AMD FX-8350 4.0 GHz - "Piledriver" for AMD Socket AM3+

[lordjohn]

I ordered 8120 with 990 combo in july, now 8350 with 970 combo cost same as what I paid in july. I should have waited

[Super XP]

Originally Posted by lordjohn

I ordered 8120 with 990 combo in july, now 8350 with 970 combo cost same as what I paid in july. I should have waited

Why? Your 990 is superior to the 970. If you really want the 8320 or 8350, just sell your 8120, add a little more cash and you got yourself a Piledriver. That is what I am planning when the time is right.

Originally Posted by os2wiz

I hope that was a joke. I have a Corsair H100 lying here ready for installation with the cpu, not a LN2 setup.. I am having installation Monday when the power supply unit is supposed to arrive. I think 4.8 GHZ is my likely target.

What with a H100? I already hit 4.40GHz running all 8-Cores with a minor bump in volts 1.375v. I can push her more, but what for. For me to go any higher running 8-Cores, I would have to up the volts a lot, something I am not willing to do. And I too have the H100.

I can easily see 5.40GHz with either of the 8320/50 CPU's with a mild bump in volts. If I can get 4.40GHz with the 8120 at 1.375v I would imagine I can get to 4.80GHzto 5.00GHz also at 1.375v. But you never know.

[os2wiz]

Originally Posted by Super XP

Why? Your 990 is superior to the 970. If you really want the 8320 or 8350, just sell your 8120, add a little more cash and you got yourself a Piledriver. That is what I am planning when the time is right.


What with a H100? I already hit 4.40GHz running all 8-Cores with a minor bump in volts 1.375v. I can push her more, but what for. For me to go any higher running 8-Cores, I would have to up the volts a lot, something I am not willing to do. And I too have the H100.

I can easily see 5.40GHz with either of the 8320/50 CPU's with a mild bump in volts. If I can get 4.40GHz with the 8120 at 1.375v I would imagine I can get to 4.80GHzto 5.00GHz also at 1.375v. But you never know.

Right now with the FX-8150 I have an H60 cooler which is much smaller than the H100. I found on Bulldozer 4.4 GHZ with my setup was the hiighest stable overclock that I could achieve. I guess if the H100 was in there right now, I could get 4.6 GHZ. I also beleive I can get 4.8 without difficulty . I'll give 5.0 a shot but I really don't want to tax the voltage much either. Record are fine , but I wouldn't want to shorten cpu life pushing higher voltages over a prolonged period. How high is your HT throughput. Right now mine is 2200.
I would have slow my memory more to get anywhere near 2600. That is the problem with the AMD design. Memeory bandwidth comes at the expense of system through put and vice versa. I think the next generation cpu will come with a new system replacing Hyper Transport. A totally new motherboard design for the full-fledged Steamroller. The refresh to piledriver may be on am3+ but in 2014 the full-fledged steamroller will usher in many changes. I must say vishera is selling very well right now. If it can maintain high sales AMD will stabillize and the accelerated improvements will be cpoming to all of us.

[Super XP]

My FX-8120 setup.

HTT @ 2600 MHz
NB @ 2400 MHz
CPU @ 1.375v
DDR3 @ 1866 MHz (16GB)
CPU @ 4400 MHz w/ 8-Cores

The IMC has been redone and enhanced with the FX-8320/50's versus Bulldozer.

[os2wiz]

Originally Posted by Super XP

My FX-8120 setup.

HTT @ 2600 MHz
NB @ 2200 MHz
CPU @ 1.375v
DDR3 @ 1866 MHz (16GB)
CPU @ 4400 MHz w/ 8-Cores

The IMC has been redone and enhanced with the FX-8320/50's versus Bulldozer.

Nice setup. I didn't do my homework properly about crossfire. I didn't realize that extra card is an extra
350 watts. I have no intention of using an extra 350 watts while I game A couple of hours a day. So now I will have to sell those 2 HD 6970 cards as well as my original HD 6950 card, So I just ordered a new HD 7950 card.With the new drivers I will be getting about 30 to 40% better performance without the expense of 2 cards
and the ongoing electricity costs. I got it from Newegg for $299 free shipping, and no taxes. Plus 4 free games.Iwill sell off the 3 cards on Tuesday after my new card arrives. I'll take a small hit over what I paid on Ebay just to move them quickly.

