Anyone remembered FX-4200?

[Sabotaged_Enigma]

Hey guys. I'm confused about how much L2 cache FX-4200 has got. But there seemed to be very few reports on this CPU. I remember it is a special one among FX family having 4 modules and 4 threads. So the question remains: how much L2 cache has it really got? 4 MB or 8 MB?
At first there was the leaked info, but afterwards it was thought to be a typo. And then it had little attention...
I can't find FX-4200 on AMD official website or TPU CPU database.

https://www.cpu-world.com/news_2012/2012032801_AMD_FX-4200_spotted_in_Gigabyte_support_list.html

 

[tabascosauz]

From what I tell the FX4200 was still Bulldozer, just an OEM part so it's obscure. So whatever the FX4100 had for cache config is what the FX4200 has. L2 did not change in the subsequent Vishera family either (FX4300/4320/4350). So 4MB.

It's not 4 modules either, that would be FX81xx/83xx.

It's 2M/4T, 2MB L2 per module. That remained the case until Bristol Ridge I think where L2 got halved to 1MB per module for whatever reason.

 

[Sabotaged_Enigma]

It's not 4 modules either, that would be FX81xx/83xx.
It's 2M/4T, 2MB L2 per module. That remained the case until Bristol Ridge I think where L2 got halved to 1MB per module for whatever reason.

Sorry but I couldn't agree with you. I saw rumour saying that FX-4200 had one integer unit disabled in each module, so it had 4 int units and 4 fp units in 4 modules. L2 cache size is uncertain, though.
Damn, need some more concrete info.

 

[Zyll Goliat]

When it comes to the L2 it should be 2 x 2 MB but L3 cache have 8 MB

 

[Sabotaged_Enigma]

When it comes to the L2 it should be 2 x 2 MB but L3 cache have 8 MB

Yes L3 cache is all the same 8 MB, but I don't know what's going on with the L2.

Here's what I've found on the web
This guy did side-by-side comparison of FX-4100 and FX-4200 at the same frequency. And it showed FX-4200 has double the performance of FX-4100.

One snapshot on CPU-world website, indicating 4 x 2 MB L2 cache instead of 2x2.

I wonder what's going on here.

 

[tabascosauz]

Sorry but I couldn't agree with you. I saw rumour saying that FX-4200 had one integer unit disabled in each module, so it had 4 int units and 4 fp units in 4 modules. L2 cache size is uncertain, though.
Damn, need some more concrete info.

So, it's basically a gimped and defective FX8150? It wouldn't be possible without using the big ~300mm^2 Zambezi die.

It hasn't happened anywhere else in the entire extended Bulldozer family, so I guess FX4200 is a unique OEM chip. Pretty rare by itself so I'm not sure if there will ever be any more testing.

It's interesting but it makes for essentially a FX8150 that can't perform like a FX8150 in integer, so that's about it when it comes to the value of it.

 

[Sabotaged_Enigma]

So, it's basically a gimped and defective FX8150? It wouldn't be possible without using the big ~300mm^2 Zambezi die.

Huh? Don't the whole lineup of Zambezi FX models use the same 315 mm² die?
Yeah this performance does indicate that there's something different with FX-4200, but there isn't too much info about it everywhere... We're guessing

 

[tabascosauz]

Huh? Don't the whole lineup of Zambezi FX models use the same 315 mm² die?
Yeah this performance does indicate that there's something different with FX-4200, but there isn't too much info about it everywhere... We're guessing

Oops, you're right about the single Orochi die.

MSI support lists had the FX4200 as being a regular 2M. So either the board vendors got conflicting info from AMD, or FX4200 was actually a normal 2M but AMD just got lazy with a select few samples and made a special CPU.

There's a thread here, I can't read German:

AMD FX 4200 gesichtet

AMD hat anscheinend vor einen AMD FX 4200 raus zubringen, welcher schon in vielen Kompatibilitätslisten aufgetaucht ist. Erstmal nicht besonderes, bis man sich den Cache und die TDP anguckt: 1MBx8 & 125W TDP. Das würde bedeuten, dass der FX 4200 ein 8-Kerner ist, was aber sehr unwahrscheinlich...

www.computerbase.de

 

[Sabotaged_Enigma]

MSI support lists had the FX4200 as being a regular 2M. So either the board vendors got conflicting info from AMD, or FX4200 was actually a normal 2M but AMD just got lazy with a select few samples and made a special CPU.

Idk man... Read this link:

https://www.cpu-world.com/news_2012/2012032801_AMD_FX-4200_spotted_in_Gigabyte_support_list.html

This is the first and the only source saying it's 8 MB of L2 cache.

There's a thread here, I can't read German:

AMD FX 4200 gesichtet

AMD hat anscheinend vor einen AMD FX 4200 raus zubringen, welcher schon in vielen Kompatibilitätslisten aufgetaucht ist. Erstmal nicht besonderes, bis man sich den Cache und die TDP anguckt: 1MBx8 & 125W TDP. Das würde bedeuten, dass der FX 4200 ein 8-Kerner ist, was aber sehr unwahrscheinlich...

www.computerbase.de

I'll try using auto-translate and see what I can find.

