Love your socket 754?

[Vette5885]

Wow, superPI seems to be sensitive to memory timings. Right now I can only get 34.67 seconds with my rig (in sig).

I had it close to 2850 with an x12 multiplier, and I only went down to 34.3 seconds.

Now I gotta try again this weekend.

 

[GreenMonkey]

Well, squeezed a little more out of my 3400 venice. 223x12 for 2682mhz Of course the math doesn't quite make sense there...

Guess that's probably the topend for me. I've never quite gotten it stable at 2.7ghz, not without bumping the voltage to 2.65 (something I'm not willing to do).

 

[BigMacAttack]

Greenmonkey - try the 11x multi. I used to run 230x12 but found the 11x multi the sweet spot for both cpu and ram. Try 240x11 and then 245x11. Who knows; you may even squeeze 250x11 out of it like me. I have had no luck on the 10x multi and the 12x means my ram runs a bit too slow - even on the DDR466 divider.
EDIT
Now isn't it just like this thing to go and make a liar out of me?! I played around quite a lot with it this morning and got it to 232x12. Here are the results:

Unclewebb - I don't think I can catch ya. That 32.30 is one smoking time!

 

[Vette5885]

You guys have me aiming at 32s right now.

I just had it up to 249 (crashed at 250) x11.5, for a 2864 cpu and 238 memory
Thats with the voltage at 1.695V (1.5*1.13)

I pulled 33.234s.

Being that the memory is rated 3-4-4-8 at 250, I might be able to pull a 3-4-3-6 or even a 3-3-3-6 at the 238. That would put me into the 32s.

Rig in sig - Cpu cooled by an AC-64 Pro with Ceramique

 

[unclewebb]

I posted a 32 second flat SuperPi once upon a time but I had to borrow my girlfriend's Crucial Ballistix 512 MB module to get me there. Now that the weather is cooling off it's time to head back to the dungeon for a suicide run to see if a sub 32 second SuperPi is possible!

High MHz, core and memory, as well as a cas latency of 2.5 has worked best for me.

What are the single core AM2s running? Maybe the low 30s with some DDR2-800 memory and their fingers crossed.

 

[Vette5885]

FINALLY

After 33.137, then 33.047, an even 33.000, I finally got into the 32's. Turns out the memory can do 3-4-3-4 at 250 (although this was at 240).

Chip was at 1.705 (1.55*1.1) - CPU-Z doesn't read over 1.69V on my MoBo

 

[BigMacAttack]

Nice going Vette! Welcome to the 32 sec. club
I was very surprised to see that my system didn't like anything over 1.64v in bios (1.664v as reported by CPU-z). If I go past 1.64v it tends to crash early and often. So after pushing to 254x11 with no luck in improving my Superpi score I tried 255x11 and it crashed no matter what voltage I tried. A quick switch to the 12x multi and the DDR466 divider gave me the 32.53 sec. run which is a personal best (I mistakenly thought that I had donbe 31.98 but it was actually only 32.98). Maybe one day I'll get lucky and get lower. This thing keeps surprising me from time to time.

 

[GreenMonkey]

Greenmonkey - try the 11x multi. I used to run 230x12 but found the 11x multi the sweet spot for both cpu and ram. Try 240x11 and then 245x11. Who knows; you may even squeeze 250x11 out of it like me. I have had no luck on the 10x multi and the 12x means my ram runs a bit too slow - even on the DDR466 divider.
EDIT
Now isn't it just like this thing to go and make a liar out of me?! I played around quite a lot with it this morning and got it to 232x12. Here are the results:

I like the 11.5x multiplier. The extra .5 x helps make for some finer control of the mhz

I'm pretty sure I can't get any higher without bumping the voltage again. I'm not too comfy with running past 1.55v. My next voltage step is 1.65 which makes me nervous. I'm afraid of reducing the life past that point - this processor will inevitably be a hand-me-down to my wife's box at some point - she's running my old AXP 1800+/9800pro.

 

[L1ght]

I broke 32 pretty easily. surpisingly.

2.9 Ghz Venice @ 1.75v
Ram @ 240 3-3-3-8(i think..) G.skill(capable of more)
12x multiplier

not the best pic, ill get a better one showing the 1m pi in detail and system settings.

 

[BigMacAttack]

I like the 11.5x multiplier. The extra .5 x helps make for some finer control of the mhz

I'm pretty sure I can't get any higher without bumping the voltage again. I'm not too comfy with running past 1.55v. My next voltage step is 1.65 which makes me nervous. I'm afraid of reducing the life past that point - this processor will inevitably be a hand-me-down to my wife's box at some point - she's running my old AXP 1800+/9800pro.

That's a fine oc for only 1.55v. Too bad your next voltage bump is so high. I don't blame you for not taking it on up. I don't mind running the extra voltage because I have 2 spare 3000+ Venices in my desk drawer and a 3400+ Clawhammer if I get desperate
I know what you mean on the hand me down. I just set up my old XP2000 (upgunned to XP2400) for my wife. It has an ATI 9200 SE card in it and is still a rock solid performer.

 

[GreenMonkey]

That's a fine oc for only 1.55v. Too bad your next voltage bump is so high. I don't blame you for not taking it on up. I don't mind running the extra voltage because I have 2 spare 3000+ Venices in my desk drawer and a 3400+ Clawhammer if I get desperate
I know what you mean on the hand me down. I just set up my old XP2000 (upgunned to XP2400) for my wife. It has an ATI 9200 SE card in it and is still a rock solid performer.

Well, I had my old 2800+ Newcastle, but I've been trying to liquidate all the spare computer junk - baby stuff is eating up a lot of room around the house.

My wife didn't want me to tear apart her box and put the the 2800+ & motherboard in it, so I traded 'em to a buddy at work for a nice hauppage card for my HTPC-work-in-progress.

If I had better cooling (as in, something other than my stock cooler) I'd give it a shot, but I don't want to pay $20 for another ~100-150mhz.

 

[unclewebb]

CPU-Z doesn't read over 1.69V on my MoBo]

The latest versions of CPU-Z seem to have a voltage reporting bug. Both versions 1.35 and 1.36 have this problem.

I guess no one thought anyone would be crazy enough to jolt their CPU with more than 1.70 volts.

Someone better tell the developer to head on over to the s754 club to see what's really been going on with these old chips.

Try downloading v1.34.1 and it should report your CPU voltage correctly.

http://www.filehippo.com/download_cpuz/?1126

Just click on the Download This Version button. They have a list of other versions of CPU-z as well.

My best SuperPi run was at about 2825 MHz. Thanks for raising the MHz bar.

 

[noobtech]

FINALLY

After 33.137, then 33.047, an even 33.000, I finally got into the 32's. Turns out the memory can do 3-4-3-4 at 250 (although this was at 240).

Chip was at 1.705 (1.55*1.1) - CPU-Z doesn't read over 1.69V on my MoBo

YOu had your HT Link at 4X? w000t! Is that okay that it's above 800!? What mobo are you using? I have some pretty good ram, so I wonder if I can do the same.

*edit... you have the same board that is in my wifes system... but the ram isn't as good I will give her a try later this week. haha

 

[Vette5885]

YOu had your HT Link at 4X? w000t! Is that okay that it's above 800!? What mobo are you using? I have some pretty good ram, so I wonder if I can do the same.

*edit... you have the same board that is in my wifes system... but the ram isn't as good I will give her a try later this week. haha

From what I've read, the primary 2 differences between the 754 and 939 single cores is the 1000MHz HT and Dual Channel. Considering that I have a pretty decend board, I figured I might as well run 1/2 way to 939.

Plus, when I originally ran 250x11.5, I had it set to 3x. Moving it to 4x gained me like .01s in Super Pi. Sounds insignificant, yes, but the run right before the one pictured above was 33.000s. I really wanted to be in the 32s.

