FX 8140 = 8 cores at 3.2GHz and 95W

There was supposed to be a 95w version of the 8120. But i have never seen one.
 
might be like the X6 1045T in that it is available only to OEM's.
 
that AMD FX-Series FX-B4150 is a four core part at much higher speed...................?
 
pretty much any 95w you can guarantee is designed for OEM since OEM's absolutely hate exceeding 95w TDP, because if they do it means dropping more money on motherboards which cuts into their profits.
 
true, but they also release retail parts with 95W tdp.

the 8140 will be an interesting chip if we are able to purchase a high-speed eight-core AMD FX part with a 95W tdp.

this is why i ask, because this is why it is interesting.
 
I'm surprised no one has commented on the increase in L3 cache. They DOUBLED the size of L3 cache, which I bet was a huge bottleneck for the first wave of BD CPU's.

Good job AMD! (Hopefully those numbers are true ;)) Also 95 watts is another improvement. Lets see what happens next, but they did the obvious so far.

FX 4100 = 6MB
FX 4150 = 12MB

FX 6100 = 8MB
FX 6150 = 16MB

FX 8120, FX 8150 = 8MB
FX 8140 = 16MB

They should easily get a 10% improvement in IPC with this new addition (If the low cache was a bottlenck for the first BD module design that was released...I hope it was), and should finally be a upgrade to Thuban owners I believe :) Maybe a couple extra 100 MHz OC compared to the first BD batches if temps are lower also. Sounds good to me.
 
the cache size was never the issue. in fact it won't change anything since very little actually uses the L3 cache except games, which it did perfectly fine. the issue was always the cache latency on the L2 and size of the L1 cache which has a much bigger effect in applications.
 
teletran8 said:
I'm surprised no one has commented on the increase in L3 cache. They DOUBLED the size of L3 cache, which I bet was a huge bottleneck for the first wave of BD CPU's.
Click to expand...

It looks like the cache is unchanged. The numbers on the chart are just the total cache (L2 + L3).

4 module: 8 + 8 = 16 MB
3 module: 6 + 8 = 14 MB
2 module: 4 + 8 = 12 MB
 
Agreed latency and bandwidth of the cache were the big problems with the cache (maybe even a big portion of the IPC problem). Doubling the size without improving both latency and bandwidth will not be very successful.
 
is there any news on the B3, or any speculation on whether these new CPU's would be examples of the B3 stepping?
 
John Vuong said:
I wonder if these new 95W TDP CPU's will reduce power draw significantly.
Click to expand...

TDP is simply a specification of the maximum amount of thermal energy a chip needs to sink. Heat being related to how much electricity you pump into a given chip...the answer is it will be less, the question is how much.
 
R3MF said:
is there any news on the B3, or any speculation on whether these new CPU's would be examples of the B3 stepping?
Click to expand...

no, there was never a B3 stepping planned. basically pile driver would be the equivalent of a B3 stepping but its going to have a lot more changes to it then what a stepping change would be. the only time you would really see a new stepping is if they changed how the processor was made, eg C2 vs C3 + efficiency. what you are seeing with the new 95w chips is just a maturing manufacturing and better yields which allows for better TDP ratings.
 
B3 never planned? Don't speak with authority on something that you clearly are clueless about.
 
sirmonkey1985 said:
no, there was never a B3 stepping planned. basically pile driver would be the equivalent of a B3 stepping but its going to have a lot more changes to it then what a stepping change would be. the only time you would really see a new stepping is if they changed how the processor was made, eg C2 vs C3 + efficiency. what you are seeing with the new 95w chips is just a maturing manufacturing and better yields which allows for better TDP ratings.
Click to expand...
Solomutt said:
B3 never planned? Don't speak with authority on something that you clearly are clueless about.
Click to expand...

More than likely a B3 revision was planned (8170, 8130, 6130, 4130), but never progressed to the design stage. The engineers more than likely decided to shitcan B3 and migrate the fixes/revisions into Piledriver (probably C2).
 
I am still not seeing how a move to introduce new, lower power, processors obviates a B3 stepping on the process. Improvements to existing lines can be made in the form of new processor numbers along with logic revisions. The Athlon series and Phenom X2 series being introduced while the PII line was still being improved is just an example. The 960T came out using less power than the PII 965, and less voltage than the 6 core processors (per volt ranges in cpuworld).....

I could be off on this, and I am always willing to learn solid facts. Making suppositions about AMD's stepping plans seems like a play for popularity more than actual information exchange, when you are asserting that it's likely there is never going to be a new stepping.

One way I can see the "no B3" play out is marketing. Since BD introduction CPUs had issues with IPC, heat, and power draw, marketing may want to move away from the series as introduced, and introduce whole lines that are actually the stepping, thus distancing themselves from a bad start. Core duo is, I believe, an example. Two P4s slapped together, and providing more performance, but still a broken architecture.

