KitGuru tested the FX9590... it's pretty bad.

[NTMBK]

Has anyone considered that AMD purposely released this joke of a processor to dampen the blow of dumping their big core architecture? because that's pretty much the only hypothetical scenario where this chip makes sense to me

They seem to be dumping their big die range, but they don't seem to be ditching the big cores as such. Kaveri is still getting released. We will have to wait and see about the Excavator APU I guess...

 

[pelov]

I was jesting, but only partly. While some of you see benchmarks and figures, all I see are graphs with pretty coloring, stating:

Look at how far behind we are.

NTMBK, you're right. I should've specified desktop/traditional server. Their APUs are still truckin', and those look much better. Delicious, in fact.

 

[USER8000]

They seem to be dumping their big die range, but they don't seem to be ditching the big cores as such. Kaveri is still getting released. We will have to wait and see about the Excavator APU I guess...

I think they are focussing on CPUs which can make it into laptops. TBH,Intel is doing pretty much the same,although they have a bigger market for their large die CPUs which means they can still justify updates. However,you can still see how the release schedule has been stretched out.

Unfortunately,this is bad news for the AM3+ line,which probably will be put on the back burner with tweaks here and there.

 

[SPBHM]

You are also ignoring the fact that millions of players of these games are not having overclocked Core i5,Core i7 or FX8300 series CPUs.

...
You talk about TF2,I have run that fine on my backup rig wich has Core2 level CPU performance during big battles(Saxton Hale maps) with loads of players crowded into one area.

again, you can play with old hardware, but it doesn't mean you get the same quality, these guys playing with Core 2 Duos are not using $200+ new graphics card, I'm saying you should go for the best gaming CPUs, when you also have a good quality VGA and is aiming for quality (shiny looking games at a high frame rate)

TF2 is a funny game for me, I've been playing since my k8 x2, it would run like crap, like sub 30fps in some areas, but after they implemented "multi core rendering" it improved, but still, even with my c2q it would go bellow 50 in bad situations, and even with the i3 2100 it's still go easily down to that range, I recorded a demo and at the same place the i3 and and an OCed PII X4 was doing avg 60 (and you really feel the slowdowns), an overclocked i5 was doing 90 (2fort with 32p fast respawn, defending the intel near the bridge)!
and that's an old game which most will tell you run at 300FPS, sure, but not always...


your CPU choice is relevant for gaming, be it old or new games... and the FX9590 gaming performance is not impressive for the money,

the FX9590 should be OK, just not good price/performance compared to i5s for gaming.

for $400 what is interesting is to look at the 9590 as a competitor to the 4770K for rendering/encoding, for gaming the 4770k is always going to be as good (with GPU bottlenecks) or much faster.

 

[galego]

Of course they could also be part of the gaming evolved program, and thus would hardly say the arch rivals CPU is a better choice.

Are you talking about facts or about suppositions? I was talking about technical facts: num. of cores, core freq., IPC, out-of-order execution,...

that is a dangerous line there galego, I am not even sure that is true...any regressions would probably be due to turbo acting up...

There are many threads about Haswell here and some of them contain benchmarks showing regressions. I add two new which I think were not given before:

http://www.pcgameshardware.de/scree...Test-Haswell-4770K-4670K-4570-Crysis-3-v2.png

The 4770k is slower than the 3770k and the 4670k is slower than the 3570k. Here one benchmark where the 4770k is slower than the 3770k

http://openbenchmarking.org/embed.php?i=1306034-UT-INTELCORE31&sha=416bee6&p=2

It is not usual, but it happens.

So the plan is to buy an underperforming, excessively power-consuming, chip now on the vague, and largely unsubstantiated, claim that it will overperform at some time in the future?

There is nothing vague and unsubstantiated. The only reason why Intel chips outperform AMD chips on gaming today is because most current games are designed for few cores. There are lots of benchmarks where a 4-core FX obtains the same FPS than a 8-core FX, because the game ignores half the extra cores. We know that the FX-8350 outperforms a 3770k/4770k when the software uses the 8 cores.

The FX-9590 will provide about a 15% IPS gain, but cannot do magic with current games, which continue to ignore the extra cores.

Future games will be more threaded, because consoles use 8-cores APUs by AMD. Then the FXs will start to show their real gaming potential.

The power-consuming myth is rebutted in the forums often, but always return as a boomerang :whiste:

 

[galego]

The Failure Xtreme 9590 is simply the worst CPU release I have ever seen.

Not even close. But the KitGuru review must be one of the worst that I have seen in recent times.

 

[BallaTheFeared]

The FX-9590 will provide about a 15% IPC gain

Wow that's amazing considering it's exactly the same uarch.

How did AMD pull that off?

 

[SlowSpyder]

DOH!

The seller said he doesn't have an ETA on when he will get these in stock, even though he has them listed as in stock (he said it is an Amazon error).

My AM3+ board and DDR3 shipped already from Newegg. So I have to figure if I want to keep those parts or RMA.

Not sure whether to go for a cheaper and generally faster Intel set up or the much more desirable (to me) AMD set up.

 

[BallaTheFeared]

Not sure whether to go for a cheaper and generally faster Intel set up or the much more desirable (to me) AMD set up.

$80 off i5, $100 off i7 on select boards at Newegg.

http://promotions.newegg.com/combo/13-2990/index.html

Though it's really only awesome if you were already looking at $200 tier mobos in the first place

 

[Essence_of_War]

The FX-9590 will provide about a 15% IPC gain

It's PD/Vishera, right?

Same arch as 8350, right?

Where in the Sam Hill are you getting a 15% IPC gain from?

 

[SlowSpyder]

$80 off i5, $100 off i7 on select boards at Newegg.

http://promotions.newegg.com/combo/13-2990/index.html

Though it's really only awesome if you were already looking at $200 tier mobos in the first place

Well, I wasn't... but I was looking at a $165 motherboard and a $400 CPU. The i5 especially would leave me with plenty left over to get a pretty good motherboard. :awe:

I was kind of considering the i7 3820 the other day, too.

 

[galego]

Wow that's amazing considering it's exactly the same uarch.

How did AMD pull that off?

It's PD/Vishera, right?

Same arch as 8350, right?

Where in the Sam Hill are you getting a 15% IPC gain from?

It was a typo. Corrected.

 

[Idontcare]

There is nothing vague and unsubstantiated. The only reason why Intel chips outperform AMD chips on gaming today is because most current games are designed for few cores. There are lots of benchmarks where a 4-core FX obtains the same FPS than a 8-core FX, because the game ignores half the extra cores. We know that the FX-8350 outperforms a 3770k/4770k when the software uses the 8 cores.

The FX-9590 will provide about a 15% IPS gain, but cannot do magic with current games, which continue to ignore the extra cores.

Future games will be more threaded, because consoles use 8-cores APUs by AMD. Then the FXs will start to show their real gaming potential.

The power-consuming myth is rebutted in the forums often, but always return as a boomerang :whiste:

If the activity factor for the software goes up, as required if the rate of output computations increases, then so too will the power consumption.

This is why less efficient code produces less heat and lower power-consumption.

What you are basically saying is that the reported power-consumption of an FX-9590 is lower than what we can ultimately expect it to be once games and apps are better designed to maximize the activity factor of the FXs.

 

[bononos]

......

The power-consuming myth is rebutted in the forums often, but always return as a boomerang :whiste:

What myth? There is no doubt that the current AMD cpus consume alot more power especially when overclocking. And in the case of the 9590, its a volcano.

 

[SiliconWars]

If the activity factor for the software goes up, as required if the rate of output computations increases, then so too will the power consumption.

This is why less efficient code produces less heat and lower power-consumption.

What you are basically saying is that the reported power-consumption of an FX-9590 is lower than what we can ultimately expect it to be once games and apps are better designed to maximize the activity factor of the FXs.

Well it is true that fully loaded PD cores are very heavy on power consumption. However there are two things we still aren't sure about. First of all, the 8350's turbo might be bringing power consumption up to the same level as all cores loaded, while obviously not giving anything like the same performance. As you know, power consumption and frequency is far from linear scaling.

The second problem is that we very rarely see power benchmarks during gaming - they are all generally given while running Cinebench or x264 because we know that they stress all cores.

It's hard to tell where the 8350's power consumption/efficiency actually is during gaming because of this. Sure it's *bad* but is it actually worse because of the high turbo, and would it be better with all 8 cores loaded? Efficiency wise it would clearly be far, far superior having all 8 cores utilised over just 1 or 2.

 

[bunnyfubbles]

It depends what tests/benchmarks you consider. There are benchmarks where a 3960X barely beats a stock 8350, and others where the 8350 at stock destroy a 3960X. Imagine a 9590...

and for any such instance where that actually happens there are 99 in favor of intel

 

[BallaTheFeared]

It was a typo. Corrected.

Cool beans, next time I'm looking to build a parallel supercomputer I'll give AMD a consideration based on your benchmarks.

 

[monstercameron]

It's PD/Vishera, right?

Same arch as 8350, right?

Where in the Sam Hill are you getting a 15% IPC gain from?

IPS maybe...

 

[USER8000]

It's hard to tell where the 8350's power consumption/efficiency actually is during gaming because of this. Sure it's *bad* but is it actually worse because of the high turbo, and would it be better with all 8 cores loaded? Efficiency wise it would clearly be far, far superior having all 8 cores utilised over just 1 or 2.

With a GTX680 running Batman:

http://img.hexus.net/v2/cpu/amd/Piledriver/FX8748/Power3.png

AFAIK,Batman uses upto 4 threads.

I think this is with an HD7970 IIRC running BF3:

http://www.legitreviews.com/images/reviews/2055/system-power.jpg

With a Geforce Titan running Crysis 3:

http://www.pcgameshardware.de/scree.../Test-Haswell-4770K-4670K-4570-x-Power-v2.png

BF3 and Crysis 3 are well threaded.

I expect for light threaded games,that power consumption for games will have a higher relative drop(due to a lower percentage of available CPU threads being HEAVILY taxed) when compared to say a Core i5.

 

[monstercameron]

With a GTX680 running Batman:

http://img.hexus.net/v2/cpu/amd/Piledriver/FX8748/Power3.png

AFAIK,Batman uses upto 4 threads.

I think this is with an HD7970 IIRC running BF3:

http://www.legitreviews.com/images/reviews/2055/system-power.jpg

With a Geforce Titan running Crysis3:

http://www.pcgameshardware.de/scree.../Test-Haswell-4770K-4670K-4570-x-Power-v2.png

BF3 and Crysis3 are well threaded.

I expect for light threaded games,that power consumption for games will have a higher relative drop(due to a lower percentage of available CPU threads being HEAVILY taxed) when compared to say a Core i5.

where is the bad?

 

[USER8000]

where is the bad?

I was just linking to the reviews I can remember which list gaming power consumption,not indicating whether it was bad or not!!

Depending on the game and how more threads are taxed(3 to 8),it seems the difference is between 22W to 78W more at the wall than a Core i7 3770K,and 42W to 84W more at the wall than a Core i5 3570K. However these are total system power figures at the wall including a taxed graphics card.

In all the cases the total system power consumption was under 295W at the wall anyway with a GTX680,HD7970 and a Geforce Titan.

 

[parablooper]

Well, it performed exactly as well as it should. It's a binned Piledriver, not magic.

Not that AMD-Intel discussions are usually rational, but there's been way too much said over something so mundane.

It was a noble effort to try and cease the war.

But war. War never changes.

 

[monstercameron]

It was a noble effort to try and cease the war.

But war. War never changes.

or the more it changes, the more it stays the same...

 

[frozentundra123456]

Are you talking about facts or about suppositions? I was talking about technical facts: num. of cores, core freq., IPC, out-of-order execution,...



There are many threads about Haswell here and some of them contain benchmarks showing regressions. I add two new which I think were not given before:

http://www.pcgameshardware.de/scree...Test-Haswell-4770K-4670K-4570-Crysis-3-v2.png

The 4770k is slower than the 3770k and the 4670k is slower than the 3570k. Here one benchmark where the 4770k is slower than the 3770k

http://openbenchmarking.org/embed.php?i=1306034-UT-INTELCORE31&sha=416bee6&p=2

It is not usual, but it happens.



There is nothing vague and unsubstantiated. The only reason why Intel chips outperform AMD chips on gaming today is because most current games are designed for few cores. There are lots of benchmarks where a 4-core FX obtains the same FPS than a 8-core FX, because the game ignores half the extra cores. We know that the FX-8350 outperforms a 3770k/4770k when the software uses the 8 cores.

The FX-9590 will provide about a 15% IPS gain, but cannot do magic with current games, which continue to ignore the extra cores.

Future games will be more threaded, because consoles use 8-cores APUs by AMD. Then the FXs will start to show their real gaming potential.

The power-consuming myth is rebutted in the forums often, but always return as a boomerang :whiste:

Not sure even if you understood what I said. Point is, this eurogamer quote you love so much is just speculation by who knows which and how many developers, who will obviously not recommend Intel if they are part of a program sponsored by amd (gaming evolved).

 

[galego]

If the activity factor for the software goes up, as required if the rate of output computations increases, then so too will the power consumption.

This is why less efficient code produces less heat and lower power-consumption.

What you are basically saying is that the reported power-consumption of an FX-9590 is lower than what we can ultimately expect it to be once games and apps are better designed to maximize the activity factor of the FXs.

LOL. This is completely unrelated to anything that I have said in this thread. I did never mention code efficiency, for instance.

What myth? There is no doubt that the current AMD cpus consume alot more power especially when overclocking. And in the case of the 9590, its a volcano.

The myth about the giant delta power consumption of the 8350, which has been debunked in the forums many times. The total power consumption is about 40-80W above a 3770k under full load, as has been shown in many places.

Moreover, many people here claimed that the 9590 was a OC 8350 and that its power consumption would be giant. Another exaggeration. Even this pretty bad review comparing an OC 9590 to a OC 3960x shows that the intel chips consumes much more power. It lacks power consumption figures of the 3960x at stock, but here you find ones

http://www.bit-tech.net/hardware/2012/11/06/amd-fx-8350-review/7

Here, an 8350 at 4.8GHz consumes only 75W more than an 3960x at stock. The delta 9590 vs 3960x (both at stock) will be less than 75W. for other workloads/hardware config. the delta could be of about 130W. Both are insignificant figure for an enthusiast user with one or two high-end dGPU and a 1200W PSU.