FWIW, I have the latest bios (6.00) for my DFI Lanparty UT NF3 250GB, with the board at 1.7V (up from 1.6). I also put a much better passive heat sink over the SB (I think its a Thermaltake that I got from Newegg for like $10). My VF900 seems to blow some air over it too.

 

[tsuehpsyde]

YOu had your HT Link at 4X? w000t! Is that okay that it's above 800!? What mobo are you using? I have some pretty good ram, so I wonder if I can do the same.

*edit... you have the same board that is in my wifes system... but the ram isn't as good I will give her a try later this week. haha

Even on 754, 1GHz is usually fine (depending on the board). I wouldn't go over 1GHz, but 1GHz and below is generally fine. Every board is different, so just test to find the peak of yours.

 

[noobtech]

From what I've read, the primary 2 differences between the 754 and 939 single cores is the 1000MHz HT and Dual Channel. Considering that I have a pretty decend board, I figured I might as well run 1/2 way to 939.

Plus, when I originally ran 250x11.5, I had it set to 3x. Moving it to 4x gained me like .01s in Super Pi. Sounds insignificant, yes, but the run right before the one pictured above was 33.000s. I really wanted to be in the 32s.

FWIW, I have the latest bios (6.00) for my DFI Lanparty UT NF3 250GB, with the board at 1.7V (up from 1.6). I also put a much better passive heat sink over the SB (I think its a Thermaltake that I got from Newegg for like $10). My VF900 seems to blow some air over it too.

Did you bump up the chipset voltage or ram? I bumped my ram to 2.7 and cpu to 1.7.

 

[cirial]

Can someone explain how they test for oc stability using Super Pi? Search didnt turn up much.. and I have only pushed my 3200+ up 200mhz so far

 

[Vette5885]

Can someone explain how they test for oc stability using Super Pi? Search didnt turn up much.. and I have only pushed my 3200+ up 200mhz so far

Super PI isn't so much for stability as it is for benchmarking. In my tests, it seems to mimic speed changes in the processor pretty well - a 15% speed increase across the board on my computer netted a 15% time decrease with PI.

Plus, it works pretty well for your e-penis (woot! I'm in the 32 second club baby!)

For stability, I use Memtest when I push the memory, then when its safe, I go into windows and run Prime95 (I alternate between small FFTs and in-place large FFTs)

And Noobtech: Memory is 2.7v (seems to crash at 2.8, must be a heat issue), CPU is 1.705 for 2880Mhz and 1.58 for 2750, and chipset is 1.7V.

 

[unclewebb]

Can you believe it?

Users can choose from cheap dual core processors from AMD in socket 939 or AM2 form.

They can also pick up an older Intel dual core that they are giving away these days or they could dig deep and pick up a high performance Conroe chip.

With all these choices what's the number one seller at Newegg?
You got it, the socket 754 Athlon 3400 Venice core chip. ( Model #: ADA3400AIK4BO )

http://www.newegg.com/Product/ProductList.asp?Submit=TOPSELLER&SubCategory=343

 

[freakycody]

lol, that is funny. You can't beat the price, mobo plus good cpu all for 100 bucks Too bad I am torn between this and a dual core......

 

[BigMacAttack]

They should have released the 3400+ Venice into the retail channel a year ago when it was first produced instead of only releasing it to OEM builders. I guess if they had done that they never would have sold any 3000+ or 3200+ Venices though. Many thanks to Intel for initiating the price war and inspiring AMD to finally loosen their grip on this great chip!

 

[Xilikon]

They should have released the 3400+ Venice into the retail channel a year ago when it was first produced instead of only releasing it to OEM builders. I guess if they had done that they never would have sold any 3000+ or 3200+ Venices though. Many thanks to Intel for initiating the price war and inspiring AMD to finally loosen their grip on this great chip!

