New A64 3200+

[wicktron]

Originally posted by: JustStarting

Originally posted by: wicktron
To the air-cooling crowd:

Shuttle AN50R rocks. Built a system for a homie and got 2.4GHz at 1.7v, not quite as impressive as Shimm's 2.5GHz at 1.7v, but hey, it's on air 🙂. Air cooled with Zalman CNPS7000A-AlCu, temps in the upper 40's at load.

How high does Vdimm go??

2.9v without mods, up to 5v with mods (VDIMM runs through +5v rail). I wouldn't go past 3v on A64 though.
The reason a bunch of A64's are dying is because of high VDIMMs. I haven't heard of any cases of high VCORE deaths yet.

 

[joe2004]

Prometeia Mach I? Now I got it, OK.
I get 2.45 on air but using Sunon Vulcano and it is like on airport here. And SH-man is right, it does flies compared to my Pentium.

 

[Shimmishim]

Originally posted by: wicktron

Originally posted by: JustStarting

Originally posted by: wicktron
To the air-cooling crowd:

Shuttle AN50R rocks. Built a system for a homie and got 2.4GHz at 1.7v, not quite as impressive as Shimm's 2.5GHz at 1.7v, but hey, it's on air 🙂. Air cooled with Zalman CNPS7000A-AlCu, temps in the upper 40's at load.

How high does Vdimm go??

2.9v without mods, up to 5v with mods (VDIMM runs through +5v rail). I wouldn't go past 3v on A64 though.
The reason a bunch of A64's are dying is because of high VDIMMs. I haven't heard of any cases of high VCORE deaths yet.

the reason people's a64 are dying is because they are using the bios to change their voltage. so people are trying to set 2.9 vdimm as 3.1. what happens when people change the vdimm to 2.9 is is that they get 3.1 but they get crazy vdimm spikes. someone reported getting a spike of up to 3.8.

so the way to fix this is to set the vdimm to auto and use the vr's to adjust the voltage.

mine is set to exactly 3.0 right now but i've had it as high as 3.4 trying to get my bh-5's stable at 250 mhz @ 2-2-2-6

 

[wicktron]

Originally posted by: Shimmishim

Originally posted by: wicktron

Originally posted by: JustStarting

Originally posted by: wicktron
To the air-cooling crowd:

Shuttle AN50R rocks. Built a system for a homie and got 2.4GHz at 1.7v, not quite as impressive as Shimm's 2.5GHz at 1.7v, but hey, it's on air 🙂. Air cooled with Zalman CNPS7000A-AlCu, temps in the upper 40's at load.

How high does Vdimm go??

2.9v without mods, up to 5v with mods (VDIMM runs through +5v rail). I wouldn't go past 3v on A64 though.
The reason a bunch of A64's are dying is because of high VDIMMs. I haven't heard of any cases of high VCORE deaths yet.

the reason people's a64 are dying is because they are using the bios to change their voltage. so people are trying to set 2.9 vdimm as 3.1. what happens when people change the vdimm to 2.9 is is that they get 3.1 but they get crazy vdimm spikes. someone reported getting a spike of up to 3.8.

so the way to fix this is to set the vdimm to auto and use the vr's to adjust the voltage.

mine is set to exactly 3.0 right now but i've had it as high as 3.4 trying to get my bh-5's stable at 250 mhz @ 2-2-2-6

Thanks for clearing that up.

 

[Shimmishim]

no problem!

🙂

still trying to fix my corrupt video issue....

let's see if different drivers help...

 

[ntrights]

Overall is it faster then your 3.0@4.0 ?

 

[Dustswirl]

What is exactly the nature of the video corruption?

 

[joe2004]

What is exactly the nature of the video corruption?

That has nothing to do with video card and everything to do with the memory and onchip memory controller. I had the same problem that was easily solved by using less memory and tweaking the timmings.

 

[Shimmishim]

Originally posted by: ntrights
Overall is it faster then your 3.0@4.0 ?

yup! 🙂

 

[Shimmishim]

Originally posted by: joe2004

What is exactly the nature of the video corruption?