I guess you could be on to something, and right-ish.
 
Core duo was a completely different core from the P4s and had a very different design philosophy.
 
Yep- was thinking of the Pentium-d. The point is that steppings and new product lines aren't mutually exclusive. Declaring AMD has no stepping B3 forthcoming seems a bit premature.
 
I think you're both right. What BD needed was major tweaking and fine-tuning that wouldn't have been achieved in a new stepping nor any short time frame. Thus they likely bypassed an official stepping and will refresh and introduce new chips made on a more mature process that should presumably offer better perf-per-watt. I still don't expect any miracles, but hopefully when they introduce the chips they can drop the prices on all of them to an appropriate level where they make sense
 
Pentium D = Dual Core Pentium 4 (AKA a chernobyl chip)
Core Duo = Original Dual Core for mobile systems (Core 2 Duo came from this)
Core 2 Duo = Dual Core for desktops based on CD
Pentium Dual-Core = Wolfdale 775 chips. E5300, etc. My E5300 clocked like a monster. 4.2GHz on air.

If these new models of BD come out, let's hope they have some sort of fix the shoddy performance. To quote logan who was at tigertv:

http://www.youtube.com/watch?v=IsjsDNjDaMw

go to 0:50 and listen for about 10 seconds... that's all you need to know ;)
 
LinkinMcOwnage said:
Pentium D = Dual Core Pentium 4 (AKA a chernobyl chip)
Core Duo = Original Dual Core for mobile systems (Core 2 Duo came from this)
Core 2 Duo = Dual Core for desktops based on CD
Pentium Dual-Core = Wolfdale 775 chips. E5300, etc. My E5300 clocked like a monster. 4.2GHz on air.

If these new models of BD come out, let's hope they have some sort of fix the shoddy performance. To quote logan who was at tigertv:

http://www.youtube.com/watch?v=IsjsDNjDaMw

go to 0:50 and listen for about 10 seconds... that's all you need to know ;)
Click to expand...

A new stepping won't fix the performance at the same clock speeds. A new stepping would only be able to improve clock speeds, overclockability, and lower power consumption.
 
either way, an appropriately priced FX-8140 combining a 95W TDP with a 3.2GHz clockspeed would be pretty tempting to me............... provided i can get a 1070 mATX motherboard to go with it!
 
Tsumi said:
A new stepping won't fix the performance at the same clock speeds. A new stepping would only be able to improve clock speeds, overclockability, and lower power consumption.
Click to expand...

True in most cases but if a processor was shipped with performance features disabled due to a hardware defect (not saying FX is like this, just in general), a new stepping to correct them along with any necessary BIOS revision could provide some IPC improvement.

I'm not holding my breath though
 
teletran8 said:
I'm surprised no one has commented on the increase in L3 cache. They DOUBLED the size of L3 cache, which I bet was a huge bottleneck for the first wave of BD CPU's.

Good job AMD! (Hopefully those numbers are true ;)) Also 95 watts is another improvement. Lets see what happens next, but they did the obvious so far.

FX 4100 = 6MB
FX 4150 = 12MB

FX 6100 = 8MB
FX 6150 = 16MB

FX 8120, FX 8150 = 8MB
FX 8140 = 16MB

They should easily get a 10% improvement in IPC with this new addition (If the low cache was a bottlenck for the first BD module design that was released...I hope it was), and should finally be a upgrade to Thuban owners I believe :) Maybe a couple extra 100 MHz OC compared to the first BD batches if temps are lower also. Sounds good to me.
Click to expand...

Sounds good, but I'm not holding my breath either.
 
It isn't doubled, they're just counting both the L2 and the L3, which is 16MB for the 4-module parts. The amount of cache will stay the same

2012021801_AMD_FX-Series_lineup_for_Q1_2012.jpg


The amount of L3 stays the same in all chips, 8MB. The L2 differs because each module gets 2MB shared, therefore a 2-module part would have 4MB L2 + 8MB L3 = 12MB cache. 6MB L2 + 8MB L3 = 14MB for the 3-module parts. 8+8 for the 4-modules
 
The 8140 looks like a normal (and somewhat belated) process improvement from GloFo. It runs a little faster than an 8120, while using a little less power. Assuming this is the case, overclockers may soon be able to squeeze a few more 100MHz out of the entire line of BD chips.

A new stepping would still be welcome if AMD were able to improve IPC in any meaningful manner. A BD varient with smaller-but-faster L2 might show improvement on typical desktop apps.
 
did this CPU get forgotten about with the release of the new FX four and six series last week.........?
 
You must log in or register to reply here.
Back
Top