Originally Posted by lordjohn

I ordered 8120 with 990 combo in july, now 8350 with 970 combo cost same as what I paid in july. I should have waited

The 990 board shoulbe a better performer than a 970 board.Why don't you put the 8350 on the 990 board?

Originally Posted by Super XP

My FX-8120 setup.

HTT @ 2600 MHz
NB @ 2200 MHz
CPU @ 1.375v
DDR3 @ 1866 MHz (16GB)
CPU @ 4400 MHz w/ 8-Cores

I see you have HT Link Speed at 2600. Perhaps my error was thinking the North bridge frequency had to match the HT Link speed. Apparently that was a misconception on my part. I will adjust that tomorrow. Thanks for the low down.

[cdawall]

Hmmm...Maybe people jumped to soon to say the FX line sucks at games

[os2wiz]

Originally Posted by cdawall

http://img.techpowerup.org/121103/Capture001.jpg

Hmmm...Maybe people jumped to soon to say the FX line sucks at games

Yes BF3 apparently has some multithreading, unlike most of the crappy gaming code that most people bllndly put their dollars down for.

[DaedalusHelios]

cdawall,

You meant "this game" not "games" right?

"Games" implies more than one. The "s" in "games" makes it plural.

[os2wiz]

Originally Posted by DaedalusHelios

cdawall,

You meant "this game" not "games" right?

"Games" implies more than one. The "s" in "games" makes it plural.

The truth is that this cpu will do very wel on probably a dozen or so well coded games out of the hundreds of well known games out there. I have a turn based startegy game called Russia Under Siege , about the Russian Civil War of 1919-1922. It most likely will do well under Piledriver because it is essentially a grand database that would have to have been multithreaded. Action and shoot-em-up games can be multi-threaded, like BF III can be multithreaded, and therefore give over-all better performance for the game itself, or they can be single-threaded and still work-only well with the Intel cpu. Apparently Intel has won the "hearts and minds" of most game designers, but more accurately their wallets. I am sure they give seed money to many of these gaming firms which influences their design decisions. That is what monopolists like John D Rockefeller or Bill Gates would do. If I was totally obsessed about this issue I would spend the time to investigate. I am not an investigative reporter and have other interests in life so I don't. When steamroller releases the single threaded performance gap should close a great deal and AMD's superior mulitithread performance edge will increase substantially as well. Sixteen months of surviving without capitulation by AMD will lead to a scare in Intel's leadership. Intel pricing structure will be smashed and the future will be up for grabs.
By the way the fact that AMD can deliver over 90 fps on BF III shows that it can do so on any game that is designed to be multi-threaded. It is the single thread and single thread alone , a moribund and antiquated programming concept , that prevents high frame rates by seizing control of the cpu. Intel solves this by throwing lots more dollars on decoders and cache size, yes it works, but it is economically an unnecessary waste of resources, if only design and code discipline could seize control of the minds that run the software industry.

[Frick]

Originally Posted by cdawall

http://img.techpowerup.org/121103/Capture001.jpg

Hmmm...Maybe people jumped to soon to say the FX line sucks at games

Doesn't that mean it's GPU bound?

[Super XP]

Originally Posted by Frick

Doesn't that mean it's GPU bound?

Not quite. I believe the software is the real culprit.
There were a few games that performed better on Intel, but most I see no big difference.

[cdawall]

Originally Posted by DaedalusHelios

cdawall,

You meant "this game" not "games" right?

"Games" implies more than one. The "s" in "games" makes it plural.

As has been said it will depend on the game that was just one of the many multithreaded games out there. Crysis is another and Crysis 3 looks like it will be as well. Hopefully this is the way all games will be going. I would rather be prepped for the future since it appears developers are getting their collective heads out of their collective butts and coding things correctly finally.

Originally Posted by Frick

Doesn't that mean it's GPU bound?

More than likely yes, but on that note even being GPU bound AMD has had lackluster performance due to a lacking single core IPC making Intel chips look immensely better performing. This is yet more proof they are not...