That has nothing to do with video card and everything to do with the memory and onchip memory controller. I had the same problem that was easily solved by using less memory and tweaking the timmings.

i'm using 256 megs of memory....

and in terms of timing... everything is set to auto in the bios...

 

[Shimmishim]

i noticed everytime that the screen went blocky the PWM temperature was upwards to 100C's.

so i took my big mother fan, set it so it's blowing on the mobo.... and prime95 is still running or running longer than it has before! 🙂

 

[sonoran]

Originally posted by: joe2004

What is exactly the nature of the video corruption?

That has nothing to do with video card and everything to do with the memory and onchip memory controller. I had the same problem that was easily solved by using less memory and tweaking the timmings.

Reading this, I can't help but think of how many people would be crucifying the P4 if its main selling point was supposedly the ability to use lots of memory in a flat address space, and the only way to get it to run stably was to "use less memory". Not meaning to start a flame war, but the irony seems a bit much to ignore. Or is the problem a poor MB design?

 

[Shimmishim]

Originally posted by: sonoran

Originally posted by: joe2004

What is exactly the nature of the video corruption?

That has nothing to do with video card and everything to do with the memory and onchip memory controller. I had the same problem that was easily solved by using less memory and tweaking the timmings.

Reading this, I can't help but think of how many people would be crucifying the P4 if its main selling point was supposedly the ability to use lots of memory in a flat address space, and the only way to get it to run stably was to "use less memory". Not meaning to start a flame war, but the irony seems a bit much to ignore. Or is the problem a poor MB design?

i dunno what the deal is with some of these a64 boards but yeah.... there are issues with ram... but then again lots of these issues can be fixed with bios updates.... there have only been 2 on the shuttle... i'm sure future ones will fix this! 🙂

 

[DAPUNISHER]

Timings make very little difference with the A64 as Tom's, Ace's, and Gamer's Planet? have shown so relaxing them to get higher ram speeds won't result in the hit the P4 takes when doing it.

 

[Duvie]

Originally posted by: Shimmishim

Originally posted by: ntrights
Overall is it faster then your 3.0@4.0 ?

yup! 🙂

What qualitative or quantitive data do you have to support this??? You said you never really liked the Asus P4p800 so you really didn't test on it....So is this just a general feeling as how snappy it feels in the desktop or in gaming alone??? I am wondering cause I hoped to get some feeling of what 2.5ghz meant in real world data.

I know at 3.5ghz I do quite well or beat similar A64 system oc'd to 2.3ghz so I would figure 500 more mhz on the P4 would still hold a lead on a 2.5ghz A64...however you may be just talking about gaming or snappiness in the OS.....

Too bad we have no real comparison data other then 3dmock.....🙁 It is not often we have ppl really jump cross platform from the two best and have some of the highest ocs on each.

 

[Shimmishim]

well if it makes you feel any better... i have my 3.0C and the p4p800 still...

if it'll make you happier, i'll run some tests...

what tests would you like to see???

give me a list and i'll do them....

 

[Shimmishim]

also solved my video problem...

i have two 80mm fans blow on the mosfets that the PWM measure and everything is fine!

so let the testing begin!

Aquamark
3dmark2k1
ut2k3
superpi
pi_fast
encoding

any other tests and links? duvie?

thanks!

 

[DAPUNISHER]

Throw in POVRAY and SETIbench Shim, that way I can see just how much the extra couple hundred mhz, 1:1 fsb:ram, and bigger cache scales things over my 3000+@2.3ghz 384DDR since I ran all the other benchies you listed aside from DIVX. Your card is faster than mine too so I can only guesstimate that part, so benchies like SETI and POVRAY will be apples to apples for me to see what the extra bandwidth and clockspeed could do for me.

 

[JustStarting]

Any idea what the PWM temp was?? How about system (case) temp? My IC7-MAX3 has a nice hardware monitor that shows PWM temps. I've seen ~ 10C above CPU temps while loaded at any given point in time.

see here.

