AM3 build, uses in 2025

[ShrimpBrime]

And why's that an issue? There's never been any consensus on defining what a "core" is, & that's because different implementations from multiple vendors can differ vastly.

By the same token Zen cores are 1.1-1.2x (normal) non HT ones, remember HT also takes up die space albeit negligible as compared to CMT.

Actually there was a big court battle over it, so yeah, definitely a consensus defining what a core is. Right there in a court room.

AMD settled to pay out.
Here, have a read

Redirect Notice

AMD Dragged to Court over Core Count on "Bulldozer"

This had to happen eventually. AMD has been dragged to court over misrepresentation of its CPU core count in its "Bulldozer" architecture. Tony Dickey, representing himself in the U.S. District Court for the Northern District of California, accused AMD of falsely advertising the core count in...

www.techpowerup.com

 

[eidairaman1]

I still have my FX rig. It has an FX8320 (stock these days), Asus 990FX Sabertooth, 16GB DDR3-1866, and a pair of GTX 970 GPUs not in SLI. I use it to run folding@home 24/7 down in the basement. Science! When I bought the hardware I knew it wasn't the fastest hardware available for a gaming rig, but I wanted it for the tweaking opportunity. It certainly wasn't a silicone lottery winner either, but it did run 4.75GHz 24/7 when it was water cooled. It had a fan mounted to the VRM and a fan on the back of the CPU socket too.

Winning combo, mine had 2133 ram but its oc at 2400 and 5.0GHz

 

[R0H1T]

Actually there was a big court battle over it, so yeah, definitely a consensus defining what a core is. Right there in a court room.

AMD settled to pay out.
Here, have a read

Redirect Notice

AMD Dragged to Court over Core Count on "Bulldozer"

This had to happen eventually. AMD has been dragged to court over misrepresentation of its CPU core count in its "Bulldozer" architecture. Tony Dickey, representing himself in the U.S. District Court for the Northern District of California, accused AMD of falsely advertising the core count in...

www.techpowerup.com

Read it again, they settled out of court. They never contested it, probably because of bad PR.

Ok looks like they did fight it but ultimately gave up ~

https://regmedia.co.uk/2019/08/27/amd-eight-core-settlement.pdf

Check the ruling and the processors listed, specifically only 7 models! Funny how come no one brought the APU's in there

 

[ShrimpBrime]

Read it again, they settled out of court. They never contested it, probably because of bad PR.

The point is that it had to go to court in the first place because people contested the fact. That's all I am saying.

To me, it's an 8 core cpu with only 4 FP units where 2 cores where called modules.

They already had bad PR otherwise...

 

[R0H1T]

Check the edit, they lost based on a bunch of buyers not getting the "performance" they wanted.

In particular, Plaintiffs claimed that they had viewed and relied on AMD’s allegedly false
advertisements when they were enticed to purchase Defendant’s CPUs. Id. ¶¶ 50-65. Plaintiff
Dickey allegedly saw representations of AMD’s CPUs that said they were “‘the industry’s first
and only native 8-core desktop processor for unmatched multitasking and pure core
performance.’” Id. ¶ 51. However, according to Dickey, this CPU did not perform as well as a
CPU would with eight independent cores. Id. ¶ 53. Plaintiff Parmer allegedly had a similar
experience. Before Parmer purchased one of AMD’s Bulldozer CPUs, he allegedly viewed
advertising that led him to believe that the Bulldozer CPUs would have eight cores independent
or capable of performing at full speed. Id. ¶ 59. Both Plaintiffs alleged that, had they known
these CPUs did not truly have eight-core capabilities, they would not have purchased the product
in the first place or paid as much for it as they did.

Following these discovery efforts, on March 27, 2018, Plaintiffs filed a Motion for Class
Certification, which sought to certify a class of consumers under California’s Unfair Competition
Law and False Advertising Law. Dkt. 118. Approximately nine months later, the Court granted
Plaintiff’s motion on January 17, 2019 and certified a class comprised of “all individuals who
purchased one or more of the following AMD computer chips either (1) while residing in
California or (2) after visiting the AMD.com website: FX-8120, FX-8150, FX-8320, FX-8350,
FX-8370, FX-9370, and FX-9590.

 

[R0H1T]

Admittedly I don't remember any ads for *dozer & it's unlikely AMD would've advertised much here. So if you're not a nerd you could feel cheated given the exaggerated claims most companies make, including AMD. So if they did that then yes users should feel aggrieved.

