Phenom II X1 570 4.2GHz single-core, good idea? So far, so good....

[kwalkingcraze]

Just checked using Asus Al Suite 3 software came with the motherboard, and its TDP is running at 150W total at 1.53V for 4.3GHz dual-core. So, that's 100W TDP more power-consumption than a Celeron G1620 to match its performance. You can see AMD is always behind the time.

 

[XavierMace]

I love that you're using a 6 year old processor to validate that statement.

 

[TurtleCrusher]

I consider the $15 Phenom II X2 570 BE the base, ultra-budget gaming CPU, due to its faster single-thread speed than FX-4100. Any K10 processor below 4.3GHz max speed, regardless of number of cores, isn't worth owning now for 2018. At least 1500 single-thread score is recommended now. Ryzen 3 1200 is rated at 1756, so it's only 255 points faster than my Phenom 570 BE at 4.3GHz.

A dual core is not recommended for gaming anymore. The G3258 is not recommended for gaming, definitely a K10 dual isn't either. There's threads on NBA 2K17 and others that games literally do not run, even on low settings. They pause for over a second at a time, 0FPS. Games need threads. A 3Ghz Q6600 is far better than a 4Ghz E8400, in modern games. Even old games such as Battlefield Bad Company 2 completely saturate a quad core if your GPU can keep up. I'd imagine that if your budget for a CPU is only $15 they'd probably be pairing it with a GTS250 or equivalent, so I guess they'd not be running anything modern anyway. I'd strongly suggest dropping $7 more on a Phenom 2 X4 for $22, a Phenom 2 X4 BE for $40, or a Phenom X6 at $45. There's a bunch of quad unlockable phenom 2 duals for $25, which would be my first choice if I had a board that unlocked. Used prices for the 8320 is $85, $99 for the 8350, so the choice is clear on that.

For kicks I just changed my 8350 to 3.2Ghz/1.125V and everything was nearly (if not identical) the same as 4.2Ghz that I ran. Rocket League, CS:GO, NBA 2K17, DiRT Rally, all good with an R9 270. I knocked it down to one module/two thread at 4.5Ghz and nothing was playable at typical settings. Even CS:GO was a stutterfest on medium. I saw what people were talking about in NBA 2K17, which runs great on quads+, but was absolutely unplayable at minimum res and lowest settings possible.

 

[kwalkingcraze]

I'd strongly suggest dropping $7 more on a Phenom 2 X4 for $22

A 3.00GHz Phenom 945 non-BE with only 1026 single-thread score, I don't think so. It'll be Celeron G470 all over again.

a Phenom 2 X4 BE for $40

Too much and overpriced. I can get FX-6300 for $10 more now. Reasonable price should be $25 now, still long way to go.

or a Phenom X6 at $45.

They don't run at 4.3GHz with 6-cores, zero percent chance.

None of these fit my needs for 2018. I don't play games.

 

[VirtualLarry]

If you need strong single-threaded, and were even considering a Phenom X6 for $45, then what about a Kaby Lake Celeron 3930 or whatnot? I've seen them new, for $37.

 

[kwalkingcraze]

If you need strong single-threaded, and were even considering a Phenom X6 for $45, then what about a Kaby Lake Celeron 3930 or whatnot? I've seen them new, for $37.

That's right, Celeron G3930's single-thread is rated at 1652. I would need to find a Phenom 570 at 4.8GHz in order to match it. The hexa-cores have 0% chance anyway.

 

[TurtleCrusher]

So the need isn't for gaming, it's for compute. I'd find different software before I limited myself to one core. That's a real niche case. For 99% of people they'd be much better off with a Phenom X6 than a high clocked single core K10.

 

[kwalkingcraze]

For 99% of people they'd be much better off with a Phenom X6 than a high clocked single core K10.

I'm running 4.3GHz dual-core right now. The OP ran it as single-core if he's still around.

 

[kwalkingcraze]

Is the K10 microarchitecture based on the K7 originally came out in 1999? I keep reading at Wikipedia with similar coincidences. "K10 cores are identical to K8". "K8 cores are very similar to K7, with 64-bit added." So, where was K10 originally born?

So, if I take one of the original 1999 Athlon "Classic" cartridge rated at 500MHz, and overclock it to 4.30GHz, does this mean this CPU architecture inside it will receive up to 1500 single-thread score like the Phenom 570 at 4.3GHz?

Interesting... I never knew K10 is really that old-school, while its maximum overclock frequency speed can compete with Ivy Bridge Celeron, all within one architectural design.

 

[Boris Morozov]

I'm running my quad core x4 965 @ 3.9ghz stable. 4GHz and up would have required more voltage and I don't think the cooling I have is sufficient.

I removed the gtx 660 and threw in a 6gb gtx 1060 for fun and the performance was actually very surprising. I have a 4k monitor and was shocked I hit 60 fps in Doom using vulkan API. Yes, there were frame drops and stuttering but the fact that it did it was a surprise. I should have recorded it. I'm using 8gb ddr 667 as well. GTA 5 in 4k at high I was getting 45fps down to the low 20s in certain areas. I love seeing how old hardware performs.

Even in 2017 I still find it a fairly capable CPU. Played all my games fine.

 

[Smoblikat]

I had a very similar setup back in the day (X2 555 that unlocked to a quad). I could run it as a quad core at 4.0, but I could get one of the cores to about 4.5, I experimented with many games (GTA IV era) and most games were playable at 768P on a high clocked single core. Those old PH2's are still great chips, ive got a couple X6's kicking around that are still in service.

 

[kwalkingcraze]

I had a very similar setup back in the day (X2 555 that unlocked to a quad). I could run it as a quad core at 4.0, but I could get one of the cores to about 4.5, I experimented with many games (GTA IV era) and most games were playable at 768P on a high clocked single core. Those old PH2's are still great chips, ive got a couple X6's kicking around that are still in service.

At dual-core mode only with unlocked cores disabled, what was your max frequency speed, and how you liked the performance?

 

[Smoblikat]

At dual-core mode only with unlocked cores disabled, what was your max frequency speed, and how you liked the performance?

I could run the same clocks/volts in dual core mode as I could in quad, so my 3 setups were 4.0 X2, 4.0 X4, and 4.5 X1. It didnt seem to care about the extra 2 cores, I probobly got lucky and got a re-binned 955BE that was otherwise perfectly functional. The performance was better than the E5200 I was using before (probobly thanks to L3 cache), but my buddys Q9550 was always faster, even when both chips were run in quad core mode. I think he was a few hundred mhz behind me too, so clock for clock I think Core 2 was faster than K10. I would consider a dual core K10 to be bordering unacceptable for modern tasks (especially gaming), but a Ph2 X4 @ 3.8+ would get most people by for a while still. The main issue with these chips is heat, they run cool but their thermal ceiling is low (62C MAX), but I have personally run 2.25V to the RAM through one of these, and the IMC and RAM itself both held up perfectly fine with no signs of degredation at all. I was using old crucial ballistix tracers with JMicron D9 chips in them. 1600mhz 6-6-6-16 @ 2.25V.

 

[kwalkingcraze]

I could run the same clocks/volts in dual core mode as I could in quad, so my 3 setups were 4.0 X2, 4.0 X4, and 4.5 X1.

When you disable more cores, you could increase your frequency speed gradually within same voltage. So at dual-core, you could get either 4.2GHz or 4.3GHz easily, since you got 4.5GHz at single-core.

 

[Smoblikat]

When you disable more cores, you could increase your frequency speed gradually within same voltage. So at dual-core, you could get either 4.2GHz or 4.3GHz easily, since you got 4.5GHz at single-core.

Uhhhh thats not really how that works. overclocking is hardly linear, especially once you approach the frequency ceiling. 4 of my cores could run 4.0, only one of them could do over that reliably. I owned the chip for years, believe me when I say I did thorough testing. and 1 core at 4.5 took more volts than 4 at 4.0, and 2 cores at 4.0 took the same amount of voltage as all 4 did.

 

[kwalkingcraze]

Uhhhh thats not really how that works. overclocking is hardly linear, especially once you approach the frequency ceiling. 4 of my cores could run 4.0, only one of them could do over that reliably. I owned the chip for years, believe me when I say I did thorough testing. and 1 core at 4.5 took more volts than 4 at 4.0, and 2 cores at 4.0 took the same amount of voltage as all 4 did.