The first shot is about 6C difference- not really loaded. That was a couple minutes after ending my 24hr prime95 run.
The second shot is about 10 minutes into a p95 run and there is a 10C difference. On the abit forums, a lot of people are seeing 10C difference while loaded, so that is fairly typical I guess.

 

[Duvie]

Well if you had ran the seti work bench unit by Mechbgon would be good....

Under encoding do an mp3 from lame encoder and and mpeg2 encoding from TMPGenc free download....enable the HT on the P4....

Shim, I noticed pwm problems with my oc before I changed out to a new case as well..I still am going to put some sinks on them and mabe direct and intake fan to blow right at it....I first noticed it with my 2.6c and if that pwm reaced 44c I would have errors in programs and prime95. I first though of the cpu temp but realised when I controlled that temp I could run 100% stable with a cpu temp 4-6c higher.



No offense on the testing....I just think to make a blanket statement when you said you really didn't do any testing on the p4p800 was a bit premature....Don't you?? It may still be true with real world data and I surely expect it to be exactly that way with gaming apps anyways....

 

[PetNorth]

Originally posted by: Duvie
Under encoding do an mp3 from lame encoder and and mpeg2 encoding from TMPGenc free download....enable the HT on the P4....

Of course, with these two, P4 will smoke A64. It's well known, TMPGenc, for example, is ultra optimized for P4 architecture.

He could use Canopus ProCoder for mpeg2 encoding with mastering quality setting (there is a free trial version http://www.canopus.us/US/support2/support_demosoft.asp ), or render a project video in Premiere or Vegas Video or Ulead VideoStudio...; or encode wav to ogg vorbis; or run the new benchmark includes with winrar version 3.3 or run benchmark that includes 7-zip, both programs to test data compression/decompression...

 

[caboob]

Nice OC Shimm. I can understand your dislike of the Springdale board. I bought an AI7 and noticed a full 1000 point dropoff in 3dmock compared to my IC7 so now its relegated to websurfing/DVD-encoding duty. As a gamer myself, I'm looking to eventually move to A64 but will do it when nv40/R420, PCI-ex, nForce-250 and a 64-bit killer app is out there. :beer:

 

[Soulkeeper]

Originally posted by: Shimmishim

Originally posted by: wicktron

Originally posted by: JustStarting

Originally posted by: wicktron
To the air-cooling crowd:

Shuttle AN50R rocks. Built a system for a homie and got 2.4GHz at 1.7v, not quite as impressive as Shimm's 2.5GHz at 1.7v, but hey, it's on air 🙂. Air cooled with Zalman CNPS7000A-AlCu, temps in the upper 40's at load.

How high does Vdimm go??

2.9v without mods, up to 5v with mods (VDIMM runs through +5v rail). I wouldn't go past 3v on A64 though.
The reason a bunch of A64's are dying is because of high VDIMMs. I haven't heard of any cases of high VCORE deaths yet.

the reason people's a64 are dying is because they are using the bios to change their voltage. so people are trying to set 2.9 vdimm as 3.1. what happens when people change the vdimm to 2.9 is is that they get 3.1 but they get crazy vdimm spikes. someone reported getting a spike of up to 3.8.

so the way to fix this is to set the vdimm to auto and use the vr's to adjust the voltage.

mine is set to exactly 3.0 right now but i've had it as high as 3.4 trying to get my bh-5's stable at 250 mhz @ 2-2-2-6

good luck with those bh-5 chips
i couldn't get mine stable over 230 with any voltage
also you might wanna try to find some good copper heatsinks to lay accross them and use artic silver
maybe then with 3.2 or more volts and an 80mm fan on them you could hit 250

good luck,
very nice oc
p.s. we all want it

 

[joe2004]

i couldn't get mine stable over 230 with any voltage

That is because you use 512 sticks with double banks.
I can get my 256 on 232-233 with 2.8V using 2-2-2-6. With 512 I get 222 at best, Mushkin even lower. Using the same voltage and 2-2-2-6.
That is the kick. AMD cannot handle 512 as high as 256 sticks. It never was able to.
People talking about overclocking memory sometimes don't know that important difference.

 

[Shimmishim]

A64 3200+ @ 2400 benches complete

here

thoughts?