AMD Launches FX-6200 and FX-4170 Processors

H

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
AMD sends word that the company is launching two new processors today, the FX-4170 and FX-6200. According to the press release, the FX-4170 is a quad core part running at 4.2GHz 4MB L2 / 8MB L3 cache with a max turbo speed of 4.3GHz and the FX-6200 is a six core CPU with 6MB of L2 / 8MB L3 cache running at 3.8GHz with a max turbo speed of 4.1GHz. No pricing information was available.
 
Last edited:
Steve, do you have any solid info on whether these are on a new stepping or just a clock speed increase with higher TDP?
 
Sweet, might get another one going soon for VM goodness. Thanks for the link.
 
This would happen a week after I put together a 4100 system. Its clearly a dumb upgrade since mine will run 4.4 ghz on stock cooler. Seems my wait for piledriver will be a long one as I'm sure they're going to Guy e these some time to sell before they push out something more powerful.
 
For the price though, the FX-4100 isn't a bad processor, especially for a small VM farm on a budget.
 
^Agree, my 4100 system will do any light duty work in a flash and I'm totally satisfied with it. Sure it doesnt look good on paper but youd never noticed the difference (usually) in real world computing vs its competition.
 
HardOnForMe said:
The 8150 @ $185 seems very tempting... for some.
Click to expand...

That's the 8120. 8150 is still at $250.

If Microcenter drops the 8120 to $185, and keeps the same $120 motherboard discount, I would be extremely tempted to get one just to play with.
 
Tsumi said:
That's the 8120. 8150 is still at $250.

If Microcenter drops the 8120 to $185, and keeps the same $120 motherboard discount, I would be extremely tempted to get one just to play with.
Click to expand...

I would drive to Atlanta and pick up a combo if they did that...lol. I'm really just holding out for the 95w 8 core to put in my whitebox. My x3450 is cool and all but I hate being hit or miss when trying to get a 1156 motherboard that supports vt-d.
 
I'm hoping this means good news for the 8 core parts so that I can have an eventual significant upgrade. My 8120 at 4.4 GHz is plenty of horsepower for now though.
 
I am wondering if they did something to improve how the cores function in windows 7? Core scheduling or something along those lines?
 
Why the heck are they even bothering to use Turbo on the 4170? A 100MHz turbo boost will pretty much do nothing to performance!
 
tangoseal said:
I am wondering if they did something to improve how the cores function in windows 7? Core scheduling or something along those lines?
Click to expand...

I would highly doubt that. I believe fixing the slow / high latency cache would improve the core efficiency and reduce the need for perfect scheduling however this most likely will come with Piledriver.

Edit: Although they could have a new stepping to fix some bug that they had to reduce the cache performance to avoid the bug.

GuruX said:
Why the heck are they even bothering to use Turbo on the 4170? A 100MHz turbo boost will pretty much do nothing to performance!
Click to expand...

Agreed. Turbo cant be that effective when its reduced to 100 MHz. Maybe they were worried about yields on a 4.7GHz turbo.
 
Last edited:
Seams like they are already using the resonant clock mesh on these ones.
Correct me if i am wrong.
 
biggston said:
Seams like they are already using the resonant clock mesh on these ones.
Correct me if i am wrong.
Click to expand...

I would also doubt that. Resonant clock mesh is for PileDriver.This to me looks like just better yields at GF.
 
drescherjm said:
I would also doubt that. Resonant clock mesh is for PileDriver.This to me looks like just better yields at GF.
Click to expand...

Well i just sed that cause i see that low turbo mode :p
Its still is possible to me that they have used it.
I could be wrong.
 
I believe the low turbo mode is a result in not being confident to yield enough chips that can do 4.7GHz in turbo mode.
 
drescherjm said:
I believe the low turbo mode is a result in not being confident to yield enough chips that can do 4.7GHz in turbo mode.
Click to expand...

Well to be honest to you,i hope its not the resonant clock mesh.why is say this?
Simple,hopefully they figured out some of the problems for the architecture.
 
Do these guys compare well against Phenom II X4 9XXs in single core strength? I'm doing a bunch of singlecore software testing (I know - 1) argh and 2) should run Intel instead) and got a 965 at the moment, dunno if I should clock that or get one of these that's nearly a gigahertz faster.
 
MrLonghair said:
Do these guys compare well against Phenom II X4 9XXs in single core strength? I'm doing a bunch of singlecore software testing (I know - 1) argh and 2) should run Intel instead) and got a 965 at the moment, dunno if I should clock that or get one of these that's nearly a gigahertz faster.
Click to expand...
Haven't you read any reviews? Bulldozer went backwards in single threaded performance compared to Thuban. Thuban is behind Sandy Bridge.
Software testing? Why? Look at the reviews, specifically Cinebench single core numbers. They are out there.

Another thing you can do is look at this SolidWorks Punch Holder benchmark.
http://www.solidmuse.com/benchmarks/solidworks-benchmark-punch-holder/
Results: https://spreadsheets.google.com/pub?key=0AssNoiBTiyqGcHo3d1RwSWtDN0xBMjh5YkV5eHlUUHc

When doing rebuilds of parts, Solidworks uses at most 2 cores, but mostly one.
This benchmark consists of rebuilding a part that was not modeled in an optimum way.
Anytime you make changes in Solidworks your part needs rebuilt, so the faster your cpu is for rebuilds the more work you can get done. The time for the rebuild is what people meaure for this benchmark. Well, Solidworks measures it, no stopwatches needed.
Overclocked Sandy Bridge is fastest, as would make common sense.
My i3 2100, stock of course since it doesn't really overclock, took 77 seconds. A 4 Ghz Thuban from someone else in the results spreadsheet took 76, which is just about the same. I had a 3.5 Ghz Phenom II a while ago and it took 94 seconds. An Athlon II at 3 Ghz or so takes about 150 seconds. The lack of L3 cache kills them. A 2500k, overclocked, flies and is a good choice for SolidWorks or any single threaded app, but this is already common knowledge. From the results spreadsheet there are 2500k and 2600k machines in the 40 to 50 second range.

The bottom line is that for single threaded performance you should go Intel for now. Not every Thuban will hit 4 GHz and even if you can it only puts you on par with an i3 2100.
No AMD will catch an i5 in single threaded performance.
 
Tsumi said:
That's the 8120. 8150 is still at $250.

If Microcenter drops the 8120 to $185, and keeps the same $120 motherboard discount, I would be extremely tempted to get one just to play with.
Click to expand...

I went off of engadget's report. They say 8150 @ $185.
 
MrLonghair said:
Do these guys compare well against Phenom II X4 9XXs in single core strength? I'm doing a bunch of singlecore software testing (I know - 1) argh and 2) should run Intel instead) and got a 965 at the moment, dunno if I should clock that or get one of these that's nearly a gigahertz faster.
Click to expand...

I did some single core testing of FX4100 vs an AMD 631. single threaded test the FX4100 was slightly faster due to its massive overclock headroom, but not clock for clock. Multithreaded (quad thread) the overclocked 3.9ghz AMD 631 beat the overclocked 4.4ghz FX4100 pretty bad 7558 vs 9552 in fritz chess.

on retail heatsink i couldnt get it stable 24x7 stable much past 4.3ghz. sure could complete benchmarks at 4.4-4.5 -- but i wouldnt call it stable.
 
biggston said:
Seams like they are already using the resonant clock mesh on these ones.
Correct me if i am wrong.
Click to expand...

No, they aren't. Bulldozer cores lack the meshing tech that is required for that clock resonance mesh tech they licensed for Piledriver. They're just higher TDP higher clocked chips, which may or may not be on 2.2ghz IMC compared to the other chips' 2.0ghz IMC. I doubt this is a new stepping considering the higher TDP, but if it is then that's actually more worse news for AMD rather than good news. 125W TDP and clock speed bump isn't what you expect from a chip that already used way too much power that was partially blamed on GloFo :p
 
No, they aren't. Bulldozer cores lack the meshing tech that is required for that clock resonance mesh tech they licensed for Piledriver. They're just higher TDP higher clocked chips, which may or may not be on 2.2ghz IMC compared to the other chips' 2.0ghz IMC. I doubt this is a new stepping considering the higher TDP, but if it is then that's actually more worse news for AMD rather than good news. 125W TDP and clock speed bump isn't what you expect from a chip that already used way too much power that was partially blamed on GloFo :p
Click to expand...

, they are releasing new 95 watt TDP parts and 125watt tdp parts. You don't go from 125watt TDP to 95 watt TDP with out a new revision could be wrong here, but it is likely these are a b3 stepping.

Performance should be identical to the current fx processors, power draw should be better, they might overclock a bit better as well. Also the resonant clock mesh can be applied to current Amd processors, there is no new tech needed for amd to add these new feature to their chips and they stated that.
 
Last edited:
The resonance mesh tech is pretty big news. Shame they couldn't get it on the Bulldozer. (I'm a long-time electronic and computer tech, and introducing a tuned inductor/capacitance "ringing" circuit is a stroke of brilliance. You have capacitance anyway, so why not use it?"

Perosnally, as far as performance goes, I think it's a secret weapon. All the news emphasis is about power savings (which it does, and does very well.) However, power reduction means heat reduction as well.

Crank that puppy up to the heat levels it was performing adequately at, and you should get a great performance boost...up to another 30%, and with Trinity and Kaveri using increased levels of parallel processing between the GPU/CPU, this should prove to be quite a punch for mobile processing.

I'm an A8-3500M user, and I'll be skipping Trinity in favor of Kaveri's GCN and increased parallelism. Should be a hell of a laptop, and especially in Crossfire with a well-matched discrete GPU.

No wonder AMD stock is up around 40% the last six weeks or so. They're showing some fantastic tech improvements which don't seem to have much of an end in sight, as well as targeting the major-money crowd: greener laptops and mainstream computing with more performance for the $.

While AMD may have given up the fastest-processor fight with Intel, it looks to me like they're going to get the lion's share of overall computing $$$, and especially if they can work a deal with Apple.
 
ebduncan said:
, they are releasing new 95 watt TDP parts and 125watt tdp parts. You don't go from 125watt TDP to 95 watt TDP with out a new revision could be wrong here, but it is likely these are a b3 stepping.

Performance should be identical to the current fx processors, power draw should be better, they might overclock a bit better as well. Also the resonant clock mesh can be applied to current Amd processors, there is no new tech needed for amd to add these new feature to their chips and they stated that.
Click to expand...

These new chips aren't 95W but rather 125W TDP. They're basically overclocked versions of the fx-6100 and fx-4100 that may have gotten a boost in their IMC. Whether that counts as a revision, I won't debate that. What I will say, though, is that revisions generally also entail lower power consumption, and that doesn't seem to be the case here. Thus you can't pin this on GloFo (which was a very popular scapegoat).

What they stated regarding resonance mesh tech licensing was directed toward Piledriver, not Bulldozer. AFAIK Bulldozer doesn't have a clock mesh where Piledriver had one anyway, which is why the implementation was referred to as seamless when they were speaking about Trinity. Piledriver had a clock mesh anyway, the integration of that tech was therefore seamless
 
Last edited:
pelo said:
What they stated regarding resonance mesh tech licensing was directed toward Piledriver, not Bulldozer. AFAIK Bulldozer doesn't have a clock mesh where Piledriver had one anyway, which is why the implementation was referred to as seamless when they were speaking about Trinity. Piledriver had a clock mesh anyway, the integration of that tech was therefore seamless
Click to expand...

I dug extensively, and I could find nothing supporting clock mesh resonance in Bulldozer, FWIW.

Funny, but I used to think AMD made a huge mistake in buying ATI. At this point, unless Intel were to purchase Nvidia and put a real burn on in to incorporate that tech with their CPU cores, I don't see as they're going to be able to keep up in the mainstream computer department.

The fact that, in a largely experimental design, AMD not only brought the APU to light, but CrossFireX to laptops, is huge and relatively unheralded, and is but the tip of the iceberg.
 
I am the one who brought this up in the first place.for the resonant clock mesh.
So the credit should go to me. :p joke
But seriously think about it.I sed this cause it was odd to me that the turbo boost was so low.
This can mean just two things,either they made some changes to the architecture to reduce power draw,
Or they could of applied this tech already on this new chips.
I see no reason to why they did not think to make this move,and they already stated that it has bean applied to the new piledriver architecture without having any issues.
So why not?
 
To apply a clock mesh would require a redesign of Bulldozer. I can almost guarantee you these chips do not have a clock mesh nor that clock mesh resonance tech. Judging by the higher TDP these are just overclocked versions of the fx-4100 and fx-6100. Redesigns don't take weeks, they take years. A tweaking amounts to Piledriver, which has been in the works since before Bulldozer was released. These things take far more time than you think they do.
 
Well we will see.
But comon man you think that the piledriver architecture would be much different than bulldozer?
Its based on the same design.
 
biggston said:
Well we will see.
But comon man you think that the piledriver architecture would be much different than bulldozer?
Its based on the same design.
Click to expand...

If I'm to judge by the clock speeds alone and even with the resonance mesh tech licensing offering a higher ceiling for the same power draw? I'm saying clock speeds because that's essentially all we have to go on. They did seem to increase the L1 data size, but that doesn't address the L1 write-through issues.

If I had to guess, Piledriver will be closer to Llano IPC but still off by ~%5. That would be the best scenario. But because the clock speeds are so high, I don't see them having done anything drastic with the L2 latency, which was probably BD's biggest downfall.

Trinity will be worth buying because of GPU performance and perf-per-watt alone, but I don't see Vishera beating Ivy or even Sandy Bridge. If AMD can make the processors cheap like they did with Deneb and Thuban, then that will be the biggest thing they'll have going for them. If you're expecting Vishera to compete with IVB or SB on performance, then no, it won't. Price? almost certainly. That's enough for a recommendation.

That's all hypothetical, though =P
 
Has nothing to do with thinking, the paths and the design have different objectives. But if you check certain posts they do not differ enough in performance to "show" something completely different.
 
Yes we are just speculating things here.
And i never said that vishera will compete with IVB.
In fact i said that i don't like the news about the resonant clock mesh,cause that just means that they focusing on alternatives to achieve some things.Not the architecture itself.
But i also did not say that the resonant clock mesh is bad tech.
 
Last edited:
What you have to grasp is that one of BD's supposedly biggest flaws was the cache latencies, and the higher the clock speed the higher the cache latency. Between the long pipeline, advertised high Trinity clock speeds and now that these chips are on 125W rather than 95W, I would think they haven't addressed the cache latencies significantly. Rather, the resonant clock mesh just allows them to clock a bit higher under the same thermal evenlope (or clocked the same under a lower thermal envelope), so it's just an asterisk rather than a reason. The tech accounts for a 100mhz bump, not 500mhz or higher and its most effective under speeds of 4ghz.

read my AT post to get a better idea.

http://forums.anandtech.com/showthread.php?t=2229071&page=2
 
PileDriver gains => FMA3 (Fast Memory Access 3) and Converged BMI (Bit Manipulation Instructions) + resonant clock mesh.

Built-in Memory Controller on Piledriver chips would be directly connected to 4 DDR3 DIMMS with stock frequency support of 1866Mhz in Dual Channel Mode

Clock-per-clock, about a (10-15% gain over Bulldozer) x (Windows 8 optimizations/fixes) x (whatever resonance clock mesh brings with increased clocks) x (higher frequencies already slated before resonant clock mesh)

Google is our friend.

1. As far as overhauled microarchitecture goes, both AMD and Intel have more limitations than ever before. The Law of Dininished Returns says the more they put directly into microarchitecture, the less they'll get out of it. As it is, both companies have made huge strides over the CPUs of yesterday.

2. Both have separate microarc directions, so they have to be careful now not to pee in each others' pools with dramatic changes. Enough lawsuits already.

3. Thinking outside of the core, there are still so many optimizations and new tech that, until now, has been largely ignored because of cost/size/tech limitations that no longer exist, thanks to mature microarc tech. Resonant clock mesh, embedded memory and i/o controllers, embedded parallel graphics that can also assist the GPU (GCN).

Parallel tech has always been faster and more efficient (and heavily used in old mainframes), it's just been more costly until component power and size could be reduced (something serial technology pioneered better, but can be applied directly to parallel now), and companies have not cooperated very well to merge their separate tech. Doubtful that AMD could have done that without acquiring ATI, because we have all seen, from companies lack of cooperation (Intel&Nvidia, Nvidia&RAMBUS, etc, etc) that partnerships just don't work as well as vertical corporations for emergent technology.
 
You must log in or register to reply here.
Back
Top