That's right. Each additional core enabled requires more power consumption with slightly lower frequency speed set if I adjust it to 1.50V exact. With two cores disabled, I always get between 200MHz to 300MHz more than with 4 cores opened. Same thing as going from dual-core to single-core. It runs at 4.4GHz single-core, but only 4.3GHz at dual-core. I don't have a quad-core Phenom to test because this 570 can't be unlocked due to its newer manufactured date, but if it does unlock, it wouldn't run at 4.3GHz quad-core anyways. I would need to lower the speed to 4.1GHZ and go from there, sometimes down to 3.8GHz if I'm unlucky.

K10 CPUs are pretty sensitive. If it doesn't work over 4.0GHz reliably, you can just forget it. They no longer compete well with the new CPUs today in terms of single-thread performance.

 

[XavierMace]

K10 CPUs are pretty sensitive. If it doesn't work over 4.0GHz reliably, you can just forget it. They no longer compete well with the new CPUs today in terms of single-thread performance.

They never competed well in terms of single thread performance.

 

[Smoblikat]

That's right. Each additional core enabled requires more power consumption with slightly lower frequency speed set if I adjust it to 1.50V exact. With two cores disabled, I always get between 200MHz to 300MHz more than with 4 cores opened. Same thing as going from dual-core to single-core. It runs at 4.4GHz single-core, but only 4.3GHz at dual-core. I don't have a quad-core Phenom to test because this 570 can't be unlocked due to its newer manufactured date, but if it does unlock, it wouldn't run at 4.3GHz quad-core anyways. I would need to lower the speed to 4.1GHZ and go from there, sometimes down to 3.8GHz if I'm unlucky.

K10 CPUs are pretty sensitive. If it doesn't work over 4.0GHz reliably, you can just forget it. They no longer compete well with the new CPUs today in terms of single-thread performance.

Ive never heard of that before, I didnt think AMD ever laser cut the phenom series. Are you sure it wont unlock, have you tried? I bet it will, what board do you have? Unlocking is only supported by SB750+ IIRC.

 

[kwalkingcraze]

Ive never heard of that before, I didnt think AMD ever laser cut the phenom series. Are you sure it wont unlock, have you tried? I bet it will, what board do you have? Unlocking is only supported by SB750+ IIRC.

Yes I did everything I can. It can be unlocked to quad-core, but when I boot, the two additional cores were missing and not soldered in from factory, so it won't boot. The newer manufactured date ones made in 2013 have only less than 10% chance in unlocking now, vs. over 60% three years earlier. I have the worst luck with the ones made in 2013, but the older ones made between 2009 and 2011 were much better to unlock and higher chance for me. AMD caught this problem and decided to cut out the core unlock freebies in the newer mfg. date CPUs. I'm not surprised.

 

[kwalkingcraze]

I transferred my Phenom 570 to my new ASRock 970A-G/3.1 today because it has a USB 3.1 Type-C port at the back, and it looks like it can run at 4.4GHz dual-core as long as the memory RAM is set to 1333MHz max speed within the max 1.55V CPU voltage the motherboard will allow. If that's true, I just increased my single-thread score to 1528 from 1492 at 4.3GHz. ASRock boards don't read 1600MHz RAM for Phenom processors, but Asus and Gigabyte can.

 

[Smoblikat]

I transferred my Phenom 570 to my new ASRock 970A-G/3.1 today because it has a USB 3.1 Type-C port at the back, and it looks like it can run at 4.4GHz dual-core as long as the memory RAM is set to 1333MHz max speed within the max 1.55V CPU voltage the motherboard will allow. If that's true, I just increased my single-thread score to 1528 from 1492 at 4.3GHz. ASRock boards don't read 1600MHz RAM for Phenom processors, but Asus and Gigabyte can.

Try the UCC feature. I think its called UCC on that board anyway.........ACC (advanced clock calibration) is what enables unlocked cores, set that to on and make sure its set for "All Core". I dont recommend running that chip at 1.55 for sustained use, if youre air cooling and care about how long it lasts, I dont generally go over 1.45 for daily use. I run 1.48 - 1.5 on my 1090T with H70.

EDIT - Just for fun I ran a passmark single threaded test on my 4.4 I7 and got 2441. These old phenom chips are fun for backup rigs/just to play with, but if youre actually trying to make a fast computer, just get an I3 2120/Pentium/somthing that isnt AMD. Oh and BTW, ASrock boards do support 1600mhz RAM on Ph2 procs, min 970 Extreme3 runs over 2000mhz just fine.