 

[Dr. Dro]

Check the edit, they lost based on a bunch of buyers not getting the "performance" they wanted.

I mean, for FX's performance to be acceptable, you either had to be a turbo AMD fanboy to the point you refused to acknowledge the existence of the competition, be on a lot of MJ, or both, usually both. Not that it makes this lawsuit valid, IMHO, but it is what it is. If you ask of me, it was always an 8 core processor and AMD did not lie there. If you released a processor that does half the work at triple the power consumption compared to the competition today, you'd be laughed out of the market and relentlessly mocked by the entire independent tech media. Core Ultra isn't half that bad and it's still by almost all metrics a complete flop that was unable to overtake even its own predecessor, which isn't exactly the absolute newest tech out there, basically Alder v2 v3 - it's still a 2021 microarchitecture.

In my opinion, it goes down in infamy as the worst processor line ever released because they repeated Intel's errors of almost a decade prior. It's not like they didn't know of NetBurst.

 

[R0H1T]

AMD's biggest issue was the IPC regression, if they could've clawed even some of that back with Piledriver or Excavator they'd have done a lot better. I bought a couple of AMD's laptops at the time for futureproofing, given their IGP's were supposed to be much better. Needless to say that didn't pan out as expected

 

[DaemonForce]

16GB DDR3-1866

I just remembered something kinda stupid. No idea if this happens on 990 boards.
Some 970 boards like the TA970 would auto-downclock memory to 1600 when populating 4 memory slots.
The FX memory controller theoretically tops out at 1866 but we run the snot out of it with so much fast memory.
I'm guessing the board topology matters a lot but it's still insane that we can even do this at all.

it did run 4.75GHz 24/7 when it was water cooled. It had a fan mounted to the VRM and a fan on the back of the CPU socket too.

My boards never saw any issues with VRMs so they went ignored. The real challenge was keeping everything within operating temps.
The RX580 raster would crash after 67℃ and would need a reboot to restore. On air the 970 chipset crashes ~92℃ which is a hard stop.

Water made things considerably better but the load temp needs to be LOW low, like 40℃. I managed but the behaviors vary per board.
My server tower has roughly 3/4" gap from door steel to back of board. If you can put a fan behind the socket, DO THAT. Airflow = good.

And why's that an issue? There's never been any consensus on defining what a "core" is, & that's because different implementations from multiple vendors can differ vastly.

Until FX. I don't understand how this braindead question manages to come up in an FX thread of all places. FX was why it was decided.
Very few of us agree with the ruling. Amazon notified me but didn't go for the payout. It's free money and prepaid but was still hella scummy.
Courts, like schools, are just not a place for smart people. Never depend on normies or a jury of normies to make good (technical) decisions.

A processor does not need to have a floating point unit to be considered a processor IMHO - otherwise chips like the 386 or 486 SX wouldn't be CPUs.

A very technical take, the one we typically adopt. Not every core needs a float (but sure helps to have dedicated units).
Normies were building junk systems and reeing about it because they could not tune worth a damn.
This FX was running circles around the 6600K, another fantastic chip. That's how dumb things were at the time.
They somehow dug the hole and put us in it. They did this to us. Remember to gatekeep. They will do it again.

Wouldn't daily an FX system if you paid me, but I'm on board with the benching fun.

Having an 8370 for daily/general/tests is fine. The 9590 for anything without the best cooling is just...Not good.
Peak 111W vs 270W is a world of difference. The kind that breaks people and their builds.

I mean, for FX's performance to be acceptable, you either had to be a turbo AMD fanboy to the point you refused to acknowledge the existence of the competition, be on a lot of MJ, or both, usually both.

Or....OR....Just be fast af boiii!

I have an Android that runs laps around the FX. It does, I know. I realize this was far behind even back then but still serves a special purpose.
The FX is among the few that outlived its purpose. Pentium 4 is dead in the 32-Bit dust and I let a squirrely 15W Athlon 64 take over for it.
Phenom II X4 was not able to keep up or compete in desktop gaming or VR. Great virtualization server and game host if we ever exit CGNAT.
Any FX 8c is slower than Ryzen but the reason I don't daily them is storage. I missed the boat for M.2 and had to wait until Ryzen to get NVMe.
20-30 years from now, FX is going to be the antique in board repair shops with bulging caps everywhere and the note "no POST, beeps."
It's going to be unrecognizable.