Sign in with
Sign up | Sign in
Your question

Fastest Computer 4 SQL Queries

Last response: in CPUs
Share
August 25, 2005 9:06:32 PM

Hi,
I need info what is the fastest computer I can get for querying a DB.
The computer will have a large DB of several GBs and will run in what I understand is an SQL Server configuration only the queries are run locally - the computer is not connected to a LAN for security purposes.
The processing is not paralleled so I'm looking at a single-processor, single-core system. Speed is most important so I'm willing to pay a pretty buck on the best CPU, but I still need to know which is better at this - Inter or AMD.

I see no reason to go with an Opteron/Xeon solution, but stay with the enthusiast's Athlon/Pentium, though I may be wrong.

How much does cache/freq./latency weigh-in? Will I benefit from 2 GBs of RAM?

Also, will I benefit from a fast HD and if so, is it worth paying for some 15K, 0+1/5 SCSI RAID?

<b>Behold, Mine anger and My fury shall be poured out upon this place upon man and upon beast and upon the trees of the field and upon the fruit of the land and it shall burn and shall not be quenched
August 25, 2005 10:53:23 PM

What SQL app doesn't support multiple CPU's ?

Fastest from Pentium/Athlon is AMD FX-57, or Opteron 254 (Opteron 254 is cheaper).
But it depends what app is used, since in some cases P4 can be faster.
But good thumb rule is: AMD is faster with bigger Databases.

You will benefit from >2GB RAM if your application and OS supports it.

<font color=red>"We can't solve problems by using the same kind of thinking we used when we created them."
- Albert Einstein</font color=red><P ID="edit"><FONT SIZE=-1><EM>Edited by HansGruber on 08/26/05 01:56 AM.</EM></FONT></P>
August 25, 2005 11:00:29 PM

>will have a large DB of several GBs

SQL Server will cache as much data in RAM as it can. If query speed is going to pay back the cost of this machine, your first investment is to max out the system RAM. Then of course you want to make sure SQL Server is configured properly to be able to use that RAM. For optimum results, it would be good to have enough RAM to hold the entire database in memory. If not, you'll probably want to consider a RAID 5 array, using the fastest drives your budget allows.

>The processing is not paralleled

You may be mistaken on that point. From the SQL Server 2000 Books Online: "Queries heavily consuming CPU cycles are the best candidates for a parallel query. For example, joins of large tables, substantial aggregations, and sorting of large result sets are good candidates". Of course SQL Server will only parallelize the execution plan if the machine has multiple processors.

But after all is said and done - how the queries are coded, and how the database is designed, can have as much - or more - effect as the speed of your hardware. I've seen queries go from hours to minutes with simple code tweaks and the addition of useful indexes.

And since this is supposed to be a CPU forum....dual core CPU's might perform better. :) 
Related resources
August 25, 2005 11:34:06 PM

If the lack of redundancy is acceptable, a raid 0 array will perform better since queries on large db's will generate a good amount of writes for scratch files and Raid 5 isn't best for writes.

Mike.

<font color=blue>Outside of a dog, a book is man's best friend. Inside the dog its too dark to read.
-- Groucho Marx</font color=blue>
August 27, 2005 5:40:30 PM

I checked with the programmer and she said that queries are not paralleled. Maybe the queries can be tweaked but as I understand they're extremely complex mathematically, and considering it took months to write them, I don't know if it's worth hiring a guy to rewrite them.

As it turns out I'm also limited to $1K-1.3K for the whole rig. So now I'm stumped whether I should invest in 2 GBs of RAM, second HD (for RAID 0, as redundancy is not necessary (redundant, you might say)), or a second core, in case there is some software parallelization.

I think I'll stick to an Athlon since the Opteron requires ECC/RE RAM (right?).

What about the Athlon not being able to use 4 modules at DDR400, has that been fixed?