[Garage1217]

Very nice... In Batman Arkham City, I run physx maxed and let the CPU handle the physx part. My 8150 and 1100T both would not go over 31fps, generally 30fps. Just benchmarked my overclocked 8350 and hit 36fps. Nice increase! Benchmark is rough, in game it is visually glass smooth.

[blibba]

Originally Posted by os2wiz

I hope that was a joke. I have a Corsair H100 lying here ready for installation with the cpu, not a LN2 setup.. I am having installation Monday when the power supply unit is supposed to arrive. I think 4.8 GHZ is my likely target.

For sub-zero overclocking, you want high-leakage. For conventional cooling, you want low-leakage.

[cadaveca]

Originally Posted by blibba

For sub-zero overclocking, you want high-leakage. For conventional cooling, you want low-leakage.

True, but can you tell me why?

[blibba]

Originally Posted by cadaveca

True, but can you tell me why?

No! I just wanted to point out that batches used in record-breaking OCs might not be the best to go for.

If I was to make an intelligent guess, I'd say that high-leakage parts, though generating more heat, get it out of the CPU more effectively. On LN2 you don't really care how much heat you generate, so this is preferable. On water or air, thermal limits are normally the absolute limits. How close did I get?

[cadaveca]

Originally Posted by blibba

No! I just wanted to point out that batches used in record-breaking OCs might not be the best to go for.

If I was to make an intelligent guess, I'd say that high-leakage parts, though generating more heat, get it out of the CPU more effectively. On LN2 you don't really care how much heat you generate, so this is preferable. On water or air, thermal limits are normally the absolute limits. How close did I get?

a chip has to leak...


it takes in all that wattage it consumes, and really, converts it to heat. Not really much else, really.

silicon is a semiconductor, and it's leakage changes according to temperatures. Cool a good "air" chip too much, and it won't be able to leak the power as heat sufficiently, and then that can causes errors or stalls, or maybe even damage? Not sure on that last bit. It simply prevents you from being able to remove the heat fast enough.


Anyway, so a "very leaky" chip, cooled to the same temps, has greater ability for leakage at the same temps, and ergo, can go a bit further before it hits that critical point where it's just too cold.


This is over and above the ability of the chip to be effectively cooled and handle the extra wattage(by removing the heat via leakage....)

handling that balance is where the real skill in sub-zero is. If you spend some time looking into LN2 cooling, you'll see how important that pots hold a specific temperature for a long time is important...this is a big reason for that need.

[cdawall]

Originally Posted by cadaveca

a chip has to leak...


it takes in all that wattage it consumes, and really, converts it to heat. Not really much else, really.

silicon is a semiconductor, and it's leakage changes according to temperatures. Cool a good "air" chip too much, and it won't be able to leak the power as heat sufficiently, and then that can causes errors or stalls, or maybe even damage? Not sure on that last bit.


Anyway, so a "very leaky" chip, cooled to the same temps, has greater ability for leakage at the same temps, and ergo, can go a bit further before it hits that critical point where it's just too cold.


This is over and above the ability of the chip to be effectively cooled and handle the extra wattage(by removing the heat via leakage....)

See the way I always looked at it was you have to have high leakage to loose some of the wattage. Low leak chips are the ones that normally fry when they are pushed to 1.8-2v under LN2. I always assumed it was the voltage drop across the chip that kept them alive at high voltage.

[cadaveca]

Originally Posted by cdawall

See the way I always looked at it was you have to have high leakage to loose some of the wattage. Low leak chips are the ones that normally fry when they are pushed to 1.8-2v under LN2. I always assumed it was the voltage drop across the chip that kept them alive at high voltage.

I'm not exactly sure what you are referring to, simply most likely because we are using different terminology.



voltage drop across a chip is kinda important too, if we are talking about the same thing, but as far as I know, it's all about leakage affecting the ability to remove heat when going really low. A chip that leaks more, you can pull more heat out of.


Do keep in mind, it's not like I went to school for this stuff, or I've had any of the "pro" clockers help...me raging over everyone's lack of help was what got me banned from XS for a year. Later I understood that the stuff I really wanted to know, was covered by NDA for those that really did know.

