? Sorry, I'm an ignant AMD troll.


RE: DDR 1400??
By Accord99 on 3/7/2006 4:32:12 PM , Rating: 2
http://www.theinquirer.net/?article=15167

Actually, at the moment they use more power than even DDR1 DIMMs.


RE: DDR 1400??
By DarthPierce on 3/7/2006 4:36:39 PM , Rating: 2
FB-DIMM is designed to improve RAS (Reliability Accessibility, Serviceability) and it is designed to allow much higher RAM density... e.g. up to 64GB/socket


RE: DDR 1400??
By masher2 (blog) on 3/7/2006 4:28:08 PM , Rating: 2
Fully-buffered DIMMs...a new memory architecture, primarily intended for reliability and stability. I don't believe FB-Dimms are intended to offer any performance increase, they'll start life in the server market segment.


RE: DDR 1400??
By highlandsun on 3/7/2006 7:31:14 PM , Rating: 2
By all accounts, even worse latency than DDR2, with latency that increases as more modules are installed.


RE: DDR 1400??
By masher2 (blog) on 3/7/2006 8:43:42 PM , Rating: 2
Not by "all accounts". Intel claims lower latency than DDR2 once you pass a certain throughput figure. And, due to much greater overall bandwidth, better performance.

In any case, the move to FB-Dimm isn't about raw performance, but rather capacity, reliability, and longevity. DDR2-based FB-Dimms will hit the server market first, and I wouldn't expect to see it on any desktops until we have DDR3-based FB-Dimms.


RE: DDR 1400??
By NullSubroutine on 3/7/2006 8:41:49 PM , Rating: 2
actually i think FB dimms (dont know if current ones do it) are slated to have a buttload of bandwidth.


By masher2 (blog) on 3/7/2006 7:09:59 PM , Rating: 4
> "Does anyone recall the two glued together chips Intel used to claim they had a dual core CPU "

Two cores in one socket = dual core chip. By definition. Now, begone troll.

> "Intel is so technologically behind AMD and so deperate to buy a clue they'll use any claims..."

Yep, they're so desperate they'll even do the unthinkable, and make a better chip.


By hans007 on 3/7/2006 9:13:31 PM , Rating: 2
first off...


lets get some stuff straight.

CAS latency is measured in cycles. a cycle can be at different speed.

an d800mhz cycle is not the same as a 400mhz cycle.

so in this case DDR400 CL2, is exactly the same latency as DDR2-800 @ CL4.


anyhow, that would mean that common consumer ram which is CL3 ddr400 is equivalent to cl3 ddr2-400 and cl5 ddr2-667 in total latency.

so for latency and not bandwidth ddr2 will have to be something like cl3 for ddr2-667 to relalyb e faster than ddr400 cl2 on a socket am2 vs 939.


that said, the socket F and 940 opterons use registered memory which adds at least one more cycle in the buffering chip.


so what this probably means, is ddr2 will do basically nothing for the amd platform unless it is bandwidth starved and in general for almost anything that doesnt matter.


and for people who do now know what FB dimms are. basically they are dimms with many chips, with some sort of controller chip . dimms are parallel devices, and the chip is also more or less a serial to parallel bridge somewhat like the ones on some earlier SATA drives to the parallel ata interface.


the serial converter allows the module to have less traces on the motherboard since serial will require less of them. its not really faster, it just allows better routing, and the bridge chip also would be able to address more memory as it would be buffered.

AMD will not have this technology as well they would need to change their memory controller again. there are 8GB FB-dimms slated for future production, so well... yeah intel would hvae an advantage in the ability to put out machines with more total memory.




By masher2 (blog) on 3/7/2006 10:29:33 PM , Rating: 2
> " its not really faster, it just allows better routing"

While correct on your other points, you're wrong on this one. FB-Dimms not only allow for higher overall bandwidth, but they allow simultaneous read/write requests against the chip. So there is a performance benefit...at least for highly loaded memory subsytems.


By butane317 on 3/7/2006 10:57:53 PM , Rating: 2
"so in this case DDR400 CL2, is exactly the same latency as DDR2-800 @ CL4."

No. DDR400 is actually at 200mhz, and DDR2-800 is also at 200mhz. That's why having high latencies on DDR-2 really kills performance.


By masher2 (blog) on 3/7/2006 11:31:21 PM , Rating: 2
> "No. DDR400 is actually at 200mhz, and DDR2-800 is also at 200mhz"

Err, no this is incorrect. DDR2-400 has memory chips running at 100 mhz and an I/O clock of 200 mhz. DDR2-800 has both rates exactly double this.



Competition is good
By lifeblood on 3/7/2006 3:35:23 PM , Rating: 2
If this is truly a reflection of performance then go Intel! I love competition. It raises the bar and gives AMD something to beat (which I am sure they will do).




RE: Competition is good
By bamacre on 3/7/2006 3:41:12 PM , Rating: 2
Exactly. Good for all of us. Except the hardcore AMD fanboys.


RE: Competition is good
By Duwelon on 3/7/2006 4:24:20 PM , Rating: 2
Yeah I don't think i've ever seen any more rabid fanboyism than I have with AMD in recent years. Don't get me wrong, I got my San Diego 3700 clocked at 2.7ghz and put my previous p4 to shame, but I still get a kick out of the AMD fanboys shrieking in pain at these benchmarks.


RE: Competition is good
By bob661 on 3/7/2006 4:32:23 PM , Rating: 2
When is Conroe officially launched?