<b>Behold, Mine anger and My fury shall be poured out upon this place upon man and upon beast and upon the trees of the field and upon the fruit of the land and it shall burn and shall not be quenched<P ID="edit"><FONT SIZE=-1><EM>Edited by ytoledano on 08/27/05 03:43 PM.</EM></FONT></P>
August 27, 2005 9:42:12 PM

It can use 4 sticks of RAM, but at 2T command rate (0-5% loss in performance).

I would invest to memory, some 4GB with 64bit OS.

But it depends how random those queries are ?
If queries go thru DB in certain order, you don't need fastest HDD's (RAM is filled as you progress).
But if queries are random, there is going to be lower cache hit rate, so highspeed HDD would help a lot.

ps. that 64bit OS is a must, it's much faster with SQL queries (if your app supports it).

<font color=red>"We can't solve problems by using the same kind of thinking we used when we created them."
- Albert Einstein</font color=red><P ID="edit"><FONT SIZE=-1><EM>Edited by HansGruber on 08/28/05 00:48 AM.</EM></FONT></P>
August 27, 2005 10:45:39 PM

For now I'll go with the 3800+ Venice and 2*1 GB 2.5-3-3-7 Roswill RAM. These 2 I'm ordering from Newegg.

The queries go over data gathered over time but don't write to any tables, at least nothing more than a couple hundred cells. About storage, will NCQ improve performance, and is it supported on most 939 boards?

I also need advice on a decent 939 board with SATA RAID (I've seen most have) and having VGA out is a plus. I also don't wanna go too much over budget and I still haven't picked an HD.

<b>Behold, Mine anger and My fury shall be poured out upon this place upon man and upon beast and upon the trees of the field and upon the fruit of the land and it shall burn and shall not be quenched
August 28, 2005 5:02:37 AM

I would get 1MB L2 instead of 512kB that's in venice core.
It should help quite a lot with SQL.

<font color=red>"We can't solve problems by using the same kind of thinking we used when we created them."
- Albert Einstein</font color=red>
August 29, 2005 9:57:56 PM

From someone who has one of the fastest SQL servers on the net. If you want to do it right the first time:

Intel E7520 chipset
Dual Nacona's
Raid 5 with 5x SCSI 320 15K RPM
4GB ~ 16GB onboard, use the memory mirror option for higher level of reliability.


<A HREF="http://www.xtremesystems.org" target="_new">www.xtremesystems.org&lt;/A>
August 29, 2005 10:57:57 PM

Judging from all these reviews, I can't see you having the fastest server with Xeons on it:

http://www.tomshardware.com/cpu/20030422/opteron-17.htm...
http://www.anandtech.com/IT/showdoc.aspx?i=1982&p=8
http://www.anandtech.com/linux/showdoc.aspx?i=2163
http://www.eweek.com/article2/0,1759,1403690,00.asp
http://www.tecchannel.de/server/hardware/402167/
http://www.open-mag.com/9086339824.shtml
http://www.sudhian.com/showdocs.cfm?aid=487
http://lists.debian.org/debian-amd64/2004/07/msg00286.h...

The good thing about Opteron servers is that they scale a lot better whit more processors on it. ;) 

I can post more and more reviews, but I ain't doing something like that (google is your friend :D  )

My Beloved Rig:

ATHLON 64 FX 55
2X1024 CORSAIR XMX XPERT MODULES
MSI K8N DIAMOND (SLI)
2 MSI 6800 ULTRA (SLI MODE)
OCZ POWERSTREAM 600W PSU<P ID="edit"><FONT SIZE=-1><EM>Edited by Bullshitter on 08/29/05 07:01 PM.</EM></FONT></P>
August 30, 2005 4:14:44 AM

Dont worry fugger, your beloved Intel will have dual core servers soon. They wont be as good as dual core opterons, but hey, you wont be able to see that anyhow.
August 30, 2005 1:17:05 PM

Quote:
I checked with the programmer and she said that queries are not paralleled. Maybe the queries can be tweaked but as I understand they're extremely complex mathematically, and considering it took months to write them, I don't know if it's worth hiring a guy to rewrite them.

The actual QUERY is not paralleled, but the SQL could analyse the query and separate it in multiple individual task.

For example, if a query request to merge data from 2 table. The SQL server can open both table simultaneously by using sending the reading command to 2 different CPU/thread.

I'm 99% sure that a DUAL-CORE system will give you a good boost in performance for the same price (compared to single core system).

NOTE : I know SQL because I program in MySQL/PHP and I also access ORACLE databases with my PHP scripts once in a while.

-
GA-K8NF-9 / <b><font color=green>Athlon 64 3200+</font color=green> @ 3800+</b>
Infineon DDR400 (CL2.5) 2x512Megs
<font color=green>GeForce 6600GT 128Megs</font color=green>
<A HREF="http://www.getfirefox.com" target="_new">Get Firefox!</A>
August 30, 2005 1:37:08 PM

the road, i just read your last statement



is there a way to contact you for support for PHP / MySQL coding help?

I'm relatively new at doing it. but I've already got an application in place for work based upon it. but the second part of it has me stumped. maybe a professional could give me an idea.

so far I've written a program that does this with an SQL database and PHP / web front end. we're a recruiting / temp agency.

it is a contractor / workee placement and trackgin system that keeps a complete history of who / where / when / we placed all workers with given payrates and pertinent information.basically all the end user has to do is click a button on workers profile enter in a pay code and all the necessary documentation is created for faxing and submittal to corporate. it also saves the information and allows for retreival anytime. i'm proud of it. but i've ran into a few little problems.

a feature i added that allows us to pull up the actual paid hours per week always times out. that particular table alone is over 1gig. since we employ approximately 5,000 / year and have history in there for over 5 years. the data is quite large.

I've tried several basic SQL queries, but they always time out because it'sjsut taking too long to find the information. i know it DOES work because it will return the start of all the pay info, but then times out and just takes a LONG time.

I also know it's not the hardware since it's a real server rig with 2 x XEON's w/ hyperthread @ 2.6ghz. 4gb RAM, and the SCSI ahrd drives in RAID 5.
August 30, 2005 2:07:16 PM

I'm not an EXPERT in database. It's more a hobby, but I read a lot and help friends/co-worker build simple DB.

I would recommend you to split your table. You could easily keep in your main table only the data related to the last months (based on your need). So, based on user input, if the requested data is not requiring old data, you will only query to main table, if your user need historical data, then you would open both table or only the "old stuff" table.

Another thing to check and make sure is how your table is indexed. You can speed up your queries by indexing the most often used fields. This take some HDD space, but this can significally improve performance on large table. But, I know there is side effects to that, but I'm not sure about all the PROS/CONS.

I think your best bet, would be to automatically move old record in a new table that would contain only old record. This process could be done via an automatic script, or you could add a step upen record addition... Something like this :

AddNewRec($new_record_info);
MoveRecToHistory($how_long);
GetRecord($parameters);

This way, everytime someone add a record, your DB would automatically move old data to the history based on your "how-long" you want to keep data in your history.

And you could use a safer approach, by running the MoveRecToHistory($how_long), every day around 00:00, this way data would always be in the correct table even during vacation/holidays/week-end/etc...

Hope this gave you good tips/ideas.

-
GA-K8NF-9 / <b><font color=green>Athlon 64 3200+</font color=green> @ 3800+</b>
Infineon DDR400 (CL2.5) 2x512Megs
<font color=green>GeForce 6600GT 128Megs</font color=green>
<A HREF="http://www.getfirefox.com" target="_new">Get Firefox!</A>
August 30, 2005 2:27:51 PM

indices (or indexes, depending on who/what you're asking :lol:  ) Would be my first suggestion.

I'm unfamiliar with MySQL and PHP, but I'm pretty experienced at optimizing queries generally though, as I've been working with Interbase/Firebird for a few years now.

Things that affect query execution times, in my experience (in order from most important to least important):
1)What Indices are defined
2)Query Design
3)Hardware

That said, it's possible to make truly terribly designed queries that take ages to run on even a well-designed DB.

