WHOA!!! running 2.635MHZ!

[Rock Hydra]

Whoa...steady there killer. You're almost pushing 3 MHz...Too bad my Gameboy is more powerful. No LL Ram, memory dividers, special cooling or additional hardware required.

 

[GRIdpOOL]

Ok... that memory is rated at cas2 at that speed. You will see better performance if you adjust your timings. You should be able to see 215mhz at C2.5. Unfortunately I don't think you are ready to overclock until you do the research. It takes years of practice to be a good system builder with the knowhow to OC. Pure clock speed doesn't mean you have a fast PC. With the Athlon64 that is more evident than with other systems as the memory controller on-chip needs low latency and timings to achieve the bandwidth desired.

 

[stevty2889]

Nice overclock but it's 2.635Ghz, not Mhz...or 2635Mhz...

 

[Shimmishim]

y'all are so mean.

you guys know he means 2.635 GHZ not MHZ

stop pulling his chain. he admits he's a "newb"

how about trying to help him out instead of bashing him?

....

like someone else said, you probably have your mem ratio set to auto in the bios and it's causing the 3:2 divider to kick in hence why your ram is running a lot slower.... and to make matters worse, you're using a half mult which technically doesn't exist and even further slows down your memory...

310 htt with 3/2 divider makes your ram somewhere around 200 mhz but you say it's 202 so that's about right...

but with the 8.5 mult... your ram is actually...

8.5 x 310 = 2635

since for some reason it rounds up, the board thinks your chip is actually running a 9x mult

so 2635/9 = 292.8 or 293 HTT

then you apply the 3/2 divider and you get a mem speed of around 195 mhz/htt

so your ram is running at 195 mhz not 202...

check this thread out

http://www.xtremesystems.org/forums/showthread.php?t=41595&page=1&pp=25

 

[flexy]

Originally posted by: csflipfreakz
hey guys...am running it now with my corsair twinx1024-3200C2PT memory at 2.5-3-3-6 202mhz and 3x ht, 8.5x multi, 310fsb!! what do you guys think?? can i push it further?? or any suggestions on improving it??

what CPU ? I ASSUME decent cooling and i ASSUME that the max for your CPU *could* be around 2600. Thats what i can get out of my 3500.
With core voltage 1.52. Its not excellent....but still better than 2400 or what some others are getting.

If you're staring with OC....read, read, read, forums, FAQs, ask. I know it sounds corny but overclocking is an "art" and mastering it might involve more than what many people think.

as a rule:

You have to know your CPUs max, and you need to know your memory's max.
BOTH things are totally different.

You have to use low multi and high HTT to find out your max memory w/ your timings....i'd say 290 could be more realistic than shooting for 300 and beyond.

THEN you need to test your max CPU within a HTT which is in your KNOWN safe memory range....eg. if you know your mem does max. 290 then use this or better lower and a multi...to find the max for your CPU.

If you test CPU max you need to stay withing safe/low memory range - if you want to check the highest memory "FSB" then do that only within your known safe/stable CPU frequency.

ONCE you find the max for both (and dont let one thing influence the other) you can combine the two. If you found your CPU's max is 2600 and your max. mem is 290...well then use a multi combo which comes close to the CPU..both doesnt exceed ANY of either CPU or mem.

read up on that...if you need to know more.

The common testing tools are:

MEMTEST for memory. ONLY for memory.

Prime95 or OCCT for CPU. Depending on config also tests some memory.

Both should work for hours without a problem to achieve a stable system.

 

[flexy]

also, stay away from the half multis like 8.5 or something.5....they are not real. Use whole numbers.
Otherwise the system will use altered frequencies/HTT and not what you set in bios.

Btw. its kinda hard to give advice since you dont list your specs, board etc..

 

[Amaroque]

Originally posted by: GRIdpOOL
the memory controller on-chip needs low latency and timings to achieve the bandwidth desired.

Not true. RAM latency doesn't have a great impact on the A64 memory subsystem because of the efficiency of the integrated memory controller. I highly suggest you, and anyone else interested in how the A64 interacts with your memory subsystem to read this.

 

[flexy]

Originally posted by: Amaroque

Originally posted by: GRIdpOOL
I highly suggest you, and anyone else interested in how the A64 interacts with your memory subsystem to read this.

i dont agree with zebos conclusion who used a lot of cpu intensive apps (raytracers, fft calculations, mp3 encoding etc.) to try to show the non importance of bandwidth.

A tool like Super PI or Prime or whatever does HARDLY utilize memory bandwidth.

Correct me if i am wrong.

And his list shows SOME gains in games and some apps where you can see quite some gains. It always depends on the app/game. He should use more game benches and maybe apps like video editing etc.

My opinion is a system should be balanced - means fast CPU, fast memory (and fast gfx . Everything will add up.

 

[Amaroque]

I'm not saying that LL RAM has no impact at all. I'm just saying that it has a very minimal impact. Basically, your money would be better spent on a faster stock CPU, or a better video card, ect...

You will see a higher return on your money spending it on hardware other then RAM. For example, you could buy a 74 GB Raptor with the money you saved on RAM. I can guarantee that you will see a much greater perceptible performance increase from a Raptor then from LL RAM.

 

[sangyup81]

Originally posted by: Amaroque
I'm not saying that LL RAM has no impact at all. I'm just saying that it has a very minimal impact. Basically, your money would be better spent on a faster stock CPU, or a better video card, ect...

You will see a higher return on your money spending it on hardware other then RAM. For example, you could buy a 74 GB Raptor with the money you saved on RAM. I can guarantee that you will see a much greater perceptible performance increase from a Raptor then from LL RAM.

Originally posted by: Zebo
Half-life 2 1024x768 in FPS
PC2300 - 3-3-3 1T- 113.3
PC3000 - 3-3-3 1T- 120.9
PC3000 - 2-2-2 1T- 131.6
PC4700 - 3-4-3 1T- 133.3
PC3800 - 2.5-2-2-1T - 134.4

I see a 10.7fps jump. I wouldn't call that minimal.

 

[Dadofamunky]

Originally posted by: Amaroque
I'm not saying that LL RAM has no impact at all. I'm just saying that it has a very minimal impact. Basically, your money would be better spent on a faster stock CPU, or a better video card, ect...

You will see a higher return on your money spending it on hardware other then RAM. For example, you could buy a 74 GB Raptor with the money you saved on RAM. I can guarantee that you will see a much greater perceptible performance increase from a Raptor then from LL RAM.

Of course, that didn't keep you from USING expensive RAM....

 

[Amaroque]

Originally posted by: Dadofamunky
Of course, that didn't keep you from USING expensive RAM....

I origionally bought the RAM for a Barton @ 2.4GHz. If I was buying RAM spicifically for an A64, I'd get something a bit more modest though.

But I admit it, I usually buy the expensive hardware, just to see for myself what all comotion is about.

 

[ChineseDemocracyGNR]

Originally posted by: sangyup81

Originally posted by: Zebo
Half-life 2 1024x768 in FPS
PC2300 - 3-3-3 1T- 113.3
PC3000 - 3-3-3 1T- 120.9
PC3000 - 2-2-2 1T- 131.6
PC4700 - 3-4-3 1T- 133.3
PC3800 - 2.5-2-2-1T - 134.4

I see a 10.7fps jump. I wouldn't call that minimal.

It's definately not minimal, I think the point is that you can get 10.7 (or much more) frames per second if you spent the money elsewhere (eg the video card).

 

[sxr7171]

Originally posted by: piroroadkill

Originally posted by: csflipfreakz
no the core speed is 2.635mhz from original 1.8mhz...memory is a little overclocked to 202mhz from original 200mhz....im jus trying to overclock as high as possible...also i dont get how memory is 202mhz...isnt 200mhz equivalent to ddr400?? bcuz when i boot with my oc settings, it says its on ddr266...shouldn't that be equal to 133mhz??

Never overclock again, infact:

Lock and ban.

My 386 ran at about 15 times the speed of that POS.

Had to do it. Someone had to.

Funny stuff!

 

[CheesePoofs]

Originally posted by: ChineseDemocracyGNR

Originally posted by: sangyup81

Originally posted by: Zebo
Half-life 2 1024x768 in FPS
PC2300 - 3-3-3 1T- 113.3
PC3000 - 3-3-3 1T- 120.9
PC3000 - 2-2-2 1T- 131.6
PC4700 - 3-4-3 1T- 133.3
PC3800 - 2.5-2-2-1T - 134.4

I see a 10.7fps jump. I wouldn't call that minimal.

It's definately not minimal, I think the point is that you can get 10.7 (or much more) frames per second if you spent the money elsewhere (eg the video card).

Its minimal if you consider that those 10.7fps just cost you about $200. Put that $200 towards the graphics card, and you will see a 10.7fps increase using the next highest res.

 

[Amaroque]

Originally posted by: CheesePoofs

Originally posted by: ChineseDemocracyGNR

It's definately not minimal, I think the point is that you can get 10.7 (or much more) frames per second if you spent the money elsewhere (eg the video card).

Its minimal if you consider that those 10.7fps just cost you about $200. Put that $200 towards the graphics card, and you will see a 10.7fps increase using the next highest res.

Exactly. If you mainly use your computer for gaming, a video card would be money much better spent. Or like my example above, a Raptor would be money better spent on a more well rounded computer (not just a gaming machine).

 

[Dadofamunky]

Originally posted by: Amaroque

Originally posted by: Dadofamunky
Of course, that didn't keep you from USING expensive RAM....

I origionally bought the RAM for a Barton @ 2.4GHz. If I was buying RAM spicifically for an A64, I'd get something a bit more modest though.

But I admit it, I usually buy the expensive hardware, just to see for myself what all comotion is about.

Yeah, fair enough. I do feel that buying cheap components is usually the wrong move....

 

[Shenkoa]

2.635 MHz, WOW THATS GREAT!

:thumbsup: