FX 8370 Review

[Phynaz]

I'm really not finding myself impressed. Lower power consumption is better, but the clocks are so low that a $199 i5 just kills the 8370 in both performance and power consumption.

 

[nenforcer]

Am I missing something?

Or does this Formula 1 game from codemasters LOSE performance when switching to SLI 2 X GTX 770's?

 

[AtenRa]

Am I missing something?



Or does this Formula 1 game from codemasters LOSE performance when switching to SLI 2 X GTX 770's?

F1 doesn't support CF or SLI

 

[Ramses]

I've about decided the FX series is a big middle finger demonstrating that they can do whatever turns them on with the profits from the graphics card division.

And I kinda like that.

 

[monstercameron]

I've about decided the FX series is a big middle finger demonstrating that they can do whatever turns them on with the profits from the graphics card division.

And I kinda like that.

interesting perspective

 

[VirtualLarry]

I've about decided the FX series is a big middle finger demonstrating that they can do whatever turns them on with the profits from the graphics card division.

And I kinda like that.

But giving the finger to whom? Intel? Or their own customers?

 

[monstercameron]

But giving the finger to whom? Intel? Or their own customers?

most likely to themselves.

 

[SlowSpyder]

http://www.computerbase.de/2014-09/amd-fx-8370e-im-test/6/

Ah, thanks. That is pretty decent, then. There are probably a handful of FX 8350's that reach 5GHz on air, but probably pretty far and between. Of course I can't read that article, so I'm not sure if they mention that as a stable OC or not. But still, not a bad OC on air.

I've about decided the FX series is a big middle finger demonstrating that they can do whatever turns them on with the profits from the graphics card division.

And I kinda like that.

I felt like that, and also liked it, about the FX 9xxx CPU's. It is like AMD said, this is all we have right now, so f*** it, crank it to 11.

 

[BigDaveX]

Kind of like what Intel did ten years ago. After they pulled the plug on Netburst, all they gave us for the next two years was Prescott 2M (a repurposed Xeon die), the Pentium D (two P4s crudely slapped together on the same package) and Cedar Mill (a dumb shrink of Prescott 2M).

Still, even THAT was more than what AMD's done with the FX line over the last two years.

 

[Abwx]

Ah, thanks. That is pretty decent, then. There are probably a handful of FX 8350's that reach 5GHz on air, but probably pretty far and between. Of course I can't read that article, so I'm not sure if they mention that as a stable OC or not. But still, not a bad OC on air.

You can use google translate, german is close to english, automated translations are quite enough.

They say that it underclock better that it undervolt given that its voltage is lower than for the other FXs.

reduction of only 0.1 volts is punished in Windows with a crash

Without much effort and special settings, the test sample is clocked at 5 GHz at 1.45 volts.

Looking at Anandtech s 127W power delta with AVX, wich is one of the worst case scenarii, the CPU TDP in said test is about 85-90W.

 

[SlowSpyder]

Kind of like what Intel did ten years ago. After they pulled the plug on Netburst, all they gave us for the next two years was Prescott 2M (a repurposed Xeon die), the Pentium D (two P4s crudely slapped together on the same package) and Cedar Mill (a dumb shrink of Prescott 2M).

Still, even THAT was more than what AMD's done with the FX line over the last two years.

AMD isn't in the same position, though. They don't have any special Opteron silicon they can pull out. They can't fit two dies on one AM3+ sized package (and if they could it wouldn't help the FX's weakest area, single threaded performance, but would certainly add to power consumption). So all they have left to do is make available more options, like a 95 watt four module CPU and a power-hungry high TDP part.

These new parts are interesting, but I think the 8320-8350 is still the way to go for AM3+, when you consider the CPU's price.

 

[Abwx]

. It's also slower than the 8150 in the single-threaded part of the 3D Particle Movement benchmark as well.

In single thread i get the same score than a FX8350 with a Pentium T4400 2.2GHz in this so called benchmark.

 

[Enigmoid]

In single thread i get the same score than a FX8350 with a Pentium T4400 2.2GHz in this so called benchmark.

Really, whats your score?

 

[BigDaveX]

In single thread i get the same score than a FX8350 with a Pentium T4400 2.2GHz in this so called benchmark.

On the other hand, the ancient Phenom II X6 clobbers the entire line-up of Intel chips (aside from the 990X) in the multi-threaded part of the test. Like many, the benchmark doesn't hate AMD in general, just the FX series (more seriously, I'm guessing it's heavy on FP code, which is FX's main weakness next to the Phenom and Intel's chips).

In any case, the point I was making is that there's no reason the 8370E should be slower than the 8150, given that it has a higher turbo speed, and from what I've read the original Bulldozer core was less aggressive when it came to turboing up than Piledriver. Given the difference between AT's Cinebench tests and Tech Report's, it looks like either AT got an awful sample, or something on their motherboard wasn't set up correctly.

 

[Abwx]

Really, whats your score?

ST score is 71.74, so it has nothing to do with hyperthreading as you said but all with intel processors, according to this bench a Bay Trail has about 40-50% higher ST FP IPC than a Piledriver.

This benchmark, or rather trashmark, was "designed" by the reviewer himself, i didnt think that Anandtech could sink so low.

On the other hand, the ancient Phenom II X6 clobbers the entire line-up of Intel chips (aside from the 990X) in the multi-threaded part of the test.

Because of the six cores but the same bench say that a Bay trail has about 107% of the ST FP IPC of this phenom wich is just ridiculous, this is a rigged bench and the numbers prove it.

As for the 8370E it is running at 3.8 in ST , likely that the bios wasnt updated according to Hfr reviewer, wich is indeed very unprofessional but i already knew about it.

 

[Rvenger]

F1 doesn't support CF or SLI

Eh? Why did they even run it then?

 

[BigDaveX]

ST score is 71.74, so it has nothing to do with hyperthreading as you said but all with intel processors, according to this bench a Bay Trail has about 40-50% higher ST FP IPC than a Piledriver.

This benchmark, or rather trashmark, was "designed" by the reviewer himself, i didnt think that Anandtech could sink so low.

Because of the six cores but the same bench say that a Bay trail has about 107% of the ST FP IPC of this phenom wich is just ridiculous, this is a rigged bench and the numbers prove it.

I dunno... seems like if you wanted to write a benchmark with the express purpose of making Intel look good and AMD look bad, you'd kind of want to avoid having it show the latest and greatest Haswell-core i5s getting defeated by the Phenom II X6, which has a CPU core dating back to 2007 (and even with the extra cores generally hasn't been comparable to any i5 since about the Sandy Bridge era).

 

[monstercameron]

what makes a good benchmark?

 

[Ramses]

But giving the finger to whom? Intel? Or their own customers?

Pretty much everyone near as I can tell. They put out a 5ghz cpu with an AIO water cooler that takes a crazy amount of unapologetic juice for cripes sake. If that wasn't a really funny middle finger I don't know what was. Maybe there extreme and edgy now, I'm old and can't tell.


I'm amused and OK with it though myself, in no small part because I'm very happy with my 8350.
Benchmarks are fine and interesting and all that, and it's another story if one is trying to do something that really needs serious single threaded power, but for what I do this thing is comically fast. Almost all modern CPU's are. I still have fifty tabs open in firefox, I still play crazy games, I do some light 3d junk for work and a ton of light image work, I spend 18 hours a day in front of this thing, and it's really fine, with an ssd and a bunch of ram I'm never waiting on the processor. If fact my cpu spends most of it's time half asleep. And it was cheap. For a less than consumer computer anyway. I didn't notice my power bill go up, and I do watch it, and it runs cool with an "enthusiast" grade cooler and case. I can't find anything to complain about, other than I want a 9590 for $200 or under just so I can say I have it. If I could have waded through the two hundred Intel motherboards out there and made sense of there processor naming scheme, I might have got out with some sort of midrange chip and board that was the same price or not much more, but I'd really rather have something higher end and more "neat" than something pedestrian just because it's a little faster. That's personal preference though, but I suspect others share it.