I agree, if this was available earlier and at a good price, i would have gotten one but i'm still happy on mine Just awaiting to get the Epox 8npa and retweaking to see if i can get better performances than my old MSI Neo Platinum.

 

[BigMacAttack]

You're getting about what I got out of my 3000+ Venice. I was able to go to 2.65 with it which is a darn good oc for a 2.0 ghz chip. That Epox should be a good board. Are you going to go the SLI route also? (yeah I know - silly question )

 

[Xilikon]

You're getting about what I got out of my 3000+ Venice. I was able to go to 2.65 with it which is a darn good oc for a 2.0 ghz chip. That Epox should be a good board. Are you going to go the SLI route also? (yeah I know - silly question )

It's not a silly question, i will go the SLI route as soon as the 7900GS cards dip around 100$. I ordered this mobo along with a new XFX 7900GS 480M Extreme... I plan to milk the s754 setup the longest possible to tide me well over the K8L/AM3 setup and skip DDR2 memory.

Since currently, the cores is going toward multi-cores instead of more speed, i won't feel left out in the speed bandwagon.

 

[noobtech]

Super PI isn't so much for stability as it is for benchmarking. In my tests, it seems to mimic speed changes in the processor pretty well - a 15% speed increase across the board on my computer netted a 15% time decrease with PI.

Plus, it works pretty well for your e-penis (woot! I'm in the 32 second club baby!)

For stability, I use Memtest when I push the memory, then when its safe, I go into windows and run Prime95 (I alternate between small FFTs and in-place large FFTs)

And Noobtech: Memory is 2.7v (seems to crash at 2.8, must be a heat issue), CPU is 1.705 for 2880Mhz and 1.58 for 2750, and chipset is 1.7V.

I'll try those settings tonight... I hope I can get in the 32 club

 

[unclewebb]

It took every last bit of performance that my 1GB OCZ Platinum could deliver but I made it.
The PC3200 - OCZ4001024ELPE was at 235.5 MHz and CL2.5-3-2-5 with a command rate of 1T.

Memory voltage was set to a default of 2.6 volts which is reported as 2.65 volts in Everest.
Increasing the voltage beyond that only results in errors.

I guess I've hit the wall but that's not too bad for an old s754 Athlon 3000.

 

[uzum4ki]

That is utterly amazing.

 

[Mr. Stryker]

Selling two AGP boards if anyone was interested.
http://www.hardforum.com/showthread.php?t=1086132

 

[Xilikon]

Got my parts (check my updated sig) and installed it in my main PC. WOW, Epox has better overclocking options than my former MSI Neo Platinum with the support of half-multipliers and more DDR speeds (added DDR150 and DDR183 between 133, 166 and 200 so I'm now able to run my RAM at 200 MHZ with a DDR150 and 274 MHz FSB).

The only bummer is that the voltage options is limited to 0.15 over default, which put me to 1.55 max but Everest tell me it's 1.59 so it's not all bad. Anyway, mine was running at 1.55 already. On the bright side, the voltage options is very clear (+0.01 to +0.15 or -0.01 to -0.25 from default) compared to the cryptic % increase or some formulas. It has also chipset voltage options when the MSI one doesn't. The temperatures on the Epox is slighty lower as well...

I didn't push it farther yet but i guess it's a much better rounded motherboard and i think I may be able to extract some more juice from my Venice

Some preliminary benchmarks with my 7900GS at 600/800 gave me 4597 3D Mark 06, 9011 3DMark 05 and 37,5 seconds SuperPI (slow but i guess it's sensitive to memory speed and timings so mine is prolly correct).

 

[Unknown-One]

Check what vCore is actualy at under "PC Health Status", or better yet use a multimeter.

These boards tend to put out more than they say they will. For example I have mine cranked all the way to "+0.15" in the BIOS (1.4 + 0.15 = 1.55v), but its actualy running the chip at 1.62v.

 

[unclewebb]

Check what vCore is actualy at under "PC Health Status", or better yet use a multimeter.