Okay after two minutes I found out those links had the same Gigabyte source...

 

[tabascosauz]

Idk man... Read this link:

This is the first and the only source saying it's 8 MB of L2 cache.


I'll try using auto-translate and see what I can find.

The only thing of any real value in that thread is the chart with gaming performance. But where the data comes from, how it was collected, what it even says on at the top of the supposed FX4200 column, no idea.

 

[AMF]

yup i had ole bulldozer! wiith an hd 4870 sapphire vapor x

 

[tabascosauz]

Idk man... Read this link:

This is the first and the only source saying it's 8 MB of L2 cache.


I'll try using auto-translate and see what I can find.

Okay after two minutes I found out those links had the same Gigabyte source...

Ah okay, the chart is just testing from another site from artificially handicapped 4M4T FX8150 to emulate a supposed FX4200, not testing the actual CPU.

I want to say that the Chinese testing looks legit and is actual FX4200 and not a emulated FX8150......but who knows. There's one really old CPU-Z screenshot of FX4200 showing 8MB L2 but it's super old (XP!) And no idea the source

 

[Vya Domus]

It hasn't happened anywhere else in the entire extended Bulldozer family, so I guess FX4200 is a unique OEM chip. Pretty rare by itself so I'm not sure if there will ever be any more testing.

It's not that odd, you had the option to disable one core per module in the BIOS for all of these CPUs.

 

[tabascosauz]

It's not that odd, you had the option to disable one core per module in the BIOS for all of these CPUs.

We're talking about the way these chips were released, not a BIOS setting.

 

[Sabotaged_Enigma]

A quick drawing to guess: is it possible to be like this? Can AMD disable only half the "shared" L2 cache?

 

[Vya Domus]

We're talking about the way these chips were released, not a BIOS setting.

Yes but the idea of having a chip with one core disabled per module isn't that outlandish since any FX CPU could be configured that way.

A quick drawing to guess: is it possible to be like this?

If this CPU was a thing that's exactly how it would have worked, again you could do this on any FX CPU, it wouldn't be a stretch to imagine one that would have those cores locked from factory.

 

[tabascosauz]

A quick drawing to guess: is it possible to be like this?View attachment 325001

View attachment 325002

In firmware it's clearly possible, but the FX4200 would have probably been physically fused that way, no? Which would be a special case amongst CMT CPUs.

 

[Sabotaged_Enigma]

again you could do this on any FX CPU, it wouldn't be a stretch to imagine one that would have those cores locked from factory.

In firmware it's clearly possible, but the FX4200 would have probably been physically fused that way, no? Which would be a special case amongst CMT CPUs.

Sure. Then I think as long as AMD can't disable only half of the shared L2 cache in each module, we can confirm it's 8 MB L2 cache then...

 

[tabascosauz]

Sure. Then I think as long as AMD can't disable only half of the shared L2 cache in each module, we can confirm it's 8 MB L2 cache then...

The plot thickens...

This one claims to be a FX8150 benching session with 1 core per module, but CPU-Z reports 2x2MB L2, it's a shame that the rest of the window is not visible

 

[Vya Domus]

Yes, this must have had the entire cache available, there is no way to configure the L2 cache that way.

 

[Sabotaged_Enigma]

The plot thickens...
This one claims to be a FX8150 benching session with 1 core per module, but CPU-Z reports 2x2MB L2, it's a shame that the rest of the window is not visible

Damn, what's going on here!

Yes, this must have had the entire cache available, there is no way to configure the L2 cache that way.

Look at that photo... Now I need to think...

 

[tabascosauz]

Damn, what's going on here!

Look at that photo... Now I need to think...

It could probably be explained if that user just misunderstood what he was doing and went the disable 2M route instead of disabling CMT. Hence 2x2MB. They were after a higher OC so conceivably both disabling modules and disabling CMT could give the same result of higher freq

I cannot for the life of me find any picture of CPU-Z for a CMT disabled FX8150. All of the OC validations just have whole disabled modules.

 

[LabRat 891]

I'm intrigued...

I have a friend w/ AM3+ kit, are these available on fleabay, affordably?

edit: nvm, 'bout ~$25-30+.

 

[MachineLearning]

I have one, it's indeed 4M/4C. 8MB L3

MachineLearning`s CPU Frequency score: 5218.33 MHz with a FX-4200

The FX-4200 @ 5218.3MHzscores getScoreFormatted in the CPU Frequency benchmark. MachineLearningranks #null worldwide and #5 in the hardware class. Find out more at HWBOT.

hwbot.org

 

[tabascosauz]

Damn, what's going on here!

Look at that photo... Now I need to think...

HWbot screenshots show 4x2MB for most of the records submitted, so safe to say you are probably right on this one. @Яid!culousOwO that one's on me, learn something new every day...

The only weird one is splave's run...2C/2T but only 1x2MB L2. There's a super old video of it on YouTube but it cuts right before it pans to the cores/thread box.

@Bones