Sometimes I look up some benches for a "fast" cpu from ten years ago that we did all the same stuff with that we do today on a day to day bases for some perspective.
I keep an old socket 478 3.2ghz HT P4 box in my garage for netflix and music while I'm working. That was a relatively fast computer in it's day. And it's faster than the dual P2's I rocked for a long time. It'd be cool to see AMD competitive with intel on the high end, but I keep myself anti-jaded enough that it's not too big a bummer.

That's my holistic approach to the whole thing.

 

[Ramses]

what makes a good benchmark?

A related question I had early on was how can one cpu/system "feel" better than
another that scores better? I grew this question back when I was getting into
SMP and Pentium Pro's and Windows NT4. To my knowledge nobody ever figured
out a way to measure that, and I suspect they are all so fast it's less significant
these days. I also suspect there was some psych involved.

 

[SlowSpyder]

Unless you're running your CPU at or near 100% load for extended periods of time, like with distributed computing, the power thing is overrated. At least from a cost perspective. Games don't load up eight cores anywhere near 100% generally.

 

[Enigmoid]

Unless you're running your CPU at or near 100% load for extended periods of time, like with distributed computing, the power thing is overrated. At least from a cost perspective. Games don't load up eight cores anywhere near 100% generally.

Depends, Atenra tested a number of CPUs with thief on DX11 and mantle.

http://forums.anandtech.com/showthread.php?p=36301071

8350 is using 50+ W more than the 3770 and ~100W more than the i3-4330 under both DX 11 and mantle.

I agree with you in a general sense but FX idles significantly higher (FM2+ is competitive with intel) and is significantly less efficient.

 

[Abwx]

8350 is using 50+ W more than the 3770 and ~100W more than the i3-4330 under both DX 11 and mantle.

.

FX has a chipset that has 25.9W TDP while a Z97 FCH is 4.1W, at full load their footprints on the power measuread at the main are 32W and 5W respectively so the 8350 does not use 50W more power but surely that the chipset is power hungry, for exemple a X99 has 6W TDP wich will translate in about 7.2W at the main, AMD better had released an adequate one even if it was a cheap solution with minimal USB3 and SATA 3 ports count.

 

[SlowSpyder]

FX has a chipset that has 25.9W TDP while a Z97 FCH is 4.1W, at full load their footprints on the power measuread at the main are 32W and 5W respectively so the 8350 does not use 50W more power but surely that the chipset is power hungry, for exemple a X99 has 6W TDP wich will translate in about 7.2W at the main, AMD better had released an adequate one even if it was a cheap solution with minimal USB3 and SATA 3 ports count.

Either way, let's round up and say the platform uses 100 watts more. Even if you game five hours (at a level that uses 100 watts more) a day at $.20/kilowatt hour that's about $26/year. I don't game near that much (35 hours a week... lol, not even close, maybe some of you do though). To me that's just not even a consideration in the grand scheme of things.

 

[Abwx]

Either way, let's round up and say the platform uses 100 watts more. Even if you game five hours (at a level that uses 100 watts more) a day at $.20/kilowatt hour that's about $26/year. I don't game near that much (35 hours a week... lol, not even close, maybe some of you do though). To me that's just not even a consideration in the grand scheme of things.

Averaging power comsumption between Cinebench single thread and multithread difference is 33W between a 4790K and a 8350 out of 107.5 and 140.5W respectively, add the GC and the relative difference will be even more pointeless, your 100W are quite generous if i can say, it s telling about how much a dead horse can be beaten..

http://www.computerbase.de/2014-09/...mm-leistungsaufnahme-volllast-cinebench-x-cpu