RE: Competition is good
By DarthPierce on 3/7/2006 4:39:34 PM , Rating: 3
H2 '06


By smokenjoe on 3/7/2006 9:47:54 PM , Rating: 3
I can understand loyalty to a ball club but to a monopolistic big money anticompetition corporation what is up with that?

OK as to intel having more on chip memory like tht really makes the diffrence google Z-ram http://www.eweek.com/article2/0,1895,1913489,00.as...

Guys it is a moving target and intel is the king of hype get over it they are a marketing company that also makes chipsexcept of the Intiam fiasco they got to spend more on advertising than on reasearch.




By JumpingJack on 3/8/2006 1:51:42 AM , Rating: 2
Nice link, I am worried by the statement

quote:
AMD has said it is aiming to roll out a quad-core-capable processor family during 2007 and then switch to an all-new processor architecture later this decade.


New architecture later this decade, does this mean 2007? 2008? 2009?

If they do not exceed the new architecture, then it could mean bad news....


By smokenjoe on 3/8/2006 2:41:53 AM , Rating: 2
No one noticed the AMD cpu was intels future processors was up aganst was not exatly top of the line even now that is more of a budget system.



News flash in a few months Intels top of the line system might be able to beat a budegt AMD server this is an improvement they were not even a close second but at least go up aganst the best they have now.

You guys are a hoot.


By masher2 (blog) on 3/8/2006 9:20:14 AM , Rating: 2
> "the AMD cpu was intels future processors was up aganst was not exatly top of the line even now that is more of a budget system. "

An AMD 280 is not a 'budget' processor. The chip alone is well over $800. The only faster Opteron is the 285... and AMD introduced that one juts a few days ago.


You must be kidding...
By Smooth2o on 3/8/2006 1:05:15 PM , Rating: 2
The only part that is not 64 bits is Core Duo (notebooks) and that will run 32 bit Vista and turn into 64 bit Merom in Q4 with the same kind of performance as Conroe.

If you haven't read the Anand review or the Hexus review, you should. I don't care what you compare to, Intel just jumped wayyyyy ahead of AMD. Otellini was right, Intel will have the lead for at least the next 1.5 years.




RE: You must be kidding...
By MrEMan on 3/8/2006 1:25:24 PM , Rating: 2
"Intel just jumped wayyyyy ahead of AMD. Otellini was right, Intel will have the lead for at least the next 1.5 years."

If the performance numbers are correct and if it ships on time then the question is: what is the big deal? AMD has had the lead for years. It is the nature of processor design and production cycles that the lead changes over time. As a matter of fact, we all benefit because of the competition.

Why I haven't had an Intel processor in years (last one was a 80386SX), I only want AMD to become prosperous enough to be able to continue to force Intel to compete technologically, and economically.

If the new Intel processors actually perform as well or better than announced, then it seems that the corporation finally turned the designs over to the engineers and not the sales and marketing department.

Lastly, finally, the "Dell will actually use AMD cpus" claim can be put to rest. As far as I am concerned they never intended to do so anyway. It was just a way for them to get better pricing out of their parent corporation (Dell using AMD is equivalent to Intel itself using AMD... never gonna happen).


RE: You must be kidding...
By masher2 (blog) on 3/8/2006 1:52:25 PM , Rating: 2
> "Dell using AMD is equivalent to Intel itself using AMD... never gonna happen"

While I agree with the rest of your analysis-- never say never.

Good advice, there.


Company-Supplied Benchmarks
By CKDragon on 3/7/06, Rating: 0
RE: Company-Supplied Benchmarks
By JackPack on 3/7/2006 3:50:49 PM , Rating: 2
Time for bifocals.


RE: Company-Supplied Benchmarks
By CKDragon on 3/7/2006 5:44:53 PM , Rating: 2
Since you apparently think I've missed something, care to share what it might be?


Missing something?
By killerroach on 3/7/2006 6:35:20 PM , Rating: 2
The one thing that Woodcrest has as an advantage isn't the DDR2, it's the gynormous L2 cache that it has. IIRC, isn't that part supposed to have 4MB of L2? Sure, it might not increase power consumption much, but something tells me having that much L2 means their yields are going to suck...




RE: Missing something?
By smokenjoe on 3/7/2006 7:06:07 PM , Rating: 2
Let us take a step back and think here Intel maid the same kind of claom when the presscots and Preslercame caim out out. Lets wait for some independant tests on real world systems. Assuming the competition will not improve is silly to fail to improve is to suffer Intel has learned ths and I am sure AMD will not forget that lesson.

I wish Intel luck the supply or AMD processors is very tight. No one in there right mind wants an Intes right now you cant give them away to performance nuts.


Memory And Bus
By Nehemoth on 3/7/2006 3:42:14 PM , Rating: 2
The Diferents Here are the memory DDR1 vs FB-DIMMs also if the diferent between FSB and the HT the which course is 1Ghz but bidiretional




that doesnt seem right...
By DeathByDuke on 3/7/2006 5:53:44 PM , Rating: 2
DDR 1400? you mean DDR 166Mhz? o_O




so is this 64 bit or not?
By shred5 on 3/8/2006 9:30:30 AM , Rating: 2
So is this 64 bit or not...




"The Space Elevator will be built about 50 years after everyone stops laughing" -- Sir Arthur C. Clarke









botimage
Copyright 2009 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki