Legit Reviews - FX-4100 Quad Core 3.6GHz Bulldozer Processor Review

[Nemesis 1]

Last I heard late 3Q early 4Q12 on the plus side, that gives AMD time to move BD to Piledriver and hopefully fix a fair number of issues. Trinity is 'supposed' to improve IPC by ~10-15%. Of course, AMD needs this now, but we'll only get BD B3 (next quarter?).

One can only hope that with the extra time BDII will see a 15-20% bump in IPC (which is reasonable considering the number of bottlenecks that reviewers have found in BDI's implementation and the fact that Piledriver has already been implemented in Trinity).

This is the problem so many had befor the BD release . Reading comprehension.

BD is what it is . Thats it . Sure it will improve but you act as if were all stuck in dead space. You assume BD II brings improvements . We all do . BUt 15% IPC no.

Than you also act as if intel is standing still . Their not! When Pile Driver shows up its going to be an even harder sell . As most people with a half a brain will wait to see IB results The only exceptions will be AMD zombie types who have no brain .

 

[Ajay]

Piledriver is supposed to improve performance per watt by 10-15%. AFAIK they've never claimed IPC would improve by that much.

I am aware of the graph you are referring to.

Some recent rumors have indicated a 10-15% performance increase for Trinity. Given the case that Trinity's top bins (w/Piledriver) will still have a 125W TDP, I'm incline to believe that that the 'watts' will be about the same, so we are likely to see a real improvement in IPC.

 

[Ajay]

This is the problem so many had befor the BD release . Reading comprehension.

BD is what it is . Thats it . Sure it will improve but you act as if were all stuck in dead space. You assume BD II brings improvements . We all do . BUt 15% IPC no.

Than you also act as if intel is standing still . Their not! When Pile Driver shows up its going to be an even harder sell . As most people with a half a brain will wait to see IB results The only exceptions will be AMD zombie types who have no brain .

I didn't even mention Intel, I was only talking about BD. I'm quite aware that Intel isn't 'standing still'.

Given that there are at least three main causes for BD poor performance (slow cache, slow integer performance and limited sustained issue width in the decoder) - there are real opportunities for substantial gains in IPC.

*By the way, IB will be out b/4 BDII so no one will need to wait for anything except BDII. Lastly, I'm no AMD zombie, my last 3 Intel systems have served me very well. I'm just rooting a bit for the underdog - glass half full.

What will we get - it really depends on the resources AMD has committed to BDII. With a substantial commitment of engineering resources, I think 15% is reasonable, especially given that Piledriver is already implemented in Trinity engineering samples.

 

[AtenRa]

What people don’t understand yet is that Bulldozer Architecture was designed for Integer calculations and FPU SIMDs, SSE, FMA4 etc.

Today’s Desktop applications are only coded for Legacy FPU (they don’t use SSE etc) and Bulldozers Integer Execution Units are under-performing.

I have said it before, BD was designed for TLP and DLP, its performance in today’s Desktop software doesn’t stretch its capabilities and it doesn’t show what the architecture is capable off.

About FX4100, yes it is slower (Default Clocks) in today’s apps against AMDs Phenom II and Intel’s Core i3 but FX4100 can be OverClocked when core i3 cant. Against Phenom II it can OC more and it has lower power consumption.

For a low budget Gaming PC, FX4100 can give you the same performance (if OCed) as any other high end CPU in the market in most of todays and future games.

I would definitely recommend it against any Core i3 and Phenom x4 for Gaming and everyday use.

 

[Gikaseixas]

For a low budget Gaming PC, FX4100 can give you the same performance (if OCed) as any other high end CPU in the market in most of todays and future games.

I would definitely recommend it against any Core i3 and Phenom x4 for Gaming and everyday use.

they do mention something like 27% improvement in one aplication just from oc it. Another positive is that it went up 1000 mhz without a single voltage increase. People aren't those good points in its favor??? Ok the arch sucks bad, i know and i'm not defending it but we do have to recognize the few positives.

 

[StrangerGuy]

they do mention something like 27% improvement in one aplication just from oc it. Another positive is that it went up 1000 mhz without a single voltage increase. People aren't those good points in its favor??? Ok the arch sucks bad, i know and i'm not defending it but we do have to recognize the few positives.

http://www.legitreviews.com/images/reviews/1766/cinebench-overclock.jpg

Even OCed to 4.6GHZ FX-4100 manages only 3.78 pts, while a STOCK, 2+ year old, 3.2GHz 955 BE already does 3.82 pts. Besides current 955 BEs easily OCs to 4GHz.

Nothing positive to this steaming pile of crap AT ALL.

 

[AtenRa]

http://www.legitreviews.com/images/reviews/1766/cinebench-overclock.jpg

Even OCed to 4.6GHZ FX-4100 manages only 3.78 pts, while a STOCK, 2+ year old, 3.2GHz 955 BE already does 3.82 pts. Besides current 955 BEs easily OCs to 4GHz.

Nothing positive to this steaming pile of crap AT ALL.

So, FX4100 is not good for Cinebench and that makes it a steaming pile of crap in general ?? :\

http://www.legitreviews.com/images/reviews/1766/cinebench.jpg
Well look at that, Intel Core i3 2105 at 3.1GHz is a steaming pile of crap because it looses from AMD's 2-year old core-Architecture Llano A6-3650 2.6GHz.

 

[Ancalagon44]

Bulldozer Defense Force deployed already I see.

 

[Gikaseixas]

Bulldozer Defense Force deployed already I see.

this is getting old, couldn't yuou come up with something better and more relevant to say???

Who's defending what? I'm just pointing out some positives, things most of the guys here seem to be ignoring like the 1000 mhz oc without a voltage increase (and with the stock cooler). The i3 won't be able to challenge it since it is locked, it's a fact.

 

[Gikaseixas]

So, FX4100 is not good for Cinebench and that makes it a steaming pile of crap in general ?? :\

http://www.legitreviews.com/images/reviews/1766/cinebench.jpg
Well look at that, Intel Core i3 2105 at 3.1GHz is a steaming pile of crap because it looses from AMD's 2-year old core-Architecture Llano A6-3650 2.6GHz.

that fella has a syndrome called "bias", just ignore him

 

[Chiropteran]

http://www.legitreviews.com/images/reviews/1766/cinebench-overclock.jpg

Even OCed to 4.6GHZ FX-4100 manages only 3.78 pts, while a STOCK, 2+ year old, 3.2GHz 955 BE already does 3.82 pts. Besides current 955 BEs easily OCs to 4GHz.

Nothing positive to this steaming pile of crap AT ALL.

How often do you use cinebench? I never do. I never use ANY of the artificial benchmarks these reviews seem to think are so important.

Why should I or anyone else care about these results?

 

[Gikaseixas]

It's a good tool to measure performance, both single and multi thread.
Nothing wrong in using cinebench in reviews.

The issue here is that people just criticize it left and right and do not aknowledge the positives. This cpu is actually making the i3 less desirable when it comes to overclocking. People on a budget will be able to spend just $110 and get decent performance, something the i3 won't be able to offer.

 

[exar333]

If we ignore performance, ignore power consumption, and ignore that we have had essentially better CPus for YEARS, BD is OK.

 

[Gikaseixas]

Ok, let's see:

Performance - better than the $135 i3 2105 in x264 HD Encoding, Handbrake, Pov-Ray 3.7, PC Mark 7, Resident Evil 5, Sandra 2011 Multi Media, Sandra 2011 SP5 Cryptography. Also better than the i3 2120 on S.T.A.L.K.E.R.: Call of Pripyat

Power Consumption - better than everything AMD and Intel i7 990X, worse than the rest of the intel lineup in the test

Let's not generalize, the focus is the FX 4100 review, not the BD line up. Would you agree that the cpu in question here performs better than the two i3 it was compared with?
That's my point, it does IMO and is priced competitively. Not bad for a $110 cpu

 

[nemesismk2]

deleted by myself

 

[Kevmanw430]

Gikaseixas, that is a very good point. It is priced competitively to the i3's, and it can overclock. However, its still slower than a similarly priced 955BE, which also overclocks decently.

 

[LOL_Wut_Axel]

Last I heard late 3Q early 4Q12 on the plus side, that gives AMD time to move BD to Piledriver and hopefully fix a fair number of issues. Trinity is 'supposed' to improve IPC by ~10-15%. Of course, AMD needs this now, but we'll only get BD B3 (next quarter?).

One can only hope that with the extra time BDII will see a 15-20% bump in IPC (which is reasonable considering the number of bottlenecks that reviewers have found in BDI's implementation and the fact that Piledriver has already been implemented in Trinity).

Trinity is supposed to improve performance by 10-15%, not IPC. If performance increases by that much then the IPC increase will be much lower. Performance can be increased via higher clock speeds AND higher IPC, but as we all know Bulldozer is made to aim for high clock speeds like the Pentium 4. Don't get excited for a big jump in IPC because it won't happen.

 

[LOL_Wut_Axel]

Ok, let's see:

Performance - better than the $135 i3 2105 in x264 HD Encoding, Handbrake, Pov-Ray 3.7, PC Mark 7, Resident Evil 5, Sandra 2011 Multi Media, Sandra 2011 SP5 Cryptography. Also better than the i3 2120 on S.T.A.L.K.E.R.: Call of Pripyat

Power Consumption - better than everything AMD and Intel i7 990X, worse than the rest of the intel lineup in the test

Let's not generalize, the focus is the FX 4100 review, not the BD line up. Would you agree that the cpu in question here performs better than the two i3 it was compared with?
That's my point, it does IMO and is priced competitively. Not bad for a $110 cpu

Biased and wrong. The 2105 is just a 2100 with HD 3000 graphics. The FX-4100 does not even come with an IGP, so right off the bat your comparison is biased. You can also get the i3-2120 for $123, making this a CPU both faster AND cheaper than the i3-2105. Price difference would be under $15 between the FX-4100 and the i3-2120.

Also, pretty much everything they used was multi-threaded, which makes it look better than what it is. Even then, it was barely faster while consuming lots more power and actually being slower than a Phenom II X4 and comparable only to an Athlon II X4. What's the incentive to buy this when the Phenom II X4 955 has comparable performance/watt, is faster, and only costs $10 more? For that matter, why not the i3-2120 instead? If we look at the overall picture the Phenom II X4 and Core i3s are pretty evenly matched, with the i3 having a big advantage in power consumption and heat produced.

 

[daveybrat]

Funny thing for AMD is this.....Bulldozer is making Phenom II Quads and Hex-Cores much more attractive and in higher demand than ever before.

I myself am now eyeing a new 1090T to replace my Athlon II cpu.

Funny how Bulldozer is increasing sales of older chips which i'm quite sure is the opposite of what AMD actually wanted. LOL!

 

[Ajay]

Trinity is supposed to improve performance by 10-15%, not IPC. If performance increases by that much then the IPC increase will be much lower. Performance can be increased via higher clock speeds AND higher IPC, but as we all know Bulldozer is made to aim for high clock speeds like the Pentium 4. Don't get excited for a big jump in IPC because it won't happen.

Maybe it won't. But the clocks on Trinity are a bit lower than Bulldozer for the top bins. That chart actually shows performance/watt, but the TDP for top binned Trinity APUs will be the same as BD (125W). Hence, that 10-15% increase in performance/watt will likely come from improved IPC. Seriously, there are plenty of areas in the implementation of BD that need improvement - so it's entirely possible to get that much IPC improvement. And since Piledriver is already implemented on Trinity Eng. samples - AMD is a step ahead already in putting out BDII. AMD's engineers aren't idiots, if we can see the problems (take a look at the Real Tech forums, or Lost Circuits) then they certainly already knew the shortfalls when designing Piledriver. There is allot of potential for improvement and a significant amount of work has already been done (designing and implementing the Piledriver module). Maybe they'll fail again, but I think it more likely that they will succeed - we'll just have to see how well.

On top of that, one would think that GF's 32nm process will be performing better 2h12 than it is now.

 

[LOL_Wut_Axel]

Maybe it won't. But the clocks on Trinity are a bit lower than Bulldozer for the top bins. That chart actually shows performance/watt, but the TDP for top binned Trinity APUs will be the same as BD (125W). Hence, that 10-15% increase in performance/watt will likely come from improved IPC. Seriously, there are plenty of areas in the implementation of BD that need improvement - so it's entirely possible to get that much IPC improvement. And since Piledriver is already implemented on Trinity Eng. samples - AMD is a step ahead already in putting out BDII. AMD's engineers aren't idiots, if we can see the problems (take a look at the Real Tech forums, or Lost Circuits) then they certainly already knew the shortfalls when designing Piledriver. There is allot of potential for improvement and a significant amount of work has already been done (designing and implementing the Piledriver module). Maybe they'll fail again, but I think it more likely that they will succeed - we'll just have to see how well.

On top of that, one would think that GF's 32nm process will be performing better 2h12 than it is now.

That's the main reason why I think the performance improvement will come mostly from higher clock speeds. They don't have too much time (only ~6-9 months) to tweak the architecture for 10-15% higher IPC and get it released.

Faster and lower latency L1 and L3 caches are the first thing they need to improve IPC, and they could certainly get a 5% boost from it. I think higher clock speeds will still account for half or 3/4 of the performance improvements.

 

[apoppin]

Biased and wrong. The 2105 is just a 2100 with HD 3000 graphics. The FX-4100 does not even come with an IGP, so right off the bat your comparison is biased. You can also get the i3-2120 for $123, making this a CPU both faster AND cheaper than the i3-2105. Price difference would be under $15 between the FX-4100 and the i3-2120.

Also, pretty much everything they used was multi-threaded, which makes it look better than what it is. Even then, it was barely faster while consuming lots more power and actually being slower than a Phenom II X4 and comparable only to an Athlon II X4. What's the incentive to buy this when the Phenom II X4 955 has comparable performance/watt, is faster, and only costs $10 more? For that matter, why not the i3-2120 instead? If we look at the overall picture the Phenom II X4 and Core i3s are pretty evenly matched, with the i3 having a big advantage in power consumption and heat produced.

No kidding the Phenom II X4 and the i3-2105 are evenly matched. Hyper-Threading plays more of a role than i originally thought.

 

[iCyborg]

No kidding the Phenom II X4 and the i3-2105 are evenly matched. Hyper-Threading plays more of a role than i originally thought.

Hm, I can't say I'm impressed with HT for gaming from that chart...

 

[Ajay]

Faster and lower latency L1 and L3 caches are the first thing they need to improve IPC, and they could certainly get a 5% boost from it. I think higher clock speeds will still account for half or 3/4 of the performance improvements.

I don't think they will have time to fix the L3 cache, it's typically done by hand - an 8MB cache would take time. Improving L1 and the decoder issue width would help allot - and I imagine some of that work has already been done for Trinity. If this Chart can be believed, the FX-8170 will be released @ 3.9GHz w/B3 stepping in 1Q12 - that's a 10% bump for the top binned part and B3 may have target implementation flaws that were limiting clock frequency.

I think AMD may have a competitive part, performance wise, when BDII comes out - but IB will kill it in power efficiency.

 

[LoneNinja]

I don't think they will have time to fix the L3 cache, it's typically done by hand - an 8MB cache would take time. Improving L1 and the decoder issue width would help allot - and I imagine some of that work has already been done for Trinity. If this Chart can be believed, the FX-8170 will be released @ 3.9GHz w/B3 stepping in 1Q12 - that's a 10% bump for the top binned part and B3 may have target implementation flaws that were limiting clock frequency.

I think AMD may have a competitive part, performance wise, when BDII comes out - but IB will kill it in power efficiency.

If AMD can really put out a competitive product, Intel could just release higher clocks sacrificing some of their power efficiency for absolute performance. Or they'll just use IB-E to crush Piledriver. I just can't see AMD becoming competitive any time soon unless Intel decides to let them.