 

[kwalkingcraze]

Try the UCC feature. I think its called UCC on that board anyway.........ACC (advanced clock calibration) is what enables unlocked cores, set that to on and make sure its set for "All Core". I dont recommend running that chip at 1.55 for sustained use, if youre air cooling and care about how long it lasts, I dont generally go over 1.45 for daily use. I run 1.48 - 1.5 on my 1090T with H70.

Looks like 4.4GHz dual-core didn't work out for me, so I'm back at 4.3GHz at only 1.51V. But it doesn't run at 1.51V constantly, V-Core downgrades automatically to 1.00V when it's idle.

UCC unlock didn't work for this processor because it was made on December 2013 "L3", AMD removed free unlock cores for the newest K10s. The older ones were better. I have several Athlon 220s that can be unlocked to quad-core with no problem, made on 2010 "H0".

 

[kwalkingcraze]

Does single-channel memory RAM help lower the CPU voltage requirement? I'm trying to see if I can get 4.4GHz dual-core by switching from dual-channel to single-channel. It misses a few inches or volts to run at 4.4GHz stable. It's 100% stable as long as the Internet is not connected or plugged in.

 

[Smoblikat]

Does single-channel memory RAM help lower the CPU voltage requirement? I'm trying to see if I can get 4.4GHz dual-core by switching from dual-channel to single-channel. It misses a few inches or volts to run at 4.4GHz stable. It's 100% stable as long as the Internet is not connected or plugged in.

Youre kind of pushing that board pretty hard as it is, those socket AM3+ ASrock boards were almost exclusively garbage except for the very high end 990FX variants. In my experience single channel/dual channel RAM hasnt yielded different results The IMC on your chip is basically bulletproof, just throw 1.65v at your RAM and run the clocks low while youre testing max CPU clocks, to rule out any possible memory errors. Most of the time what you consider to be "idle" is not what these chips would consider it to be. I would seriously suggest not running 1.51v through this chip on this board if you care at all about how long it lasts, and considering youre trying to squeeze that last 1% out of an antiquated chip, my guess is you do care how long this lasts. My personal recommendation would be see how far you can get with 1.45v. Ive played with just about every one of these chip/board/memory combinations you can imagine, and once you get over 3.8ghz the voltage scaling starts to get crazy. So an increase from 3.8 -> 3.9 might take an extra .125v, but going from 3.9 -> 4.0 would take an extra .2.

 

[kwalkingcraze]

Youre kind of pushing that board pretty hard as it is, those socket AM3+ ASrock boards were almost exclusively garbage except for the very high end 990FX variants. In my experience single channel/dual channel RAM hasnt yielded different results The IMC on your chip is basically bulletproof, just throw 1.65v at your RAM and run the clocks low while youre testing max CPU clocks, to rule out any possible memory errors. Most of the time what you consider to be "idle" is not what these chips would consider it to be. I would seriously suggest not running 1.51v through this chip on this board if you care at all about how long it lasts, and considering youre trying to squeeze that last 1% out of an antiquated chip, my guess is you do care how long this lasts. My personal recommendation would be see how far you can get with 1.45v. Ive played with just about every one of these chip/board/memory combinations you can imagine, and once you get over 3.8ghz the voltage scaling starts to get crazy. So an increase from 3.8 -> 3.9 might take an extra .125v, but going from 3.9 -> 4.0 would take an extra .2.

I tried looking at ASRock 990FX, but none of them offers USB 3.1 Type-A & Type-C ports at the back. By 2020, all new boards will have them standard, and this 970A-G/3.1 is the only AM3+ board that is 2020 ready, plus all the used FX CPUs will become cheaper starting at $10 on my next upgrade. What an investment... Sometimes the Asus 970 PRO GAMING/AURA I tried earlier feels more-obsolete than this ASRock 970A-G/3.1, after testing both, because it lacks USB 3.1 Type-C. That's why I switched to ASRock. I'm open to all brands, doesn't matter the reliability.

I'm fine running at 1.51V. It doesn't run at 1.51V 98% of the time, anyways. At normal usage to idle, it averages only 1.15V, according to ASRock Extreme Tuner meter. Frequency speed is also variable, doesn't lock at 4.3GHz.

K10 CPUs do prefer running at less than 3.8GHz, but I would be at LGA1156's Celeron G1101 single-thread level. Any K10s that can't run over 4.0GHz should be in the trash can already.