My G3258 was a poor overclocker - not bad with 1.3 Vcore though

[slag]

Excuse me, where that chip came from? That would help the rest to avoid those chips.

http://forums.anandtech.com/showpost.php?p=37738101&postcount=22

Microcenter in Overland Park, Kansas. A suburb of KC.

 

[slag]

4.0 ghz

1.3 volts

updated below.. got to 4.2 @ 1.3 seemingly stable.

 

[VirtualLarry]

4.0 ghz

1.3 volts

Wow, and I thought mine was a bad OCer. I couldn't get 4.2 at 1.300V stable, so I backed off to 4.0 and 1.200V, and it seems stable to me.

 

[slag]

I am settling at 4.2 ghz @ 1.3 volts. 4.3 gives a whea uncorrectable error. I might try 1.35 volts when I get the hyper 212 in, but good with 4.2 for now.

My testing at 4.2 ghz was looping the heaven screen and then benchmarking it 4 times in a row.

 

[VirtualLarry]

My testing at 4.2 ghz was looping the heaven screen and then benchmarking it 4 times in a row.

Mine was running BOINC and WCG.

 

[slag]

well, it looped the cpuid stress test and the heaven benchmark all night long and had no issues at 4.2 ghz. I wanted to add that I only saw a marginal score increase in the heaven benchmark when going from 3.8 to 4.0 ghz, and very very slight increase going to 4.2 ghz. I'm using an ATI 7870 video card, but it seemed to not scale up as I increased the cpu frequency as much as I had hoped. I'm going to download MSI afterburner today after I get home from work and see what I can push the video card to. My son is happy to game at 1680x1050 with high detail and its pretty smooth.

 

[Justinbaileyman]

Why are you guys so scared to pump the volts through these cheap cpu's? They are made to be abused. I would go up to 1.45v and push this cpu as far as it can go. I bet these could easily hit 4.6-4.8Ghz. I mean come on they are only $65-$70.00 and can be easily replaced.

 

[slag]

Why are you guys so scared to pump the volts through these cheap cpu's? They are made to be abused. I would go up to 1.45v and push this cpu as far as it can go. I bet these could easily hit 4.6-4.8Ghz. I mean come on they are only $65-$70.00 and can be easily replaced.

Diminishing returns. 4.6 ghz won't do a darn thing for gaming except raise temps and prematurely wear out the cpu more than its already being stressed. Going from 4.0 to 4.2 ghz gained about 6 points in the heaven benchmark. It's for my son who only games and types word docs so 4.2 is fast enough for him. If I could get 4.6 ghz at 1.3V, I'd go for it, but not worth 1.45 volts in my opinion.

 

[soccerballtux]

That sounds like about how my overclocking went with a G3258 with my GA-H81M-DS2V v1.0 board when I got it. It had BIOS F5 out of the box, which supported G3258 overclocking.

I tried just increasing the multi, was able to boot Win7 64-bit at 38x, blue-screen at 39x, and no POST at 40x. But running DC load on it, I had to downclock to 36x, otherwise it would crash.

When Gigabyte released the F6 BIOS for my board, to fix Win10 issues, I found that now, I could magically overclock higher. I could also increase the vcore beyond 1.200V.

I was able to boot Win7 64-bit at 42x, but it wasn't 100% stable running DC load, even boosting the vcore from 1.200V to 1.300V. So I had to settle for 40x at 1.200V, which is where I'm at now.

So check for a BIOS update, it might improve overclocking.

Edit: Oh, there was something else I discovered, with the newer F6 BIOS. Setting XMP, would prevent me from being able to overclock as high, for some reason.

what do you have to say for yourself? this model is turning out to be a joke. not worth the buy unless you can hit 4.5. I paid $80 after tax for my FX-8310 and it hits 4.6ghz

 

[slag]

what do you have to say for yourself? this model is turning out to be a joke. not worth the buy unless you can hit 4.5. I paid $80 after tax for my FX-8310 and it hits 4.6ghz

huh? Why would it be a joke if it cant hit 4.5 ghz? Show me another $50 chip that has the performance this one does.

You paid $30.00 more for another chip (over half as much more than I paid for this processor) and both have a 7.7 rating on cpu boss....

cpu.userbenchmark.com says this:

For the vast majority of users, assuming an overclock of 4.3 GHz, the G3258 is the best value processor on the market by miles. On the other hand, users that encode multi-media or run SLI/Crossfire setups should look elsewhere. http://cpu.userbenchmark.com/Compare/Intel-Pentium-G3258-vs-AMD-FX-8310/2434vsm18534

My son games with a single card and types word docs. This chip is a much better bang for the buck than the AMD chip is for what he does.

 

[VirtualLarry]

what do you have to say for yourself? this model is turning out to be a joke. not worth the buy unless you can hit 4.5. I paid $80 after tax for my FX-8310 and it hits 4.6ghz

...and? So? The G3258 still has faster ST peformance, due to MUCH higher IPC. Combined with the value of having a heatsink that is ready-made for OC, and much lower power consumption, I think that the G3258 is a winner.

PS. Didn't you indicate that you could only hit your max OC on that chip, with only one module active, and that you had temp and mobo VRM stability-related issues?

Edit: Sour grapes from an AMD user that secretly wished they went Intel.

 

[LTC8K6]

My G3258 ran at 4.8 very nicely.

That's a 50% overclock.

Are we complaining about 50% overclocks on stock coolers now?

 

[Leyawiin]

what do you have to say for yourself? this model is turning out to be a joke. not worth the buy unless you can hit 4.5. I paid $80 after tax for my FX-8310 and it hits 4.6ghz

lol - and in many/most gaming situations a G3258 @ 4.6 Ghz can pimp slap it to the ground.

 

[Justinbaileyman]

My G3258 ran at 4.8 very nicely.

That's a 50% overclock.

Are we complaining about 50% overclocks on stock coolers now?

Sweet!! Thats what Im talking about!! 4.8Ghz ain't nothing to cry about at all specially for a el cheapo $70 chip. Kind of thinking about buying a couple just to putz around with. How are encoding times with handbrake? Right now Im using a AMD 5350 to encode a dvd to MVK 720p takes about 45 mins to an hour depending on the length of the film.

 

[LTC8K6]

Sweet!! Thats what Im talking about!! 4.8Ghz ain't nothing to cry about at all specially for a el cheapo $70 chip. Kind of thinking about buying a couple just to putz around with. How are encoding times with handbrake? Right now Im using a AMD 5350 to encode a dvd to MVK 720p takes about 45 mins to an hour depending on the length of the film.

http://www.anandtech.com/show/8232/...y-edition-review-the-intel-pentium-g3258-ae/2

 

[RickH]

It's a gamble.
I spent $150 for CPU, MB and memory to upgrade an old box. From a single core Pentium to G3258 @ 3.8 for less than I usually spend on a MB. Not bad, but it depends on what intend to use it for. Surfing the web or gaming--big difference in your requirements. R

 

[know of fence]

I am settling at 4.2 ghz @ 1.3 volts. 4.3 gives a whea uncorrectable error. I might try 1.35 volts when I get the hyper 212 in, but good with 4.2 for now.

My testing at 4.2 ghz was looping the heaven screen and then benchmarking it 4 times in a row.

I'm going to say it outright I am a little peeved at your and Larry's use of "1.3 volts". People will use just one decimal place to describe their OC, but that just to give an idea. Overclocking at "1.3" volts is like saying "Officer I wasn't speeding, I wasn't driving more than a 100 mph back there!" It's not nearly descriptive or informative enough.
The FIVR lets you adjust Voltage in 0.004 or 0.005 volt steps.
It only takes an average increase of about 0.019 V from one multiplier to the next, highest increase was 0.085 V for me, but never 0.1 V.

If you don't want spend the time fine-tuning, use 0.025 V steps. But 0.1 V jumps that just so crude and coarse.

What if your OC was stable at 1.205 V but you run it at "1.3 V". I'm not even talking about 5 to 10 watts and degrees saved, it's just misleading to round voltages up this way.

 

[VirtualLarry]

Huh? I've been careful to specify 1.300V, and not just 1.3V. And I tried intermediate voltages as well, between 1.200V and 1.300V. My CPU just isn't stable under DC load at 4.2Ghz, seemingly regardless of how much voltage I put through it.

Remember, my stress-test is a lot more grueling than the trivial testing that you've done, to test many clocks and voltage combos.

 

[slag]

I'm going to say it outright I am a little peeved at your and Larry's use of "1.3 volts". People will use just one decimal place to describe their OC, but that just to give an idea. Overclocking at "1.3" volts is like saying "Officer I wasn't speeding, I wasn't driving more than a 100 mph back there!" It's not nearly descriptive or informative enough.
The FIVR lets you adjust Voltage in 0.004 or 0.005 volt steps.
It only takes an average increase of about 0.019 V from one multiplier to the next, highest increase was 0.085 V for me, but never 0.1 V.

If you don't want spend the time fine-tuning, use 0.025 V steps. But 0.1 V jumps that just so crude and coarse.

What if your OC was stable at 1.205 V but you run it at "1.3 V". I'm not even talking about 5 to 10 watts and degrees saved, it's just misleading to round voltages up this way.

Be peeved all you want. I don't want to, nor do I need to, spend multiple evenings playing with all the voltage settings to figure out what works. 1.3 volts works for me, it doesn't overheat.


For you, as you put it, it only takes 0.019 V increase, but as you can see, I hit a wall earlier and couldn't even boot at 1.25 volts @ 4.0 ghz. Yet, I changed the voltage to 1.3 volts.. not 1.301 or 1.302, or whatever, but 1.3 volts, like I said, and now its smooth as butter and works great at 4.2 volts.

Lastly, I find it humorous that you are telling me what increments to overclock with and somehow I'm doing a disservice to people across the world because I didn't use small enough increments when I changed the voltages...

PFFT

 

[soccerballtux]

...and? So? The G3258 still has faster ST peformance, due to MUCH higher IPC. Combined with the value of having a heatsink that is ready-made for OC, and much lower power consumption, I think that the G3258 is a winner.

PS. Didn't you indicate that you could only hit your max OC on that chip, with only one module active, and that you had temp and mobo VRM stability-related issues?

Edit: Sour grapes from an AMD user that secretly wished they went Intel.

what's the ipc on yours?

 

[soccerballtux]

single core cinebench is cool by me for bench

 

[crashtech]

single core cinebench is cool by me for bench

Funny, you've never bothered to submit a score to my Cinebench thread, ST or otherwise. Should I start soliciting everyone's ST scores to encourage more participation?

 

[know of fence]

Lastly, I find it humorous that you are telling me what increments to overclock with and somehow I'm doing a disservice to people across the world because I didn't use small enough increments when I changed the voltages...

PFFT

I am a person living across the world, who found it very helpful to have a reference (AT review) for overclocking. So helpful that I set out to create an even more precise reference, to my surprise the results are entirely consistent. Having never owned an unlocked CPU with an OC board, it was a very rewarding experience. An experience that no fellow nerds cared to share, sadly.

In turn it raised more questions, especially in regard to variance in the silicon lottery, as well as cooling solutions.
To know where your CPU stands compared to others, which involves a certain gutsy resolve to face facts, all that is takes is to precisely determine the lowest voltage it will run at, even at stock frequency, or a frequency slightly higher than stock that isn't yet completely bogged down by thermals.

Mainstream culture panders to stupidity, so we all grow up with a bias against precision and all the right stuff, completely unjustified! Never mind the invention of the integrated circuit, the printing press or the wheel, it is quantitative measuring and subsequent standardization of units that created modern science. Precise quantifying is the difference between an exploding meth-cook and Walter White, the difference between astrology and astronomy, between Lavoisier and ... i don't know... Merlin.

Thanks to those 18th century Frenchmen, we already have the same volts across the globe, and it takes no time at all to adjust down your Volts and see if your PC boots into your OS. It is actually simpler to find the voltage curve and increase voltage from a position of certainty (using the reference to guide your voltage jumps). Rather than using the "see-saw" method, and torturing your CPU for days with prime crunching, ray tracing or protein folding, not knowing if you are at 5 mV or 95 mV above what is stable.

Your 4.2@"1.3" OC, it could be 30 mV to 130 mV worse that mine 4.2@1.170 [60°C]. With the added uncertainty of different cooling and presumably very different temps, the top OC makes for a bad point of comparison. Your chip may even be better or average, and it all comes down to cooling. I too ran into a (thermal) limit at which voltage gave no added stability long before reaching 100°C TJmax.

I have a selfish interest to find out of course, because neither Crashtech nor Ian Cutress or anyone else have provided data precise enough at the very low end (3.2 GHz), so I'm just kind of curious if the voltage curves all run into one another. I suspect they actually don't, and that a "bad" chip will have a higher voltage throughout the frequency range. But other than pure academic interests I also would like to see a really bad chip, properly tested (even at a single frequency, which is all it takes) so I can point and laugh, and feel good about my own luck.