Using 4 Memory Slots (A64)

[Rodz]

I am sure this has already been asked but I couldn't seem to find via search.

Anyway here goes:

Will my A64 Winchester suffer any huge performance losses by using up all the 4 memory slots to reach 2 GB DDR? (512 x 4)

Good day and thank you in advance.

 

[Bar81]

No, you'll never notice the difference in the real world between 1T and 2T nor even between DDR400 and DDR333 if you're so inclined. Enjoy.

 

[bjc112]

http://forums.anandtech.com/messageview...atid=28&threadid=1577496&enterthread=y

Should give you an idea.

 

[Rodz]

So basically, the difference would be:

Using 2 memory slots only would give me the possibility of a 1T command rate

while

Using 4 memory slots would automatically give me a maximum of 2T command rate.

Am I getting it at all?

 

[Bar81]

Exactly, and no performance difference you'll ever notice.

 

[Rodz]

Clears up a lot.

Thank you very much for all your help!

 

[tdawg]

what about cas latency differences? say one dual channel kit running at cas 2 and another kit running at cas 2.5 or 3? any difference? and filling all four banks will cut ddr400 to ddr333, right?

 

[Bar81]

The difference between DDR400 and DDR333 on an A64 is 0-3% (that's right you read zero) I've done my own testing on this matter. I think DDR333 or 400 depends on the mobo implementation strangely. I know AMD recommends DDR333 with all four banks filled but people have run DDR400 in that scenario.

As to CAS latency, that also is largely irrelevant on an A64 platform; you won't ever notice a real world difference. In fact even on an Intel platform the only place you really saw a real world detectable gain was Quake III, and maybe some encoding apps???

 

[tdawg]

Thanks, Bar. You've been very helpful in this thread.

 

[mwmorph]

Originally posted by: Bar81
The difference between DDR400 and DDR333 on an A64 is 0-3% (that's right you read zero) I've done my own testing on this matter. I think DDR333 or 400 depends on the mobo implementation strangely. I know AMD recommends DDR333 with all four banks filled but people have run DDR400 in that scenario.

As to CAS latency, that also is largely irrelevant on an A64 platform; you won't ever notice a real world difference. In fact even on an Intel platform the only place you really saw a real world detectable gain was Quake III, and maybe some encoding apps???

A few corrections:
PC2700 speed is max for 4 stick 2 sided ram
PC3200 speed is max for 4 stick 1 sided ram

2T is 5% slower than 1T in most cases
PC3200 aand PC2100 is 7% difference in no synthetic benchmarks

Cas 2 is 5% faster than Cas 3

Venice core changed this. Check if your Cpu is venie(ex, manufactured in the last month or two). Use CPU-Z to be sure.

You will se a difference but it wont be that much.

Total, you lose maybe 3% in less memory intensive apps and maybe 7% in more memory intensive apps(of course SisSoft Sandra is an entirely different matter. 30-40%loss is more likely, but it's just a synthetic benchmark so it dosent matter that much)

the above is of course witth similar cas latency ram and pc2700 at 2T

if you had Cas 2 ram and bumped it down to Cas 3 ram at pc2700 ant 2T, speeds will decrease more around 10%

 

[Bar81]

Sorry, that's simply not accurate. Where you're pulling those numbers from I'd like to know?

 

[theMan]

yeah, if ur not obcessed with benchmarks dont worry about it. whats a few seconds? especially if you need the extra GB.

 

[mwmorph]

Originally posted by: Bar81
Sorry, that's simply not accurate. Where you're pulling those numbers from I'd like to know?

search the forums and Anandtech.com itself

From Anandtech.com article

Socket 939 Dual-Channel motherboards were found to exhibit a very wide performance difference between a Command Rate setting of 1T and a setting of 2T. The impact on memory bandwidth is dramatic between these 2 settings. In SiSoft Sandra 2004 standard buffered Memory Benchmarks, a 1T command rate showed a Sandra bandwidth of 6000 Mb/sec, while a 2T rate with the same 2 DIMMs in Dual-Channel mode was only 4800 Mb/sec. This is a huge difference in memory bandwidth and the Command Rate setting definitely impacts performance test results on Socket 939 motherboards. All AnandTech benchmarks were run at a Command Rate setting of 1T. This includes all benchmarks that were run in the CPU tests, as all benchmarks were rerun in the CPU tests as soon as we had verified the performance impact of Command Rate settings.

From Forums.Anandtech.Com

***********MAY 1 2005**************

Edit: Since we have not found anything around the net or at reviewers sites dispite thier 1T "massive" benefit claims our members have done some tests!!! Wooo woooo.. I am leaving out synthetics since we already know the crazy story they like to tell; Sandra a whopping 35%, everest a massive 25% etc etc etc. Screw that! We don't want to freighten people into running 1T But we rather give a realistic performance benefit they see moving from 2T to 1T all things being equal.

GuitarDaddy System: CPU - Winch@2548Mhz | Memory- 2-3-2-7 X @232Mhz
SuperPI 8m 1T- 6:03m
SuperPI 8m 2T- 6:09m
1T is 2.4% faster
Aguamark3 1T- 58156
Aguamark3 2T- 57985
1T is 0.3% faster
3dmark 03 1T- 6908
3dmark 03 2T- 6872
1T is 0.5% faster
HL2 1T- 97.5fps
HL2 2T- 95.5fps
1T is 2.2% faster
X2 the threat 1T- 87.4fps
X2 the threat 2T- 85.8fps
1T is 1.9% faster
Doom3 1T- 85.9fps
Doom3 2T- 84.3fps
1T is 1.9% faster

Not looking good for 1T supposed "have to have" (thanks GD)
-------------------------------------------------------------------------

Sir Avalon System: CPU - Sempy128KB@1800Mhz | Memory- 2-2-2-10 X @225Mhz
SuperPi 1M 1T- 49s
SuperPi 1M 2T- 50s
1T is 2.0% faster
HL2 (coast 05, 1024x768, 2xAA/4xAF) 1T- 83.5fps
HL2 (coast 05, 1024x768, 2xAA/4xAF) 2T- 81.0fps
1T is 3.0% faster
Doom 3 (1024x768, 2xAA/4xAF) 1T- 74.8fps
Doom 3 (1024x768, 2xAA/4xAF) 2T- 73.5fps
1T is 1.2% faster

Avalon's using a 128KB LVL2 chip which probably accounts for the "massive" 3% improvment in one game
-------------------------------------------------------------
Insomniak System: CPU - Winch@2400Mhz | Memory- 2.5-3-3-8 X @200Mhz

Battlefield Vietnam 1T- 106.5FPS
Battlefield Vietnam 2T- 102.1FPS
1T is 4.3% faster*
Source VST 1T- 164.7
Source VST 2T- 158.3
1T is 4% faster*
Super Pi 8M 1T- 6:49
Super Pi 8M 2T- 6:58
1T is 1.4% faster
3dMark03 1T- 9258
3dMark03 2T- 9187
1T is 0.8% faster
Far Cry 1T- 105.2
Far Cry 2T- 103.4
1T is 1.7% faster

Insomniak definity showed the most pronounced differences so far, had a couple games in the 4% range. As and average though he's still less than 2%.
----------------------------------------------------------------------------

blckgrffn System: CPU - 754NC@2600Mhz | Memory- 2-3-3-5 X @195Mhz
Doom3 1T- 103.1
Doom3 2T- 99.2
1T is 3.7% faster
3DMark2001SE 1T- 22543
3DMark2001SE 2T- 21970
1T is 2.6% faster

DUVIE (from anotherthread) System: CPU - Winch@2640Mhz | Memory- u/k timings X @275Mhz
No data Duvie does the math for us: 1T vs. 2T
cinebench 2003 = 0%
superpi 2mb = <3%
prime95 benc 2048K = >1%
Povray 3.5 chess2 = 0%
3dsmax7 rendering = 1%
winrar (file compression) = 7%
the TMPGenc was 1%
"I dont see it ppl...tell me apps this will show up.... "

Benchmarked by various anandtech members. Obviously this is a limited application benchmark. Most tend to be non memory intensive apps like Video Card Benchmarks(games, 3dmark) and CPU Benchmarks(superpi,etc) so it isnt too accurate a test if memory intensive speed which would explain the low differences HL2 tends to be more CPU/Mem Intensive(physocs, ai) as opposed to the more video limited benchies like DOOMIII and 3DMark. Win RAR shows 7% difference which is max, but i remember memory intensive programs like Photo/Video Editing have even bigger differences.

From an Forums.Anandtech.Com Post

Came across a Corsair Whitepaper that has some interesting info. They used Doom3 time demo on 3500+ 939 setup. Summary of conclusions pertaining to the topic at hand:
1) 2-2-2-5 1T vs. 2-2-2-5 2T, -3.29% performance loss in Doom3 timedemo, more so than any other timing change.

2) 2-2-2-5 1T vs. 3-3-3-8 2T, -17% in Sisuck Sandra, 90% attributable to CPC change.

From an Forums.Anandtech.Com Post

Zebo, I had run a couple gaming benchmarks in my Sempron 2800+ 90nm thread, and found that dropping your memory command rate to 2T was the equivalent of a 400mhz drop in CPU speed at 10x7. At least I think it was about 400mhz. It's in my thread somewhere, but I remember it being definitely noticeable.

Posted on hardware-help.org

After reading Toms hardware article on memory timings I stand corrected. It[Cas 2] appears to gain about 6% over CAS 3.

 

[Bar81]

Okay, so you quoted Sandra numbers several times. Big deal as it's not real world. And you posted some numbers by people whose testing acumen I can't vouch for to disprove my own numbers. Sorry, not nearly good enough. Besides the numbers don't even support your claims. I stand by my numbers. On the A64 1T and 2T settings will exhibit no real world noticeable difference due to the A64 architecture and it's onboard memory controller. Similarly the difference between DDR333 and DDR400 is 0-3%.

Just so it's clear to take the most extreme example, pit a CAS2 DDR400 system against a CAS3 DDR333 system and you'll see anywhere between a 1-5% difference. Outside of a framerate counter no person in the real world will be able to tell the difference.

 

[mwmorph]

Originally posted by: Bar81
Okay, so you quoted Sandra numbers several times. Big deal as it's not real world. And you posted some numbers by people whose testing acumen I can't vouch for to disprove my own numbers. Sorry, not nearly good enough. Besides the numbers don't even support your claims. I stand by my numbers. On the A64 1T and 2T settings will exhibit no real world noticeable difference due to the A64 architecture and it's onboard memory controller. Similarly the difference between DDR333 and DDR400 is 0-3%.

Just so it's clear to take the most extreme example, pit a CAS2 DDR400 system against a CAS3 DDR333 system and you'll see anywhere between a 1-5% difference. Outside of a framerate counter no person in the real world will be able to tell the difference.

um... a lot of these benchers are some of the more trusted, senior and knowledgeable people on thes forums
DUVIE,Insomniak, Avalon have all been around a long time and are know for posting helpful accurate replies.

THG, a REPUTABLE, Major hardware revier(albeit a little biased toward some brands, but still MUCH better than almost every other site out there besides anand[shudder:xbit]) reports 6% from changing latencies from cas 2 to 3 over a wide array of tests(opengl games, sissoft and 3dmark2001(a primarily cpu limited test now)

as for results dont support my claims, uh... no sh!t man. These are, read my post, Game and CPU benchmarks. Of course there will be no difference in superpi and minimal difference in games. THEY ARE VIDEO AND CPU LIMITED BECAUSE THAT'S WHAT #DMARK(Video), Super pi(CPU), Games(Yet gain Video, HL2 is usually moe impaced by cpu than most games). Test Farcry. It is actually fairly memory dependant.

the math is 5% for 2t, 5% of cas and 1-2% for speed(linearly calculated from the memory matrix)

 

[Bar81]

Sorry, when your own numbers don't back up your claims your credibility goes right into the toilet. btw I never questioned the competence or integrity of the testers, I just simply stated that I don't know them; learn to read. Besides, they back ME up, so for the sake of argument they're the best testers in the world You really need to work on the whole supporting your argument thing, there are some kinks you need to work out.

Okay, so again let's take an even more extreme scenario and compare a DDR333 CAS3 2T system to one that has DDR400 CAS2 1T which is in itself a ridiculous scenario as we're leaving out the fact that in the real world a person running 2T is generally going to have more memory than one with 1T (and that will probably influence benchmarks) but for the sake of argument let's assume all else the same. Only then would you start approaching something that may, in certain programs, be a detectable difference *on the high end* of the 3-8% difference between the setups. But even then, remember you're looking at very specific apps. The majority of what's out there is handicapped by other factors given the extremely efficient nature of the A64 architecture and its onboard memory controller.

 

[mwmorph]

Originally posted by: Bar81
Sorry, when your own numbers don't back up your claims your credibility goes right into the toilet. btw I never questioned the competence or integrity of the testers, I just simply stated that I don't know them; learn to read. Besides, they back ME up, so for the sake of argument they're the best testers in the world You really need to work on the whole supporting your argument thing, there are some kinks you need to work out.

Okay, so again let's take an even more extreme scenario and compare a DDR333 CAS3 2T system to one that has DDR400 CAS2 1T which is in itself a ridiculous scenario as we're leaving out the fact that in the real world a person running 2T is generally going to have more memory than one with 1T (and that will probably influence benchmarks) but for the sake of argument let's assume all else the same. Only then would you start approaching something that may, in certain programs, be a detectable difference *on the high end* of the 3-8% difference between the setups. But even then, remember you're looking at very specific apps. The majority of what's out there is handicapped by other factors given the extremely efficient nature of the A64 architecture and its onboard memory controller.

read my posts. and see how you support me
me:
"You will se a difference but it wont be that much"


me:
"Total, you lose maybe 3% in less memory intensive apps and maybe 7% in more memory intensive apps
if you had Cas 2 ram and bumped it down to Cas 3 ram at pc2700 ant 2T, speeds will decrease more around 10%[referring to max speed obviosuly]"
you: after i post, you change answers multiple times
"on the high end* of the 3-8% difference between the setups."
you again
"[same comparison situation]you'll see anywhere between a 1-5% difference. Outside of a framerate counter no person in the real world will be able to tell the difference."
----------------------------------------------------------------------------------------------
um. okay.. reword me and try to prove me wrong with it. also, change your answers.

me:
"Most tend to be non memory intensive apps like Video Card Benchmarks(games, 3dmark) and CPU Benchmarks(superpi,etc) so it isnt too accurate a test if memory intensive speed which would explain the low differences"
youafter i posted)
"The majority of what's out there is handicapped by other factors given the extremely efficient nature of the A64 architecture and its onboard memory controller."
----------------------------------------------------------------------------------------
wonder what i said multiple times?

Me:

"PC3200 aand PC2100 is 7% difference in no synthetic benchmarks"
"1-2% for speed(linearly calculated from the memory matrix)[referring to decrease form 400mhz to 333mhz""
You: after i posted my math(second quote)
"the difference between DDR333 and DDR400 is 0-3%."

If you hadn't noticed, all your posts support my answers. you just dont read it right.

you:
"No, you'll never notice the difference in the real world between 1T and 2T nor even between DDR400 and DDR333 if you're so inclined."
again: you cahnge your answers to suit you
"pit a CAS2 DDR400 system against a CAS3 DDR333 system and you'll see anywhere between a 1-5% difference"
again: whoops, changes it again!:Q
"the 3-8% difference between the setups"
Senator Kerry, is that you?

you really should develop some reading comprehension and stand behing your findings.
5% is huge! 8% as you stated later is even bigger. the upgrade from a 9800pro(last gen) to a 6600gt(this gen) is around 5-8% in most cases. I dont think it is your place to question my credibility.

Sorry if i sound like an ass, but if you just left the personal attacks at home..

 

[Bar81]

Take a reading comprehension course. You're way too braindead to have a discussion with.

 

[rise]

1t vs 2t on an a64 means next to nothing in real world apps.

 

[BeakerChem]

Now I am really confused. I almost had it there before everyone got mad.

Tell me if I am getting this right: (All comparisons are for an Asus A8N-SLI Deluxe motherboard with an FX-55 proc if that last bit matters any, which I don't think it does. Also, all mem is PC3200 DDR400)

--
2 sticks of 2-3-2-5-T1 (1 GB each) would run faster than 4 sticks of 2-2-2-5-T1 (512MB each)

Reasoning: With filling two extra slots, the T value drops to T2, and the 400MHz speed drops to 333MHz.


--
2 sticks of 2-3-2-5-T1 (1GB) would run faster than 4 sticks of the same

Reasoning: Same as above. But 4 GB worth of memory would allow more junk to be run at the same time.


--
Corrections?

 

[mwmorph]

"Just so it's clear to take the most extreme example, pit a CAS2 DDR400 system against a CAS3 DDR333 system and you'll see anywhere between a 1-5% difference."
"Only then would you start approaching something that may, in certain programs, be a detectable difference *on the high end* of the 3-8% difference between the setups."
"No, you'll never notice the difference in the real world between 1T and 2T nor even between DDR400 and DDR333 if you're so inclined"

hmm. okay.
you know what you're right. everyone else MUST be wrong because no one can possibly be smarter than YOU. in fact, in reality 0=5 and 5=8 because whatever idiot invented numbers got it all wrong.

lol you are hilarious. exhibiting the classic behaviors of trolling.
Phase 1. Make up stuff/change information
Phase 2 Pretend that anything contrary to what you said is stupid
phase 3: Personal insults

I know your kind. I tried to let you off the hook on this one by avoiding your bull as long as possible but i guess the saying still goes "Dont Feed The Troll"

I'm not actually making anything up, though he can be helpful most of the time, he just dosent back down no matter how wrong he is. look at a lot of previous posts by bar, espically the "via go pop" FUD post he made. I guess he just goes around from time to time spewing wrong info, criticizing others that dont agree.

seriously, 8% difference is somewhat noticible Rodz, it will actually be like 3-5 fps in most games, also 2gb is not need for xp or games. 1gb is entirely for all games/xp. unless you need photoshop or video editing. if you od, by all means go for it. 2gb at 2t is still going to be faster than 1gb a 1t for all apps, just that you dont get as much linear increase as if using 2x1gb sticks vs 2x 513mb sticks.

 

[Bar81]

Phase 4. Ignore idiot.

 

[BeakerChem]

Originally posted by: mwmorph

seriously, 8% difference is somewhat noticible Rodz, it will actually be like 3-5 fps in most games, also 2gb is not need for xp or games. 1gb is entirely for all games/xp. unless you need photoshop or video editing. if you od, by all means go for it. 2gb at 2t is still going to be faster than 1gb a 1t for all apps, just that you dont get as much linear increase as if using 2x1gb sticks vs 2x 513mb sticks.

This may be a silly question, and outside of this discussion, but what about:

2 GB from sticks of 1024MB at CAS 2-3-2-5-T1 vs. 1 GB from 2x512MB sticks at CAS 2-2-2-5-T1

Thanks!

 

[mwmorph]

Originally posted by: BeakerChem

Originally posted by: mwmorph

seriously, 8% difference is somewhat noticible Rodz, it will actually be like 3-5 fps in most games, also 2gb is not need for xp or games. 1gb is entirely for all games/xp. unless you need photoshop or video editing. if you od, by all means go for it. 2gb at 2t is still going to be faster than 1gb a 1t for all apps, just that you dont get as much linear increase as if using 2x1gb sticks vs 2x 513mb sticks.

This may be a silly question, and outside of this discussion, but what about:

2 GB from sticks of 1024MB at CAS 2-3-2-5-T1 vs. 1 GB from 2x512MB sticks at CAS 2-2-2-5-T1

Thanks!

2GB at 2-3-2-5 1T will be faster than 1gb at 2-2-2-5 1T.
Cas latency(1st number) is more important than the other 3 numbers. command rate si more important than cas latency

"When we see what a RAM?s CAS latency is, it is 4 numbers, such as 2-2-2-5, which correspond with CAS ? tRCD ? tRP ? tRAS. You should be able to change the CAS Latency in your BIOS. In most BIOSes, it can be found under the ?advanced chipset? menu, though it can be found in different areas in different BIOSes.

CAS stands for Column Address Strobe. This is the number of memory cycles that pass between the time a column is requested from the active page and the time the data is ready to send across the bus. This number is usually 2, 2.5, and 3, on DDR memory. This is actually the last part to come into effect.

RAS to CAS Delay is referred to as tRCD. This is the delay in memory cycles between the time a row is activated and when data within the row can be requested. This only happens when data is not on the active row.

tRP is the time for RAS Precharge. This is the time in memory cycles that is required to clear out the active row out of the cache, before a new row can be requested. In other words, it?s the time it takes for the memory to stop accessing one row and start accessing another. Once again this only takes place it the data is not in the active row.

tRAS refers to the minimum time that a row must remain active before a new row can be activated in each memory bank. A new row can not be opened until the minimum amount of time has passed. If there is more than one bank on memory, this will help the performance of the tRAS. If there is only one active bank, then the need to change rows is guaranteed, and if there is more than one bank with memory, then there is only half the chance that there will be a need to change rows. In turn, the tRAS will only come into effect half the time. The tRP and tRAS together are often referred to as the Row Cycle time, because they happen together.

Let's compare this to a library. Let's say you want a book about Florida. Let's say the librarly is putting books away, and you ask for the book on Florida. The librarian would have to get off the ladder, move the ladder to the travel section, and then get back up on the ladder. This would compare with the tRP. The tRCD would be the time the librarian is at the right shelf and right row of the bookcase. The tRAS would be the time when the librarian finds the books about Florida, and when he finds the specific book you want. The CAS latency in this case would be the time from when he takes the book off the shelf, and the time it takes him to walk back to you. If he walks fast, it might be 2; if he walks slower, the number would be larger.

Let's say the library is putting books on Florida back; then the time required would be a lot less. All of the moving of the ladder would not be needed. This could happen in the memory if the data needed is on the active row. I hope this helps you understand how memory latency works. "

read this:How memory latencies affect speed
While for many programs, there isnt too much of a difference between 2-2-2-5 and 2.5-4-4-8, winRAR cut 1 minute, from 8:30 to 7:34(13%) for compressing a 360mb file, 3Dmark2001se(a cpu/memory limited benchmark, now that all the newer cards have more than enough power to kill this 4 year old benchie) goes from 15483 to 16287(5%), DOOM 3 is the difference between 58 and 62 so latencies and command rates are important.

hope this helps!

 

[BeakerChem]

Wow.


That is the most stunningly complete and helpful answer I have ever seen to any question.

Thanks!!!