Was the AMD FX ahead of it time in Integer Performance?

[bonami2]

Had a random idea — ran PassMark on my 13900K as if it were an old FX-9590.

 

I disabled Hyper-Threading and the E-cores to mimic the FX’s setup, just out of curiosity.

Attached is the result. Now I’m wondering: are we seeing real gains in integer workloads over the past 12 years, or has most of the progress been in floating point performance?

Gonna dig into some real-world integer-heavy tasks next. Suggestions are welcome!

 

Top = 13900k

Bottom = 9590

 

 

Edited April 12 by bonami2

[pio]

This sounds like an interesting test!  Let us know what you find please.

[bonami2]

Left 13900k 8 P-Core 0 E-Core no HT

Right FX 8350 on a Formula Z at 4.7 supposedly.

Geekbench 6

 

Sadly seem like the FX was a lot slower! Still not bad for a 10yo+ Cpu

 

 

Edited April 12 by bonami2

[pio]

8 hours ago, bonami2 said:

Left 13900k 8 P-Core 0 E-Core no HT

Right FX 8350 on a Formula Z at 4.7 supposedly.

Geekbench 6

 

Sadly seem like the FX was a lot slower! Still not bad for a 10yo+ Cpu

 

 

Neat results.  Expected, but neat.   

It's amazing how quick and snappy those FX chips can still feel today honestly (and how SURPRISINGLY capable they are).  They really do surprise you, even if the benchmark numbers show "bad" numbers.

[ShrimpBrime]

I think the integer core was fine, a little small, pipelines a tad too long, but then they couldn't have done 5ghz boxed processors and 8ghz world records. So at least there where upsides to the story for sure.

[Alastair]

I loved my FX. Such fond memories of my good old Vishera. I still have her sitting in a drawer, missing a pin, but still working. 

[neurotix]

I had an FX-8350 gaming and benching stable at 5GHz. It was also lapped. It ended up going to damric, then pio, then back to damric (I think.)

 

Good memories too, the thing was fun to overclock especially with 2400MHz RAM and a HT and CPU/NB OC.

 

Unfortunately, I moved on because it severely bottlenecked two 290s in Crossfire:

 

Valley 1080p FX-8350 5GHZ with two 290s: 95 fps

Valley 1080p i7-4770k at stock: 110 fps

Valley 1080p i7-4770k at 4.5GHz: 135 fps

 

Yeah they were good chips and fun to OC, but I don't really think they were ahead of their time.

 

[pio]

On 02/05/2025 at 15:11, neurotix said:

I had an FX-8350 gaming and benching stable at 5GHz. It was also lapped. It ended up going to damric, then pio, then back to damric (I think.)

 

Good memories too, the thing was fun to overclock especially with 2400MHz RAM and a HT and CPU/NB OC.

 

Unfortunately, I moved on because it severely bottlenecked two 290s in Crossfire:

 

Valley 1080p FX-8350 5GHZ with two 290s: 95 fps

Valley 1080p i7-4770k at stock: 110 fps

Valley 1080p i7-4770k at 4.5GHz: 135 fps

 

Yeah they were good chips and fun to OC, but I don't really think they were ahead of their time.

 

Yes, Damric has your old CPU again.  And how did it bottleneck two R9 290's?  I had crossfire RX 580's with mine, did a great job.  That's how I blew mine up.  

[bonami2]

31 minutes ago, pio said:

Yes, Damric has your old CPU again.  And how did it bottleneck two R9 290's?  I had crossfire RX 580's with mine, did a great job.  That's how I blew mine up.  

Unigine heaven was bottlenecked on my 4.4ghz fx 6300 with a gts 250.

 

It still ran at 4k with a 7950.

My 4790k produced higher score with the gts 250 in unigine.

 

Depend on game or benchmark as usual.

[neurotix]

15 hours ago, pio said:

Yes, Damric has your old CPU again.  And how did it bottleneck two R9 290's?  I had crossfire RX 580's with mine, did a great job.  That's how I blew mine up.  

If you read my post all the way, you'll see it was bottlenecking 2x 290s in Unigine Valley (and most other benchmarks and games). A single one did like 60fps, two did 95 fps which is pretty bad scaling. This was remedied by upgrading to a 4790k. More info in the post I made earlier.

Edited May 4 by neurotix