I mean, really, I'm just making shit up here. I dunno wtf is the truth, honestly. I still don't get any more answer than I did before doing reviews, really

[cdawall]

Originally Posted by cadaveca

I'm not exactly sure what you are referring to, simply most likely because we are using different terminology.

Could be I tend to look at things a little differently thanks to work and school.

Originally Posted by cadaveca

voltage drop across a chip is kinda important too, if we are talking about the same thing, but as far as I know, it's all about leakage affecting the ability to remove heat when going really low. A chip that leaks more, you can pull more heat out of.

AFAIK a chip that leaks more would have more voltage drop so we are talking about the same thing I think.

Originally Posted by cadaveca

Do keep in mind, it's not like I went to school for this stuff, or I've had any of the "pro" clockers help...me raging over everyone's lack of help was what got me banned from XS for a year. Later I understood that the stuff I really wanted to know, was covered by NDA for those that really did know.

I have had some help thanks to blatant disregard for NDA's which I helped with in all honesty. Lets just say "cdawall" is on file with AMD@Dallas.

Originally Posted by cadaveca

I mean, really, I'm just making shit up here. I dunno wtf is the truth, honestly.

Sounds like you are correct for the most part. End of it all is there was a reason the Phenom II TWKR chips were high leakage silicone.

[cadaveca]

Originally Posted by cdawall

Sounds like you are correct for the most part. End of it all is there was a reason the Phenom II TWKR chips were high leakage silicone.

I know that...long before TWKR chips were made, I was asking AMD reps directly for high-leakage chips for OC'ers..the chips they'd normally bin as useless, which was EXACTLY what AMD did with the TWKR chips. Which is why I say that the "TWKR" chip was my idea...

I "got" all this stuff long ago...as far as I understand, it's just something about silicon in general, and not really anything new. Where those critical points are has changed over time, but the general nature of what this behavior is, has not.



And really..just me making shit up...if I'm close, I guess that's because it's logical, and Occam's Razor wins over all. Pure luck.

I mean really, bin a bunch of chips under LN2, no matter the chip ,and this trend emerges, which is where I got the idea from. For all I know, could be some board thing, memory...some PLL..I don't have a clue, really.

[cdawall]

The old 945ES chips were the same way as the TWKR models. Every single one seemed binned more or less the same stable on water at 4-4.2, but required 1.65v on average.

[swaaye]

Originally Posted by cdawall

http://img.techpowerup.org/121103/Capture001.jpg

Hmmm...Maybe people jumped to soon to say the FX line sucks at games

CPU performance doesn't seem to affect the game much. It's probably very GPU bound at high detail + 1080p even with a 7950 3GB.
http://techreport.com/review/23750/a...sor-reviewed/7

[cdawall]

Originally Posted by swaaye

CPU performance doesn't seem to affect the game much. It's probably very GPU bound at high detail + 1080p even with a 7950 3GB.
http://techreport.com/review/23750/a...sor-reviewed/7
http://techreport.com/r.x/amd-fx-8350/bf3-fps.gif
http://techreport.com/r.x/amd-fx-8350/bf3-99th.gif

Originally Posted by cdawall

More than likely yes, but on that note even being GPU bound AMD has had lackluster performance due to a lacking single core IPC making Intel chips look immensely better performing. This is yet more proof they are not...

Already said that

[xenocide]

Originally Posted by cdawall

Already said that

It's worth noting that BF3 is also probably the most heavily threaded game on the market. If I recall it's optimized for up to 10 or 12 threads--leaning towards 12 since that would make it ideal for really any CPU on the market (6 core Intel's with HTing). I've said it before and I'll say it again; if you can max the threads on a BD\PD CPU, they are good options for the price. The problem is unless you're working in rendering or play specific games that are heavily threaded (a lot of Strategy, Simulation, and a handful of FPS') you're not using the best case scenario that makes your product worthwhile.

For most run of the mill games Intel CPU's handily outperform AMD options, just because they only use 1-4 threads, which with Intel's superior IPC, makes the AMD parts look bad.