If you can get Mysql to tell you what Query PLAN it's trying to use, that's usually a good place to start - see if it's using any "natural" joins. That almost always means an index would be handy.

If you like you could post/PM me a brief overview of the database design, and a problematic query and I'd have a look, but I wouldn't be able to respond all that quickly of course. :smile:

---
<font color=red>"Life is <i>not</i> like a box of chocolates. It's more like a jar of jalapeńos - what you do today might burn your a<b></b>ss tommorrow."
August 31, 2005 5:20:15 AM

thanks for the tips guys...

I've done some twealking and it improved everything traumatically


it looked like the main problem was my original design of the query. it was basically doing redundant queries for some of the data.

I had basically for some obscure reason unkown to me done the query to pull out the array of the dates of all the pays, and then using that date, did another query in the same table to pull up the hours associated with that date.

the double query is what killed it.

what i've NOW done is one query to pull up both the feilds of Date and Hours, and just put a simple loop in there to display the next record in the query.

sometimes ya try too hard and do things and overcomplicate things. simplifying the query and just using the loop was the easiest method for some reason I thoujght it didnt originally work and used the long ass double query
August 31, 2005 7:06:14 AM

>and it improved everything traumatically

Oh dear.. sorried to hear that LOL.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
August 31, 2005 7:29:44 AM

A little drama, a little trauma, so who's to know?

BTW, I had a crush on a girl in HS named Linda Pasternak, but she wouldn't even give me the time of day. That was traumatic. I wonder if she's related to m.
<P ID="edit"><FONT SIZE=-1><EM>Edited by endyen on 08/31/05 03:34 AM.</EM></FONT></P>
August 31, 2005 10:06:51 AM

Well, if you look at the amount of posts in the 'Archive' forums, you'll see that we seem to be overdue for some maintenance in that area.

I guess there's probably been a server upgrade so it's happier with the increased amount, but all the same, it would probably help quite a bit :smile:

---
<font color=red>"Life is <i>not</i> like a box of chocolates. It's more like a jar of jalapeńos - what you do today might burn your a<b></b>ss tommorrow."
August 31, 2005 12:38:58 PM

no relation

and i'm a very traumatic dramatic sorta git

but SQL / PHP is killing me. i'm hours away from deadline on this thing and I can't get one function working

need to upload a file to a directory and reference it in the SQL Database so that when certain profile is selected, that persons image is displayed
August 31, 2005 3:47:17 PM

Can't you save the image or a link to the image in the table so it is part of the query?

I'm assuming that when a operation requests a certain profile it pulls:
1. First Name
2. Last Name
3. Address
4. ...
5. Picture

Its been awhile since I've done DB stuff but I know you can pull images. It seems like it would be equivalent to a query you run on New Egg. The Item picture as well as Item Details are displayed.
August 31, 2005 3:57:38 PM

very simillar, i figured it out without even touching the database so no worries.

since every profile hasa unique identification number in our systems, when the user uploads an image file, it automatically renames the file to their unique ID and stores it in a directory on the server for it.

now when someone prints off a specific report, the image is printed inline automatically. i don't use anything in the SQL except the Unique id of the prfile and it finds the image
August 31, 2005 4:34:07 PM

Glad you figured it out.

I don't think you could afford my consulting fees. :evil:  :smile:
August 31, 2005 5:57:53 PM

heck, i couldnt afford [-peep-]... nor would ym company pay it

I wrote this program because it would help out our jobs here. not because i'm paid to do it. i'm still being only paid my "office admin" salary
August 31, 2005 8:49:26 PM

=) i'm not sure he could afford that...having a 1-1.5k budget lol. Have you got some updated sites with info where the xeons are outperforming similarly scaled opterons? I'm all ears if you have some. I don't take anybodies word on anything, so you'll need to back it up with real data.

