• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

1:1 VS 5:4...How much difference does it really make??? ***added test***

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Yes, you are misunderstanding me. I apologize for my lack of clarity.

Most pc3500s and lower will not do 5:4 at 280+ fsb on P4 systems utilising 2.4C with the lower multiplier. I am not arguing that 5:4 is worse than 1:1. All I'm saying is that in order to run high fsb, no pc3500 or lower can do even 5:4, or require very very high vdimm (3+) and some luck.

So, pc4000 which can do ddr500+ should definately be able to do 5:4 at very high fsb. Hell, my pc4000s can do DDR550! Please do not be-little me about simple mathematics. I think I have done my research to know what these numbers all mean, and never said DDR600 was even a possibily. 🙂
 
For the sake of further study and clearer understanding of what this means I will be adding o more tst situations....

Add#1

2.6@3.033ghz 5:4 cas 2,5,2,2 CPC enabled

Add#2

2.6@3.033ghz 5:4 cas 2.5,7,3,3 CPC disabled (matching specs of the 466mhs ddr)


Lets see the real difference and then conlcude if someone who dumps old pc3200 cas 2 ram form much more expensive pc3500 to pc3700 stuff is worth the real world gains....
 
Originally posted by: Thor86
Yes, you are misunderstanding me. I apologize for my lack of clarity.

Most pc3500s and lower will not do 5:4 at 280+ fsb on P4 systems utilising 2.4C with the lower multiplier. I am not arguing that 5:4 is worse than 1:1. All I'm saying is that in order to run high fsb, no pc3500 or lower can do even 5:4, or require very very high vdimm (3+) and some luck.

So, pc4000 which can do ddr500+ should definately be able to do 5:4 at very high fsb. Hell, my pc4000s can do DDR550! Please do not be-little me about simple mathematics. I think I have done my research to know what these numbers all mean, and never said DDR600 was even a possibily. 🙂
Click to expand...

That may be true but you are missing the point of this article!!!

It was about whether or not one should sacrifice higher fsb for the simple fact 1:1 ran out on WHATEVR particular pcXXXX the person was running.

If your pc4000 ran out at 1:1 at 275fsb on you 2.4c for a total of 3.3ghzand maybe doing this with cas 2.5 to cas 3 levels...YET>>>You could drop the raio to 5:4 and run that same ram at 440mhz ddr which it clearly would do and maybe even do cas 2 to cas 2.5 and yield even 10 more fsb then as shown above in similar apps this system would be the better overall performer....

That is the main message in this thread.....


The side mesage Oldfart is bringing up is more of an economic one in which he is stating the numbers show such a small gap even if running large differences in memory speed that it just doesn't always justify getting the expensive ram when dropping to a 3:2 (by the way which thugs does with his 2.4c at 280+ fsb and still gets life out of his HyperX pc3000) and getting same overall or slightly better OC will slightly better, equal, or slightly worse results....

Money talks ppl, as most AMD fans here will atest to...
 
Maybe I'm just not experienced in Intel technology, but that's rather difficult to understand Duvie... I don't know if it's the way it's laid out or what... I'm still having a bit of a hard time figuring out what exactly you were trying to prove.

To disspell the myth of a handful of knuckleheads who think cause they run 1:1 though the cas timings may be weak means it equal to a much higher clocked cpu
Click to expand...
Does this mean you're trying to prove running 1:1 at a lower clock speed isn't as beneficial as running a different ratio with a higher clock speed?
 
I think this is an important question to clear up, and Duvie thanks so much for doing this test and sharing your results.

I think your results are consistent with a similar test done by XbitLabs (link). What I take from both tests is that, while results vary depending on the application, in general 1:1 with loose timings is better (by a small margin) than 5:4 or 3:2 with tight timings at the same FSB. If you can up your FSB appreciably by going to a ratio, though, you definitely should do so, especially if you can use tight timings when doing so.

If any of you experts disagree with this, I would be curious to learn from you. This is an area that I am trying to learn more about.

Last point: Duvie what video card are you using? Those are some sorry 3dMark01 scores, and they suggest to me that the video card is the limit there rather than the CPU or memory. I wonder if the results would have been different if you had used a video card that made the CPU and memory the limiting factor. For instance, in the XbitLabs test, the 2.4C overclocked to 3GHz with DDR500 at 2.5-4-4-5 (1:1 ratio) scored 17,254, while the same setup with DDR400 at 2-2-2-5 (5:4 ratio) scored 16,872. In tests with Quake 3, Unreal Tournament, and Serious Sam, the difference was even more pronounced, although the percentage differences are still not that big. Here is the difference between those two setups in the XbitLabs tests:

Benchmark Performance gain with DDR500 SDRAM (at 2.5-4-4-5, 1:1 ratio, versus DDR400 at 2-2-2-5, 5:4 ratio)
Business Winstone 2002 0.6%
Multimedia Content Creation Winstone 2003 1.83%
Data Compression, WinRAR 3.0, Best, sec 2.83%
MPEG-4 Encoding, FlasK 0.78.39/DiVX 5.02, fps 0.67%
3DMark03, Default 0.33%
3DMark03, Default, CPU score 1.56%
3DMark2001 SE, Default 2.26%
Quake3 (four), High Quality, 1024x768x32 3.98%
Unreal Tournament 2003 (dm-antalus) 3.95%
Serious Sam 2 (The Grand Cathedral) 2.45%
CINEMA 4D, CINEBENCH 2003, Raytracing, CB 0%
CINEMA 4D, CINEBENCH 2003, Shading, CB 1.25%
CINEMA 4D, CINEBENCH 2003, Lighting, CB 1.75%
On average 1.8%
 
Originally posted by: Jeff7181
Maybe I'm just not experienced in Intel technology, but that's rather difficult to understand Duvie... I don't know if it's the way it's laid out or what... I'm still having a bit of a hard time figuring out what exactly you were trying to prove.

To disspell the myth of a handful of knuckleheads who think cause they run 1:1 though the cas timings may be weak means it equal to a much higher clocked cpu
Click to expand...
Does this mean you're trying to prove running 1:1 at a lower clock speed isn't as beneficial as running a different ratio with a higher clock speed?
Click to expand...



In part Yes, and in part you are right this is more of something discussed more in INtel P4 based threads....

I see many ppl and I can find you links where ppl who think since they are running 1:1 ratio fsb/ram even if the timings are loose or looser then they would be if they ran at 5:4 ratio that it makes it equivalent to a higher overclock that did in fact use a 5:4 ratio...I have seen people state sometimes a 1;1 ratio system was equal to a system 300-400mhz higher in overall speed simply cause that person ran a 5:4 ratio...

NOw you undersand that when the dual channel system is running at a 5:4 ratio once again we are back to the days of the p4 capable of handling more bandwidth then the memory....Ie rambus versus single channel ddr days...

What I am showing here in these particular apps is that ram speed is...

a) Not all it is cracked up to be
b) tight timings can often make more of a difference then overall mhz speed or 100's of pts in Sandra Mem scoress
c) can limit your oc and ultimately limit you from a better performing and higher oc'd system.


If I followed many ppl 's logic I would have stopped at 3.033ghz with 466mhz ddr 2.5,7,3,3 and thought it was the best performance I could get. Instead I dropped it to 5:4 and ran it out until the cpu truly became the limiting factor and thus ran my memory at slower speed but much faster timings....In the end you can see i was better off...

So until Ishow ppl what this real world difference really is in these type of situations many will continue to belive iti s 1:1 or nothing. That 5:4 is something to look down on...To spend 100's of more dollars then needed to garner a few % pts of actually real world performance.

We as hardware users need to get past the Sandra mem scores, the other sissoft benches, PCmark, etc and see what it really means....
 
Originally posted by: FPSguy
I think this is an important question to clear up, and Duvie thanks so much for doing this test and sharing your results.

I think your results are consistent with a similar test done by XbitLabs (<a class=ftalternatingbarlinklarge href="http://www.xbitlabs.com/articles/memory/print/p4-mem.html" target=blank>link</A>). What I take from both tests is that, while results vary depending on the application, in general 1:1 with loose timings is better (by a small margin) than 5:4 or 3:2 with tight timings <STRONG>at the same FSB</STRONG>. If you can up your FSB appreciably by going to a ratio, though, you definitely should do so, especially if you can use tight timings when doing so.

If any of you experts disagree with this, I would be curious to learn from you. This is an area that I am trying to learn more about.

Last point: Duvie what video card are you using? Those are some sorry 3dMark01 scores, and they suggest to me that the video card is the limit there rather than the CPU or memory. I wonder if the results would have been different if you had used a video card that made the CPU and memory the limiting factor. For instance, in the XbitLabs test, the 2.4C overclocked to 3GHz with DDR500 at 2.5-4-4-5 (1:1 ratio) scored 17,254, while the same setup with DDR400 at 2-2-2-5 (5:4 ratio) scored 16,872. In tests with Quake 3, Unreal Tournament, and Serious Sam, the difference was even more pronounced, although the percentage differences are still not that big. Here is the difference between those two setups in the XbitLabs tests:

Benchmark Performance gain with DDR500 SDRAM (at 2.5-4-4-5, 1:1 ratio, versus DDR400 at 2-2-2-5, 5:4 ratio)
Business Winstone 2002 <STRONG>0.6%</STRONG>
Multimedia Content Creation Winstone 2003 <STRONG>1.83%</STRONG>
Data Compression, WinRAR 3.0, Best, sec <STRONG>2.83%</STRONG>
MPEG-4 Encoding, FlasK 0.78.39/DiVX 5.02, fps <STRONG>0.67%</STRONG>
3DMark03, Default <STRONG>0.33%</STRONG>
3DMark03, Default, CPU score <STRONG>1.56%</STRONG>
3DMark2001 SE, Default <STRONG>2.26%</STRONG>
Quake3 (four), High Quality, 1024x768x32 <STRONG>3.98%</STRONG>
Unreal Tournament 2003 (dm-antalus) <STRONG>3.95%</STRONG>
Serious Sam 2 (The Grand Cathedral) <STRONG>2.45%</STRONG>
CINEMA 4D, CINEBENCH 2003, Raytracing, CB <STRONG>0%</STRONG>
CINEMA 4D, CINEBENCH 2003, Shading, CB <STRONG>1.25%
</STRONG>CINEMA 4D, CINEBENCH 2003, Lighting, CB <STRONG>1.75%</STRONG>
On average <STRONG>1.8%</STRONG>
Click to expand...


Sorry, I am not a gamer....This is a Radeon 8500AIW DV card no opitmizations done in settings which I have as all optimal quality...The radeon AIW DV card has much lower core and mem speeds then regular radeon 8500...I have 230/190 versus 275/275 rtail and 250/250 oem. I have used radtweak in the past and bumped it up to 250/230 (the max it would do) and went aorund 10500.....

I like your results as well....It is clear a person saving some bucks and sticking with that older pc3200 could have had a system almost neglible to the 1:1 system for a fraction of the cost.....Once I add the 2 above test systems I think I will see same results.

 
I'm probably one of those people you are referring to... although my logic was that if you could run the processor at 3 Ghz at 1:1 with tight timings, it would probably perform better, or the same as running 3.2 Ghz at 5:4 with the same timings. But since the RAM you have doesn't seem to be capable of running the same timings at 1:1 and 5:4, I guess we won't know =)
 
Originally posted by: Jeff7181
I'm probably one of those people you are referring to... although my logic was that if you could run the processor at 3 Ghz at 1:1 with tight timings, it would probably perform better, or the same as running 3.2 Ghz at 5:4 with the same timings. But since the RAM you have doesn't seem to be capable of running the same timings at 1:1 and 5:4, I guess we won't know =)
Click to expand...


Actually NO....

I have ran the 3.2ghz at same cas 2.5,3,3,7 and it was still faster then the 3.033ghz cpu 1:1 with same timings...


Also...I will run the 3.033ghz at 5:4 and you will see even at same speed the difference is going to be small as fpsguy shows. I will take a guess but be pretty sure that a 3.0ghz at 1:1 with any timings will never be equal to the 3.2ghz unless you may be running some gaming apps which I admit I have not tested and in the past seem to like the bandwidth a bit more....

PLus I couls only run 1:1 to 2.9ghz(223fsb for 446ddr) with cas 2,7,3,3 with CPC enabled and trust me the scores were worse then the 466mhz ddr so that would only prove the point even more.....I was limited in my graphing ability but I have tested this and the pcu at many speeds tiings and ratios. I have 5 pages of notes I put together for the past 3 days...
 
excellent experiment backed with excellent data Duvie.:beer:

oldfart: Nice to have someone else backing me up on this for a change! As you know, I've been fighting this myth for a long time. I've been posting about this here, ABXzone, and BE forums. All you have to do is actually test it and the results are right there in front of you. The guys who are out there pimping overpriced "high speed" memory test with SiSoft mem bench, Aida and the like since these synthetic apps show huge gains but have no relevance to real world apps.

That's exactly why i'm saving my money for OCZ PC3500 Platinum or OCZ PC3700 Gold.
 
I have ran the 3.2ghz at same cas 2.5,3,3,7 and it was still faster then the 3.033ghz cpu 1:1 with same timings...
Click to expand...
See, it's all about things running in sync... keep in mind, this is just my theory because I don't have a P4 or anything to actually test it... by running in sync I don't mean 1:1... take a clock for example... the hour hand runs at 1/60th of the minute hand, and 1/360th of the second hand. Those are all running in sync... the hands "touch" at regular intervals. Now if you speed up the second hand and not either of the others, it may hit the minute hand and the hour hand more frequently, but what we're looking for is all the hands to meet at once, that's when data can be transmitted. And in this situation, if you speed up the second hand say, 10%, all 3 won't meet as often... when the minute and hour hands meet, they may have to wait for the 2nd hand to come back half way around the face of the clock for all 3 to meet.

That might be a really bad analogy... some of you are probably looking at the computer screen thinking "no wonder he spends the majority of his day at a computer" 😀 It make sense to me in my head, but it's hard to put it on paper... I'll work on it and see if I can get some kinda chart, or animation, or math forumulas to back up my theory. Maybe if I didn't sleep through Algebra/Geometry/Statistics in high school I might be able to put it on paper or on a graph easier =)
 
Originally posted by: Jeff7181
I have ran the 3.2ghz at same cas 2.5,3,3,7 and it was still faster then the 3.033ghz cpu 1:1 with same timings...
Click to expand...
See, it's all about things running in sync... keep in mind, this is just my theory because I don't have a P4 or anything to actually test it... by running in sync I don't mean 1:1... take a clock for example... the hour hand runs at 1/60th of the minute hand, and 1/360th of the second hand. Those are all running in sync... the hands "touch" at regular intervals. Now if you speed up the second hand and not either of the others, it may hit the minute hand and the hour hand more frequently, but what we're looking for is all the hands to meet at once, that's when data can be transmitted. And in this situation, if you speed up the second hand say, 10%, all 3 won't meet as often... when the minute and hour hands meet, they may have to wait for the 2nd hand to come back half way around the face of the clock for all 3 to meet.

That might be a really bad analogy... some of you are probably looking at the computer screen thinking "no wonder he spends the majority of his day at a computer" 😀 It make sense to me in my head, but it's hard to put it on paper... I'll work on it and see if I can get some kinda chart, or animation, or math forumulas to back up my theory. Maybe if I didn't sleep through Algebra/Geometry/Statistics in high school I might be able to put it on paper or on a graph easier =)
Click to expand...


NO sometimes speed is the factor and just raw cpu power....In gaming I bet the higher bandwidth makes more of a difference but just not in the programs above....

Also in synch is great and 1:1 is optimal as it perfectly mathes the cpu to the ram for optimal bandwidth throughput...1:1 should always be better when cas timings and cpu speed are the same. BUt timings also matter and if one has better timings and the programs ran are not bandwidth hungry lik most are not then timings will make more of an issue and the added bandwidth is all for not...

Wait until you see my added test it really shows the higher speed and nice Sandra Mem benches don't mean crap....




 
Originally posted by: Jeff7181
I have ran the 3.2ghz at same cas 2.5,3,3,7 and it was still faster then the 3.033ghz cpu 1:1 with same timings...
Click to expand...
See, it's all about things running in sync... keep in mind, this is just my theory because I don't have a P4 or anything to actually test it... by running in sync I don't mean 1:1... take a clock for example... the hour hand runs at 1/60th of the minute hand, and 1/360th of the second hand. Those are all running in sync... the hands "touch" at regular intervals. Now if you speed up the second hand and not either of the others, it may hit the minute hand and the hour hand more frequently, but what we're looking for is all the hands to meet at once, that's when data can be transmitted. And in this situation, if you speed up the second hand say, 10%, all 3 won't meet as often... when the minute and hour hands meet, they may have to wait for the 2nd hand to come back half way around the face of the clock for all 3 to meet.

That might be a really bad analogy... some of you are probably looking at the computer screen thinking "no wonder he spends the majority of his day at a computer" 😀 It make sense to me in my head, but it's hard to put it on paper... I'll work on it and see if I can get some kinda chart, or animation, or math forumulas to back up my theory. Maybe if I didn't sleep through Algebra/Geometry/Statistics in high school I might be able to put it on paper or on a graph easier =)
Click to expand...


Jeff, You bring up some good points but! the P4C rigs are soooo much different, in terms of how the memory is handled, to the AMD systems. Like Oldfart and Duvie have pointed out, you can search my threads as well. I stated this when I first got my P4C (from and AMD box) how the timings were not that much of a factor in comparison. It is all about the FSB.

BTW my scores I listed above were nearly identical with or w/o running the F@H client at the same time. This HT is amazing more and more.

Sunny129, I am not so sure it is worth it to get the Gold (I can't believe I just said that) unless you know your CPU will hit 260+FSB. I think I am lucky to have my Gold do 275 1:1 but my PC3500EL will not do 5:4 above 260 (w/o a volt mod). I am so tempted to do the volt mod to see if I could get the gold to do 300 1:1.
 
Originally posted by: orion7144
Sunny129, I am not so sure it is worth it to get the Gold (I can't believe I just said that) unless you know your CPU will hit 260+FSB. I think I am lucky to have my Gold do 275 1:1 but my PC3500EL will not do 5:4 above 260 (w/o a volt mod). I am so tempted to do the volt mod to see if I could get the gold to do 300 1:1.
Click to expand...

i take it you mean more than just a voltage adjustment in the BIOS when you refer to a volt mod? and why would the memory need more juice at 260+ fsb on a 5:4 divider? at 260fsb, memory is running at 416mhz, which is already slower than the 433mhz specification of PC3500 memory.
 
Is there any chance you could post a little more information regarding your set up and the files you used? If our systems are close enough, I'd like to test my 2.4C@3.2 w/DDR 533 on a few of them to see what kind of difference it makes. My rig is in my sig.
 
P4 2.6c
Abit IC7 (Bios version 14)
Enermax 430watt PSU
2x512mb PC3500 Geil Ultra Platinum
WinXP


What do you need to know for files???

TMPGenc version 2.5
I am using a Divx 5.02pro codec AVI....I can send you the template for encoding. HT enabled in enviromental.

BeSweet GUI ver0.6 Besweet v1.5b22
Converting a stripped Wav file from the Divx movie above to AC3 using AC3enc (built into Besweet)

Cinebench2003

Autocadd2002
Rendering a perspective scene of a 30MB file with photo raytrace (shadows on/cache on/best quality)

SuperPI ver1.1
Prime95 ver22.12.1

DVDshrink3.0beta
I took the movie Mummy Returns DVD rip (6.4gb film alone) and reudced it 39% to fit on 1 4.3gb disk.

WinRaR version 2.9
 
OK here are some quick results I got with my setup:
Test config:
P4 2.4C, HT-enabled
Abit IS7, BIOS 16
Radeon 8500LE @ 275/275 (8500 non-LE)
2x512 MB Hynix PC3200 (NO CPC, NO pseudo-GAT)
WinXP Pro



#1) RaynorWolfcastle's config: p4 2.4@3.3 GHz w/ 3:2 for 440MHz DDR CAS 2.5,7,3,3
#2) Duvie's config: p4 2.6@3.2ghz w/ 3:2 for 38mhz ddr cas 2,5,2,2 spd
#3) RaynorWolfcastle's config: p4 2.4@3.207 GHz w/ 1:1 for 535MHz DDR CAS 3,8,4,4
#4) Duvie's config: p4 2.6@3.2ghz w/ 5:4 for 394mhz ddr cas 2,7,3,3 CPC enabled (Bios version 14)

N/T = not tested


-------------------------3.3 GHz w/ 440ddr-------3.2ghz w/ 338ddr------3.2ghz w/ 567ddr------3.2ghz w/ 394ddr (CPC)

Sissoft MEM---------------5590's------------------------4800------------------------5790's---------------------5420's

Tmpgenc-------------------N/T-----------------------------2:18------------------------N/T--------------------------2:17

SuperPi 1M----------------0:43---------------------------0:44-------------------------0:43-----------------------0:43
superPI 2M---------------1:39----------------------------1:43-------------------------1:40-----------------------1:42
Prime95 (ver22)
1792k----------------------71.755------------------------73.320----------------------72.626--------------------73.137

Besweet Wav-AC3------N/T-----------------------------4:11------------------------N/T--------------------------4:04
DVDshrink 3.0------------N/T-----------------------------17:28----------------------N/T--------------------------17:32

Cinebench2003
single cp/multi------------331/393---------------------323/384-------------------322/380-------------------329/388
AutoCAD 2002------------N/T-----------------------------0:46-------------------------N/T-----------------------0:45

WinRAR---------------------N/T----------------------------3:08------------------------N/T-------------------------3:11



I should note that I had a bunch of programs running in the background while running this incl. Norton Antivirus Auto-Protect, Zone Alarm, and ActiveSync...

I own the Mummy Returns so I'll try to do the DVD-shrink later, if you PM me the config info for TMPGenc, BeSweet, and WinRAR incl. test file used (hopefully you're also using the Mummy returns for TMPGenc), I'll give them a try as well.

I'll try to test out 3.2 GHz w/5:4 later today 🙂
 
In tmpgenc and mummy returns lies your problem...It is an encoded Divx title from the movie with vbr.(high quality)...Then I took a small 2 minute clip removed by nanub. It will be hard to impossible to duplicate and to email unless you can take a 25,282kb file....

Then the actual Mummy returns is the dvd-shrink of the the tmpgenc of the entire movie after running 3+ hour vbr encode on it....

Lets pick another DVD title I can do....Do you have Gone in 60 seconds or Bourne Identity....Both of those are biger the 4.3gb for the movie alone and will need compression.

The divx thing is pretty specific to me, sorry!!!
 
Also that besweet file is the stripped wav file from the Divx-Mummy Returns then converted to ac3....

The winrar is obviously specific cadd files for me so that one will be hard to duplicate as well...


You pick one of those 2 files I suggested above (or another one) and we can run dvd2avi and strip the ac3 then run ac3 to wav for the reverse...then we can run dvdshrink as well...


Also results fall in line....make sure you look at my same speed same cas timings and same speed different cas timings cause I think it shows that as your does in some of the programs cas timings make all the difference and in others memory speed does...

 
Duvie, I'm looking at the results again...

What happens if you run 3.033ghz w/ 5:4 for 373mhz ddr cas 2,5,3,3 with CPC disabled?

And what happens if you run a 456 FSB 1:1 where you can still run the RAM at 2-7-3-3?

Or how bout 452 where you can run 1:1, GAT, and 2-7-3-3?
 
Hate to ask a potentially noobish question, but does the 5:4 with tighter timings > 1:1 with looser timings that duvie is testing only hold true for P4s or for athlons as well?
 
Originally posted by: Jeff7181
Duvie, I'm looking at the results again...

What happens if you run 3.033ghz w/ 5:4 for 373mhz ddr cas 2,5,3,3 with CPC disabled?

And what happens if you run a 456 FSB 1:1 where you can still run the RAM at 2-7-3-3?

Or how bout 452 where you can run 1:1, GAT, and 2-7-3-3?
Click to expand...


Damn good questions!!! Hehehe Wow, I would need a pretty big chart....

What I can tell you is with the cpc disabled the mem score changes down by about 30 pts in sandra and the tmpgenc test increases by 1 sec. No change in pi 1m and 1 sec diffeence in 2m. I may run more apps for you on this one...Thugsrook has already ran some test where he showed the cpc being enabled had a positive effect. My gut tells me it will beworse performance but the level may be quite small.

UT2003 (3.033ghz cas 5:4 373ddr w/ cas 2,5,3,3
640x480 res due to my Radeon8500dv and wanting to isolate the cpu and system and not limit with the vid card

---------------Gat enabled---------Gat disabled

Flyby------------186.32---------------185.80(-0.3%)
Botmatch-------82.46-----------------81.60(-1.1%)

In comparison the 3.033ghz 1:1 466ddr cas 2.5,8,3,3 did 186.82/83.65



Wrond data above I will correct....Actually I ran it was 223fsb(2.9ghz) that 2,7,3,3 could still pass memtest or 446ddr and in the app I ran TMPGenc it was 2:29 or 5sec more then the higher speed memory and higher overall clock....I would have to test more but the pattern suggest that the cpu speed will trump the cas timngs going from 2 to 2.5 by itself.

UT2003demo 185.45/81.42

Actually 446ddr was last place for GAT enabled as well.....so at 452ddr we took a drop to cas 2.5 and cpc disabled....

Tmpgenc was 2:27 or slightly faster, but tmpgenc obviously as seen in all the testing just doesn't care about memory speed as much as overall clock

UT2003 demo 185.90/81.42 very close to the cpc disabled 5:4 cas 2,5,3,3...so obviously the ddr speed made all the difference here and making up for a 90mhz shortfall




 
The reason I asked is because you were only gaining a little more speed by maxing out the RAM, and if you backed it down just a tad, you'd be able to run CAS 2, which is a GREAT benefit... and also run GAT (I'm not exactly sure what that does but I know it increases performance 😀 )
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top