The toy after a week of use

[BTRY B 529th FA BN]

Originally posted by: Rubycon

Originally posted by: geokilla

Originally posted by: jaredpace
lopri, what are your overcloaking results?

He's probably too busy jizzing his pants to post back. His OCs are just that high

As long as it stays out of the power supply he should be fine.

lol, thanks for the laugh

 

[heyheybooboo]

Originally posted by: lopri
There is definitely something wrong. Gotta go out for now but will probably back quick so I can figure this thing out. Until then all I have is a 1M.

What exactly are your problems?

4GHz+ @ 1.34v looks like a good start

 

[BTRY B 529th FA BN]

Originally posted by: heyheybooboo

Originally posted by: lopri
There is definitely something wrong. Gotta go out for now but will probably back quick so I can figure this thing out. Until then all I have is a 1M.

What exactly are your problems?

4GHz+ @ 1.34v looks like a good start

I think the 1m Pi times are a little slow, thus "something wrong"

 

[Flipped Gazelle]

Originally posted by: BTRY B 529th FA BN

Originally posted by: heyheybooboo

Originally posted by: lopri
There is definitely something wrong. Gotta go out for now but will probably back quick so I can figure this thing out. Until then all I have is a 1M.

What exactly are your problems?

4GHz+ @ 1.34v looks like a good start

I think the 1m Pi times are a little slow, thus "something wrong"

AMD pretty much tanks at Superpi, compared with Intel CPU's. I think those results are pretty normal.

 

[lopri]

Here is the bottom line: The BIOS wasn't ready for 955BE. For whatever reason, it would seem to read things correctly, but for only ONE core. The system was running 'Turbo Mode' and 'Cool & Quiet' at the same time.

It's no wonder why the system would fly one minute then crawl the next. I even ran Prime95 @4.0GHz and it ran for 2 hours+! Of course with only one core @4.0GHz and the rest stuck with x4 multiplier (0.8GHz).

I flashed the new BIOS from Gigabyte (published 4/24 it says, but I didn't see it till yesterday) and now the reality sets in.

I am right now just jumping around the clocks with multipliers and vCore only, but the report on AT review seems correct. There looks to be an issue with 4.0GHz/64-bit OS and Phenom. (at least with all 4 cores)

3.6 GHz 1.312V
3.8 GHz 1.392V

3.6 and 3.8 seem rock-stable although I didn't run any 3D bench yet. I have a feeling it will take a while to learn this chip. I didn't even started NB/memory/HTT overclocking yet. (and eventually 8GB RAM) I'm also thinking about switching the HSF. (currently using Scythe Samurai)

 

[jandlecack]

Originally posted by: Idontcare

Originally posted by: soonerproud
Another thing you got completely wrong is that backwards compatibility does not hold the AM3 processors back in performance since the AM3 and the AM2+ controllers on the CPU are entirely different controllers. In many benchies there was a slight increase in performance using AM3 due to the extra bandwidth of DDR3.

I think Jandle's argument is more that if AMD had ditched the AM2+/DDR2 hardware and transistor needed to for that compatibility and then used those freed-up transistors to create a tri-channel DDR3 AM3 cpu plus bolster any other xtor starved tweaks on current PhII (within the same die-size and cost-budget) then such a CPU would have likely had higher IPC.

He's making the argument that the presence of AM2+/DDR2 hardware means something else could not be present, and that something else could have meant higher performance.

That was what I had in mind, yes. Thanks for clearing it up.

 

[lopri]

...

 

[Flipped Gazelle]

Originally posted by: lopri
Here is the bottom line: The BIOS wasn't ready for 955BE. For whatever reason, it would seem to read things correctly, but for only ONE core. The system was running 'Turbo Mode' and 'Cool & Quiet' at the same time.

It's no wonder why the system would fly one minute then crawl the next. I even ran Prime95 @4.0GHz and it ran for 2 hours+! Of course with only one core @4.0GHz and the rest stuck with x4 multiplier (0.8GHz).

I flashed the new BIOS from Gigabyte (published 4/24 it says, but I didn't see it till yesterday) and now the reality sets in.

I am right now just jumping around the clocks with multipliers and vCore only, but the report on AT review seems correct. There looks to be an issue with 4.0GHz/64-bit OS and Phenom. (at least with all 4 cores)

3.6 GHz 1.312V
3.8 GHz 1.392V

3.6 and 3.8 seem rock-stable although I didn't run any 3D bench yet. I have a feeling it will take a while to learn this chip. I didn't even started NB/memory/HTT overclocking yet. (and eventually 8GB RAM) I'm also thinking about switching the HSF. (currently using Scythe Samurai)

Interesting. When the 955 was released, I saw reviews with OCing ranges from 3.6 - 4.3 Ghz. 3.8 seems pretty normal, which I think might disappoint AMD OCing fans a bit.

 

[lopri]

Quick 3D test was fine @3.8/1.392V. I wonder if I need a beefier PSU (like the one in Ruby's video) once I install a 4890? Right now I'm running the system on a 450W PSU.

 

[lopri]

Originally posted by: Flipped Gazelle
3.8 seems pretty normal, which I think might disappoint AMD OCing fans a bit.

Yeah I guess. But it's fast as hell even @3.6GHz. Granted the fastest 'feel' is felt @4.0GHz. It's crazy.

 

[Tempered81]

yah, if you're doing an overclocked quadcore and overclocked 4890, you will need more than a 450watt

 

[soonerproud]

Originally posted by: lopri
Quick 3D test was fine @3.8/1.392V. I wonder if I need a beefier PSU (like the one in Ruby's video) once I install a 4890? Right now I'm running the system on a 450W PSU.

I would most definitely get at least a 80% certified 550 watt psu for the X4 955. I bought a Antec EA650 Earthwatts 650 Watt psu myself at Best Buy to give myself a little extra wiggle room. The 4890 may require at least a 500 watt psu if not higher.

 

[soonerproud]

Originally posted by: jandlecack
That was what I had in mind, yes. Thanks for clearing it up.

I see where you are going now. ADHD sometimes affects my reading comprehension because I get in a hurry to read something and don't take enough time to digest it. I also get bored of things easily and move on to the next thing at the drop of a hat.

Anyway, I should have not posted my response in the first place since the OP was displeased with the direction the thread was turning. If you want to start a new thread to continue the discussion further, I am more than willing to participate, because I do learn a ton from the incredibly intelligent individuals that post in this section.

@ Lopri:

I apologize for continuing to stay off topic earlier as I read your wishes after my reply. Now back to your new toy!!!!!!!!!!!!!!!!!!!!!!!!! :beer:


Corrected typos

 

[lopri]

Just logged back on to report back. Oh thank you much and no worries there, soonerproud. Trust me I was all ready to participate in the debate. I just happened to get busy so I just missed the right timing. We all have opinions about AMD vs. Intel, ATI vs. NV, Apple vs. Microsoft,.. the list goes on and on.

Back to the 955BE. This chip is really an interesting chip. Frustrating, yet rewarding. I have had this chip now only for 2 days, but one thing I can say for certain is that my Gigabyte board is not ready for prime time overclocking 955BE. It is an excellent board and I made sure everything works right (including SB750's RAID/AHCI) using X2 4050e, but its BIOS is not tuned well for AM3 just yet. (I suspect most AM2/AM2+ boards would be similar, particularly with 955BE)

I see this through the parameters (NB/memory/HT) changing when I leave things @Auto, and apparently Gigabyte didn't spend much time on the base code as it could. Going further, I'm thinking that even the base code from AMD probably is far from being polished.

There must be some sort of 'hook' (like 'straps' in Intel chipsets) as the CPU/NB multiplier and HTT change, and it looks to be that's what makes it so difficult to break 4.0GHz even though 3.8~3.9GHz can be achieved by near-stock voltage.

I thought about trying it on 32-bit OS, but because I have no plan to use this CPU in a 32-bit environment, I decided not to waste time. It's kinda disappointing, however, that this chip is not clocking as well it does in 32-bit, considering AMD pioneered 64-bit in the x86 world.

In the meantime, I have an interesting question for all. (Take a guess before you click on the answer)

What is the most difficult stress test for 955BE, when overclocked?

Answer

 

[Flipped Gazelle]

Originally posted by: lopri
Just logged back on to report back. Oh thank you much and no worries there, soonerproud. Trust me I was all ready to participate in the debate. I just happened to get busy so I just missed the right timing. We all have opinions about AMD vs. Intel, ATI vs. NV, Apple vs. Microsoft,.. the list goes on and on.

Back to the 955BE. This chip is really an interesting chip. Frustrating, yet rewarding. I have had this chip now only for 2 days, but one thing I can say for certain is that my Gigabyte board is not ready for prime time overclocking 955BE. It is an excellent board and I made sure everything works right (including SB750's RAID/AHCI) using X2 4050e, but its BIOS is not tuned well for AM3 just yet. (I suspect most AM2/AM2+ boards would be similar, particularly with 955BE)

I see this through the parameters (NB/memory/HT) changing when I leave things @Auto, and apparently Gigabyte didn't spend much time on the base code as it could. Going further, I'm thinking that even the base code from AMD probably is far from being polished.

There must be some sort of 'hook' (like 'straps' in Intel chipsets) as the CPU/NB multiplier and HTT change, and it looks to be that's what makes it so difficult to break 4.0GHz even though 3.8~3.9GHz can be achieved by near-stock voltage.

I thought about trying it on 32-bit OS, but because I have no plan to use this CPU in a 32-bit environment, I decided not to waste time. It's kinda disappointing, however, that this chip is not clocking as well it does in 32-bit, considering AMD pioneered 64-bit in the x86 world.

In the meantime, I have an interesting question for all. (Take a guess before you click on the answer)

What is the most difficult stress test for 955BE, when overclocked?

Answer

I guessed ...! Do I win the kewpie doll?

Have you tried setting the NB/HTT to 1.6, and then start overclocking from there?

 

[nyker96]

Originally posted by: lopri


Back to the 955BE. This chip is really an interesting chip. Frustrating, yet rewarding. I have had this chip now only for 2 days, but one thing I can say for certain is that my Gigabyte board is not ready for prime time overclocking 955BE. It is an excellent board and I made sure everything works right (including SB750's RAID/AHCI) using X2 4050e, but its BIOS is not tuned well for AM3 just yet. (I suspect most AM2/AM2+ boards would be similar, particularly with 955BE)

that is too bad. but since you had become the offical guinea pig for 955BE around here this is to be expected. I think many am2+ board makers still testing new bios for this cpu. even am3 boards are still quite new. no doubt, things will be better after several bios tweaks few months down the line. still your initial results already pretty decent. have you tried any linkpack test on this cpu? I wonder how much temperature wise. linpack rack up the highest temp of any program I've seen. Thanks for the reports.

 

[lopri]

Linpack runs pretty quick and painless, as long as things are stable. (duh) It does 20-passes of Linpack in no time, yet it fails (insert the answer) within 20 secs if it doesn't like something (100% sure it's related to NB)

Thus I am not sure how I should treat Linpack under AMD platform. As a matter of fact, it used to fail with 4050e (instant lock-up's) which was perfectly stable for everything else.

Gazelle: You spoiler!

 

[Flipped Gazelle]

Originally posted by: lopri
Linpack runs pretty quick and painless, as long as things are stable. (duh) It does 20-passes of Linpack in no time, yet it fails (insert the answer) within 20 secs if it doesn't like something (100% sure it's related to NB)

Thus I am not sure how I should treat Linpack under AMD platform. As a matter of fact, it used to fail with 4050e (instant lock-up's) which was perfectly stable for everything else.

Gazelle: You spoiler!

What, me?

 

[lopri]

Thinking about it, I wonder the 4.0 GHz has something to do with the 2.0 GHz default NB frequency? As if it's hard to reach certain FSBs under certain straps, maybe something in the NB refuses to allow the execution unit to go past its x2 frequency? (This is a pure guess on my side)

Edit: Oh and temperatures - I don't know exactly. Software reads pretty low temps in a predictable manner (~50C as long as NB is not OC'ed, add 2~3C per every 200MHz increase of NB, maybe?) but I do not trust it. I am using a Scythe Samurai which I've been using for a couple of years now, and I have a good feel of how it reacts to what it's mounted on. I think actual temps, or differently put, temps calibrated to Intel's reading, should be at least 10C higher.

 

[Idontcare]

This 4GHz "upper limit" with PhII reminds me of the 400MHz "upper limit" in win95 with AMD K6-2 due to a timing loop erroring out for taking too little time to complete.

 

[lopri]

A quick update for anyone interested: I am going to start testing with 8GB (4x2GB) of RAM and a HD 4890 later tonight. The max OC I was able to achieve with 2GB (2x1GB) was 3.8 GHz core / 2.4 GHz NB. I called it good enough after 2 hours of Linpack, two full runs (1.5 hours each) of PCMark Vantage suit, as well as many Cinebench R10 runs. Core voltage = 1.376V, NB = 1.175V. More than anything, the NB clock is far short of what I hoped for. But I do believe it is not the chip, but rather BIOS (or possibly cooling) so I hope Gigabyte will release a better BIOS soon. Highest NB frequency I could boot the system with was 2.9 GHz, and it ran 2.6 GHz sometimes OK and sometimes not. Max HTT on this board was around 320. (but past 300 the board won't sleep correctly)

Obviously more RAM won't make things any easier, but I decided it'd be better working with the final setup I wanted than wasting time on some wishful numbers.

I know some folks were hoping to see higher clocks, but personally I am happy with the result/performance of the 955BE so far and benchmark numbers seem to back it up. (better than Q6600/Q9400 setups I had before, both were @3.6 GHz) And what I wanted, before 955BE reviews were out, was 3.6 GHz-ish clock speed and 2.8 GHz-ish NB speed, so overall it is not that far off. I am glad to finally have an AMD chip that's capable of running @ 3.6 GHz (it's my magic number), and it being a quad-core makes the deal sweeter although I sincerely hope AMD (or anyone for that matter) can solve this 4.0 GHz / 64-bit mystery. Oh and AMD should definitely rectify this huge discrepancy between core and NB clock speed. I can't help but wonder how an imaginary K10 running a same core/NB frequency like K8 would perform.

 

[Idontcare]

Originally posted by: lopri
I can't help but wonder how an imaginary K10 running a same core/NB frequency like K8 would perform.

Generate benchmark data with your bench of relevance/choice with the core/uncore clockspeeds identical at three (or more) different clockspeeds (say 1.8, 2.2, and 2.4GHz) and then extrapolate the results to 3.8GHz (use a best fit equation in Excel or equivalent) and compare the extrapolated result to that of your current setup (3.8/2.4) with the same bench.

The extrapolated result should be higher (better) than the 3.8/2.4 real-world result but we are interested in how much higher. I'd like to hope/believe it isn't more than a 5% performance (IPC) delta.

 

[xxTurbonium]

Originally posted by: lopri
long time, AMD
It's my first AMD CPU in a long time. (not counting the 4050e which came as a combo with the board) I'm excited.

Umm, I hate to be the one to break it to you, but I think there's something wrong with your CPU: it's upside down.

 

[Idontcare]

Originally posted by: Turbonium

Originally posted by: lopri
long time, AMD
It's my first AMD CPU in a long time. (not counting the 4050e which came as a combo with the board) I'm excited.

Umm, I hate to be the one to break it to you, but I think there's something wrong with your CPU: it's upside down.

Fixed now

 

[Flipped Gazelle]

Originally posted by: Turbonium

Originally posted by: lopri
long time, AMD
It's my first AMD CPU in a long time. (not counting the 4050e which came as a combo with the board) I'm excited.

Umm, I hate to be the one to break it to you, but I think there's something wrong with your CPU: it's upside down.

No, it's not upside-down; you are simply not looking at it from AMD's perspective!