F@H:
AMD: [64 3000+][2500+][2400+][2000+][1.3][366]
Intel: [X 3.0x3][P4 3.0x2][P4-M 2.4][P4 1.4]

"...and i'm not gay" RX8 -Greatest Quote of ALL Time
August 31, 2005 8:58:51 PM

I'm sorry you're going to make me cry... Raid 0 on a sql server? I hope this is for a Web Site where you keep track of people's stats for CS:S or something...Cuz otherwise i might have to smack U. If the mobo you picked up was a nf4 asus board, you could buy 3 hds and put them in raid 5 (at the expense of the cpu...unless you went with an x2 4200 or something...[asus nf4 boards have built in raid 5]). To put together a server for under 2000 is hard... Unless it's a gaming server, or a school project.

The more i think about it...You said this is not on a LAN...What good is a database if it's not on a LAN...Well, it could be used for querying over a WAN...but that's hardly better then a LAN lol. It must be a stand-alone machine, where one person sits and inputs data from printed db's. God why... Say it ain't so?! More data please, allay my fears!

F@H:
AMD: [64 3000+][2500+][2400+][2000+][1.3][366]
Intel: [X 3.0x3][P4 3.0x2][P4-M 2.4][P4 1.4]

"...and i'm not gay" RX8 -Greatest Quote of ALL Time
September 1, 2005 9:54:27 PM

Budget oriented,

820 D processor
P5WD2 mobo (supports raid 5)
a couple of sata drives (3+)
2GB of DDR2

Pretty close to my old setup other than I used raptor 10K drives in raid 5.

SQL is all about IO and machine bandwidth, not cpu power.

Chances are my site hs 500+ online and go do a search. My search times are less than 1/10 of a second and I have a very busy forum.


<A HREF="http://www.xtremesystems.org" target="_new">www.xtremesystems.org&lt;/A>
September 2, 2005 2:06:52 PM

Heh that's better heh. Yea i've been out to your site a few times, i have it linked in my favs. I haven't posted out there though (er i don't think i have....maybe i did). I'm limited while at work by my bw, so i usually can't tell if a site is running slow or not. That's the problem with colleges (and community colleges...), you're sharing your bw with god knows how many people at a time.

How'd you like those raptors (74s?) in raid 5? Did u actually use onboard raid 5? Or did you buy a card? I haven't been terribly impressed by the onboard stuff so far. At least with the boards i've used.

F@H:
AMD: [64 3000+ (down)][2500+][2400+][2000+][1.3][366]
Intel: [X 3.0x3][X 2.8x2][P4 3.0x2][P4-M 2.4][P4 1.4x5][P4 1.3]

"...and i'm not gay" RX8 -Greatest Quote of ALL Time
September 5, 2005 11:51:03 AM

The raptors are nice in raid and very dependable so far. Huge improvement in database read times with 10K and 15K drives. You would have to spend a lot to get 15K U320 SCSI setup w/ RAID controller compared to the SATA Raptor solution.

I have run both 4x raid 0 and 4x raid 5 on the onboard controller.

<A HREF="http://www.xtremesystems.org" target="_new">www.xtremesystems.org&lt;/A>
September 6, 2005 1:44:46 PM

Yea i haven't heard anyone complain about them so far (raptors). My little 36 has been a rock <knocks on wood>...and i travel fairly often with it, so it doesn't just sit around.

F@H:
AMD: [64 3000+ (down)][2500+][2400+][2000+][1.3][366]
Intel: [X 3.0x3][X 2.8x2][P4 3.0x2][P4-M 2.4][P4 1.4x5][P4 1.3]

"...and i'm not gay" RX8 -Greatest Quote of ALL Time
September 9, 2005 12:43:10 AM

First, I'm only home once every 2 weeks, so I couldn't get around to reply.
Second, a DB on a stand-alone workstation is good if the DB is static and only the queries change. Also, the DB comes on DVD and the queries can be recovered, so RAID 0 is an option.

For now, I'm taking the X2 3800+ with 2*1GB CL2.5 Roswill memory. I could take the 4000+ Venice, which is a bit less and features a 200Mhz high freq and double the cache, I wonder which is better for this sort of thing.
This is what I'm buying from Newegg - only the small expensive stuff which are worth buying abroad. The rest is not really dependant on these two.

If someone has a better idea, I'd like to know.

<b>Behold, Mine anger and My fury shall be poured out upon this place upon man and upon beast and upon the trees of the field and upon the fruit of the land and it shall burn and shall not be quenched
September 9, 2005 2:13:23 PM

It's cool, i'm on the road more then i'd like too...but not nearly as much as you are.

So security or size is an issue with this? Size probably or you wouldn't have said dvd. So is this like a logistics machine or something, where you're checking orders to make sure they got though...though every major vendor i've worked with has everything in databases off of either mainframes or unix boxes with lapdogs. We've used stand-alone stuff with things like pay stubs, where a high speed scanner looks like it's shuffling cards... Sorry you just have my curiosity up. =)

Sounds like a pretty sweet machine. So you know, going with 2gb may mean that you have to use 333 instead of 400mhz ram. There are ALOT of ram out there that won't do 400 because of the controller on the amd processors in a 2x1gb or a 4x512mb formation. There are a few reviews out there that have information about the big ram, but not many. The only one i remember off-hand that did well in a 2x1gb, 2 3 2 1T 200fsb (3200), was the mushkin high perf stuff. Runs about $300 from newegg. But i haven't used any formation of 2x1gb so i can't personally vouch for any of them. I have 1gb in 4 machines heh, and 4x512 in a xeon system (that does squat...but by god we hit our <i>standard</i>).

I'd stick with the x2, it's fairly cheap, and will perform pretty well for you. (better then the 4000+ on most of your db work, dependant on how the db was setup and how you access it).

edit: Also, i saw no one had recommended a mobo for you. The Asus nf4 boards all* support raid 0, 1, 0+1; The Deluxe and Premium boards support 5 as well, on the Sata controller only. Obviously the sata would somewhat limit you to sata drives. Almost all of the nf4 boards have some flavour of raid on them, raid 5 is a little more rare. There are only two companies i can think of that had it, with an nf4. I think the high-end dfi board had raid 5 available on it...

F@H:
AMD: [64 3000+ (down)][2500+][2400+][2000+][1.3][366]
Intel: [X 3.0x3][X 2.8x2][P4 3.0x2][P4-M 2.4][P4 1.4x5][P4 1.3]

"...and i'm not gay" RX8 -Greatest Quote of ALL Time<P ID="edit"><FONT SIZE=-1><EM>Edited by apesoccer on 09/09/05 10:25 AM.</EM></FONT></P>
September 9, 2005 7:18:09 PM

"On the road"? More like "in the army" :(  I'm home thu 16:00 through sun morning though only once every 2 weeks.

Security is a big issue, hence the stand-alone-ness.

I thought taking 2*1GB instead of 4*512 solved the 333Mhz problem the hammers have, are you sure of this?

'bout RAID, I can always use Windows' built in RAID support, THG has a great guide on how to enable it (exists only in 2K Server\Advanced Server). I've been using it's mirroring on a computer and I'm very pleased. But that probably won't be necessary since RAID 5 is a bit too expensive for me and 0 comes with most mobos.

<b>Behold, Mine anger and My fury shall be poured out upon this place upon man and upon beast and upon the trees of the field and upon the fruit of the land and it shall burn and shall not be quenched
September 9, 2005 8:04:14 PM

Well, this has sort of been an on-going debate here. Since none of us actually have 2 - 1gb sticks, we can't verify or denigh either way. So, we've spend a good deal of time looking at tech review sites. Where there is very very little reviews about 2 gb, most of which are in the 4x512 variety. And the only ones capable of 400mhz, so far, in a 4x512 are tccd chips, and at a downgraded 2T over the 1T timings. There are a lot of people who are making claims that 2x1gb will work at 400, but not a single person has been able to come up with a review, or a screenshot of their own ram running at those speeds. I've come up with one review, and it was to that mushkin stuff i quoted to you earlier. Hrm...<A HREF="http://forumz.tomshardware.com/hardware/modules.php?nam..." target="_new">linkage (discussion)</A>...<A HREF="http://forums.rojakpot.com/showthread.php?p=215193" target="_new">Makes statement that the reviewed Mushkins ran at 1T</A>...<A HREF="http://www.rojakpot.com/default.aspx?location=3&var1=22..." target="_new">Link to Mushkin 2x1GB review...</A>...shew and their you have it...if you do find more information specifically regarding 2x1gb, by all means pass it on. I'll do the same. There have been several <i>discussions</i> on this issue heh. Do a ssearch in the memory section.

F@H:
AMD: [64 3000+][2500+][2400+][2000+][1.3][366]
Intel: [X 3.0x4][X 2.8x2][P4 3.0x2][P4-M 2.4][P4 1.3]

"...and i'm not gay" RX8 -Greatest Quote of ALL Time
September 11, 2005 3:04:44 PM

Intel CPU's have MUCH WORSE memory performance than AMD, so what the hell is that "AMD sucks at 2T" crap ?

<font color=red>"We can't solve problems by using the same kind of thinking we used when we created them."
- Albert Einstein</font color=red>
September 12, 2005 3:10:14 PM

That "crap" is the FACT that running 233 fsb @ 1T is FASTER (ie does more work) then 250 fsb at 2T. Noob. Do some research. Every site that has done a review on the 64bit 939 chips says that 1T timings are the MOST important factor in the timing's for AMD processors. When OC'ing you can get a higher bus speed by going to 2T, however, you still hit a wall before you overcome the loss you incur by changing from 1T. I'm not saying you suddenly lose 50% of your memory speed/bw...i'm saying its the most important timing change you can make. Again...do a little research before jumping to conclusions.

F@H:
AMD: [64 3000+][2500+][2400+][2000+][1.3][366]
Intel: [X 3.0x4][X 2.8x2][P4 3.0x2][P4-M 2.4][P4 1.3]

"...and i'm not gay" RX8 -Greatest Quote of ALL Time
September 12, 2005 3:18:47 PM

That was true for xp's maybe...i don't know...i haven't oc'd a xp. I'm sorry but you're just wrong in regards to 64's.

Quote:
<A HREF="http://www.anandtech.com/memory/showdoc.aspx?i=2226&p=2..." target="_new">A Command Rate of 1T is significantly faster than memory running at 2T.</A>


edit: Just reread the last two responses i sent out...heh. I'm not going to take them back, but i think i might have gotten a lil sand in my vagina...god i need some sleep.

F@H:
AMD: [64 3000+][2500+][2400+][2000+][1.3][366]
Intel: [X 3.0x4][X 2.8x2][P4 3.0x2][P4-M 2.4][P4 1.3]

"...and i'm not gay" RX8 -Greatest Quote of ALL Time<P ID="edit"><FONT SIZE=-1><EM>Edited by apesoccer on 09/12/05 11:25 AM.</EM></FONT></P>
September 13, 2005 4:33:46 AM

I was comparing it to INTEL.

So the verdict is: AMD has much better memory performance than Intel, it does not matter if AMD has 1T or 2T.
AMD can keep it's core working, because RAM is not lagging behind so much, as it is lagging on Intel.

<font color=red>"We can't solve problems by using the same kind of thinking we used when we created them."
- Albert Einstein</font color=red><P ID="edit"><FONT SIZE=-1><EM>Edited by HansGruber on 09/13/05 07:55 AM.</EM></FONT></P>
September 13, 2005 4:57:04 AM

I wonder how Intel does with 4 sticks of RAM, and with over 2GB ?

<font color=red>"We can't solve problems by using the same kind of thinking we used when we created them."
- Albert Einstein</font color=red>
September 13, 2005 8:33:07 PM

http://www.xtremesystems.org/forums/showthread.php?t=68...

2GB 4x 512MB 400+ FSB 1:1

I have 8GB in one of my boxes, 2GB x 4

Works fine.

Bus bandwidth, Intel wins. If you got some benches you ran yourself we would love to see what you can muster up since I know you cannot backup what you preach. It takes good hardware and good ram in the first place, you cant throw crap in a box and expect it to turn to gold.

<A HREF="http://www.xtremesystems.org" target="_new">www.xtremesystems.org&lt;/A>
September 13, 2005 8:45:26 PM

FUGGER, you do know that, internet is full of reviews that proof the fact, that AMD performs much better.
It's especially true with dual core CPU's, in which Intel is suffering because of lack of bandwith, and because the lack of low latency memory access.

edit: My grammar etc sucks. so do your best to understand this correctly.. :lol: 

<font color=red>"We can't solve problems by using the same kind of thinking we used when we created them."
- Albert Einstein</font color=red><P ID="edit"><FONT SIZE=-1><EM>Edited by HansGruber on 09/13/05 11:55 PM.</EM></FONT></P>
September 13, 2005 8:58:59 PM

You have not been paying attention.

Find any review and notice one obvious.

Intel DDR2 will be set to SPD super slack timing (4-4-4-15 or 5-5-5-15) and to top that off it will be using a ratio.

Notice on the AMD rig it is benched against, it is using 2-2-2-5 timing.

It takes the same effort to set timing on the intel as the AMD. But in benchmarks it truely hurts Intel performance.

I know you may be able to find one review with tight timing but 99% are done the way I described above. DO the reviewers do this to make AMD look better? Are they too stupid to know how to change from SPD on the intel system even though they did it a million times before? Maybe they wanted AMD to fare better in bandwidth related tests?

The heat sink crap was more bs.

Intel has made decent changes to the new dual cores such as 1066FSB and 2x2MB cache. The .065 drops the voltages and a modified layout shoud improve thermal disipation.

<A HREF="http://www.xtremesystems.org" target="_new">www.xtremesystems.org&lt;/A>
September 13, 2005 9:17:49 PM

It's the performance in real world applications that matter, and NOT some "OMG, i have poor real world performance with meaningless numbers".

Quote:
The heat sink crap was more bs.

ROFL.

Get a life and stop lying. :lol: 

<font color=red>"We can't solve problems by using the same kind of thinking we used when we created them."
- Albert Einstein</font color=red>
September 13, 2005 10:00:58 PM

"It's the performance in real world applications that matter, and NOT some "OMG, i have poor real world performance with meaningless numbers".

then why do you use Sandra a synthetic benchmark as a good rule of measure for bandwidth? Sure the on die controller helps latency and that inflates the synthetic. Kick over to Everest and the roles reverse and Intel takes the lead. Still synthetic tests but different test algo.

You know I am strong in bandwidth, I come from two bandwidth companies. SQL queries is also something I am familiar with.

I am sure stability is also a factor the originator of this thread is seeking.

<A HREF="http://www.xtremesystems.org" target="_new">www.xtremesystems.org&lt;/A>
September 14, 2005 2:47:11 PM

Quote:
Intel CPU's have MUCH WORSE memory performance than AMD, so what the hell is that "AMD sucks at 2T" crap ?

Ok...first off, I'm comparing amd to itself, in it's regular form and in a reduced form. You wanted to know why Amd sucks at 2T...so i answered you, since you felt the need to jump on my case. My answer obviously revolves around the fact that running 1T timings are almost always going to be better then running 2T timings. It makes no statement whatsoever about amd being better or worse then Intel, which is what you're trying to make this in to. Troll.

If you had just wanted to know why 1T runs better then 2T, you would have stated the question differently. Don't try to change your reasons now, everything you've said is already out there.

F@H:
AMD: [64 3000+][2500+][2400+][2000+][1.3][366]
Intel: [X 3.0x4][X 2.8x2][P4 3.0x2][P4-M 2.4][P4 1.3]

"...and i'm not gay" RX8 -Greatest Quote of ALL Time
!