I know how to check the 12V rail, the 5V and how to check for 3.3V with a multimeter but how the heck do you check vCore with a multimeter?

 

[Unknown-One]

I know how to check the 12V rail, the 5V and how to check for 3.3V with a multimeter but how the heck do you check vCore with a multimeter?

The only way I know of is to run a wire (shielded to avoid shorting across other pins) from the voltage pin on the processor so you can get a direct reading.

Strip an end of a VERY THIN wire and shove it into the correct pin hole on the socket, then smash the processor down on top of that. Once you put everything back together you can measure between the lose end of the wire and ground to get a direct voltage reading.

The last time I tried this was on a P4 though...not exactly sure if there is an easier way to do it for s754.

 

[unclewebb]

The only way I know of is to run a wire (shielded to avoid shorting across other pins) from the voltage pin on the processor so you can get a direct reading.

That's kind of what I thought. I'm not that brave so I think I will stick to using the older versions of CPU-z.

The best advice I have for new users is to set the CPU voltage as high as the motherboard will go. That works best for me!

 

[Unknown-One]

The best advice I have for new users is to set the CPU voltage as high as the motherboard will go. That works best for me!

Yea, its not like the top voltage on these EPoX boards is very dangerous. These Venice chips appear to take 1.6v just fine (heck, BigMac runs his at 1.7v).

 

[BigMacAttack]

Actually I'm only running at 1.62v atm but I have pushed beyond 1.7v a time or two with my 3000+ and 3200+ Venices. The 3400+ does not seem to like anything over 1.66v (bios) anymore. I was playing with it last weekend and noticed it would crash if I went to 1.68v or 1.7v but would go to 232x12 or 255x11 on 1.64v. Of course it was nowhere near Prime stable but I could get Superpi runs off it and could run normal apps.
I don't dare go to the max vcore my board allows, though. It goes from 1.7v directly to 1.8v and maybe to 1.85! That's a bit too much for any Venice IMO! Unless one has a water cooling setup. I wouldn't be opposed to that possibly sometime in the future

 

[Xilikon]

Did a check via the pc health monitor in bios and it's 1.6v

The bummer right now is that at 2603 MHz (9.5x274), it's not even prime-stable (fail within 1 hour or shout it has a bad hardware problem). Memory is running at 200 MHz 2.5-3-3-6, perfectly within the specs. I started Memtest86 1.65+ this morning before going to work so it will test during the whole day to find out if it is the memory.

I need to find what cause this since i'm very sure my Venice can run 2.6 prime stable and my possible guess is that it doesn't really like the half-multiplier or the htt bus is a bit high ( currently 3x for 822 MHz, which should normally be no biggie ). If this doesn't work, i'll try either 10x or 9x

 

[BigMacAttack]

Yeah - half-multi's haven't been too successful for me either. My 3400+ seems to like the 11x multi best but my 3000+ liked the default 10x best, perhaps because my mobo didn't like the higher cpu frequency settings. Half- multi's also affect ram speeds too so that might be a fly in the ointment for you as well.

 

[Unknown-One]

Yeah - half-multi's haven't been too successful for me either. My 3400+ seems to like the 11x multi best...

I ran into the same thing, I can clock higher at 11x than I can at 10 or 12.

8x was another good multi, I actually hit 2.7GHz (8 x 337 = 2696) prime stable (only tested 2 hours) but the nForce4 chipset was getting MUCH hotter than I was comfortable with.

 

[cirial]

How are you guys unlocking multipliers? I haven't read anything about unlocking them, since back in the Socket A days of doing the "pencil trick".. anyone have any documentation that I can read up on?

 

[HydrasunGQ]

How are you guys unlocking multipliers? I haven't read anything about unlocking them, since back in the Socket A days of doing the "pencil trick".. anyone have any documentation that I can read up on?

They're not unlocking. AMD CPU's are unlocked by default at multipliers below stock. You can not change to multipliers above stock with the exception of the FX line.