fx-9590 benchmarks surface

So Steamroller for AM3+ will have many things disabled, like GPU and southbrige?

It's two entirely different platforms and sockets, always has been since Bulldozer FX/APU.

Steamroller modules will be coming in two varieties in both APU and FX form. Kaveri, the codename for the APU, will be the first to have the Steamroller modules hopefully released by the end of 2013. Socket is going to be socket FM2+. That platform supports on-die PCI-Express 3.0 due to lack of HyperTransport since Trinity, and on-die Radeon graphics support.

Steamroller-based FX series is coming out supposedly in 2014 but socket is still unknown. We might get it on Socket AM3+ or a new socket. All indications hint at Socket AM3+ still. Who knows, maybe it'll be Socket AM4? It's too early to say. But, that platform only supports PCI-Express 2.0 because of HyperTransport still exists, and no on-die Radeon graphics. And, with the leaked info of the 1070 and 1090FX chipsets, I'm going to assume Steamroller FX will coincide with that release.

But, if you look at it, Steamroller is going to be a rather substantial improvement over Bulldozer and PIledriver modules. MicroOP cache similar to Sandy Bridge and newer Intel processors. A doubling and widening of decoder, integer, and floating point units. Power improvements, hopefully.

By the time Excavator (or whatever its codename is now), it'll have that plus something called "condensed libraries" according to the AnandTech article. The idea is simple to understand: Squeeze more into less space. It'll probably be the first FX series to have an on-die GPU for compute and HSA (heterogenous system architecture).
 
Well, that 9370 looks good. :D I already have an 8350 so I will just wait for steamroller anyways.
 
I'm just finally glad that 1 module = 2 cores in Steam Roller so people will have to post some other 1 liner in AMD threads.
 
Would A.M.D. possibly release newer AM3+ chipsets that could use 32-bit HyperTransport links, not dual 16-bit links such as its current chipsets?
 
Would A.M.D. possibly release newer AM3+ chipsets that could use 32-bit HyperTransport links, not dual 16-bit links such as its current chipsets?

With the same # of pins and retain backward compatibility?
 
With the same # of pins and retain backward compatibility?

Seriously, AMD needs to integrate a new socket, hopefully an LGA socket, believe it or not, I still get cases of bent pins from people trying to 'DIY' an AMD system for tge first time.
 
I believe backward compatibility will have to end when DDR4 comes. Although that may be 2 to 3 years away.
 
I believe backward compatibility will have to end when DDR4 comes. Although that may be 2 to 3 years away.

Yeah, that's going to be the case with AMD. We now know that Haswell-E will be the first DDR4 processor out sometime next summer 2014. For non-enthusiasts, Skylake will be the next to add DDR4 compatibility. DDR4 isn't expected to ramp up production until 2015 with 2014 being first introduced to server processors.

We might see a new socket and DDR4 memory controller on Excavator whenever that is released within the next two years.

My best bet, with all the things going on at AMD, we will more likely see this happen:
  • Kaveri APU released - Late 2013
  • Steamroller FX released - Early to mid-2014 (New chipset-- 1070 and 1090FX)
  • Steamroller APU refresh (similar to Trinity to Richland) - Mid-2014
  • Steamroller FX refresh - Mid- to late-2014
  • Excavator FX/APU released - Late 2014 to mid-2015 (New socket, DDR4, PCI-Express 3.0 or 4.0 (final specs in 2014-2015)
 
I thought the plan was to just keep one platform. So were going to have APU in the future rather then a CPU?

The current APU plan with AMD seems (mainboard wise) more or less like a clusterfuck , FM2+ will be followed by a new version with DDR4 in 2014 , very short life again , after FM1 they said this wouldn't happen..

AMD new chipset on AM3+ is somewhat doubtful they already let everything slip on that platform and it seems not to be something where they are going to put extra effort in.
 
Max clipper over at VR zone has his hands on an fx-9590 and puts it threw its paces.
http://forums.vr-zone.com/hardware-depot/2884721-the-amd-fx-9590-fx-centurion-preview.html

I find this part most interesting

"Looks like the full test rig is already able to run at ~170W on full load off the wall so the official figure of TDP 220W just for the CPU seems simply extreme overkill attention seeking behaviour from AMD! "

Wow, people have access to these CPUs and they AREN'T overclocking them!? This is the only info I want! I know what an 8350 can do at 5.0GHz, I want to see of this chip can go further!
 
Wow, people have access to these CPUs and they AREN'T overclocking them!? This is the only info I want! I know what an 8350 can do at 5.0GHz, I want to see of this chip can go further!

NDA? It seems there is a very strict guideline what is permitted to be shown in both of the reviews that I have seen.
 
Price isn't really that bad when Intel Extreme CPU's are $1000+

No, a 700-800 price is not acceptable as it's not extreme edition performance. Maybe 3 years old extreme edition performance, but not current gen. Max OC vs MAX OC, it looks similar to gulftown encoding performance. I have my gulftown doing a 33% OC - I doubt the 9590 can match that kind of increase considering how far the chip is already being pushed.

That's really not too bad and will be plenty of performance for a lot of people. But anyone crazy enough to spend $800 on a CPU (like me :D) probably already has something not worth sidegrading to the 9590.

Pricing will make or break this product.
 
If AMD is blocking OC performance in their NDA, I don't expect much headroom. 8350s already can reach 5.0ghz on all cores, so reviews showing benchmarks are really just telling us what we already knew. If these chips consistently hit 6.0ghz you think AMD would be preaching it from the mountaintop, begging people to test their claims.
 
If AMD is blocking OC performance in their NDA, I don't expect much headroom. 8350s already can reach 5.0ghz on all cores, so reviews showing benchmarks are really just telling us what we already knew. If these chips consistently hit 6.0ghz you think AMD would be preaching it from the mountaintop, begging people to test their claims.

Why would there be an NDA? I can get the 9370 at TD. 9590 was on there, but its gone now.
 
his power meter seems awful optimistic. Mine shows idle power draw of a stock core i5-3570k around 77watts i realize our systems are different .... but dang.

My MSI 990fx-gd65v2 + 8320 at 4.9ghz would idle around 90? i forget, but i do remember was when i started cinebench 11.5 watts went up by 335 to something like 400 for total system.

I find a total system draw of 170 watts pretty far fetched for 8 cores fully loaded and 1.5V+ core.
 
his power meter seems awful optimistic. Mine shows idle power draw of a stock core i5-3570k around 77watts i realize our systems are different .... but dang.

My MSI 990fx-gd65v2 + 8320 at 4.9ghz would idle around 90? i forget, but i do remember was when i started cinebench 11.5 watts went up by 335 to something like 400 for total system.

I find a total system draw of 170 watts pretty far fetched for 8 cores fully loaded and 1.5V+ core.

I agree, my 8350 at 4.7 at 1.45v gets up to and over 400w at load. Although that's with two LED monitors attached to the same battery backup, but they don't draw too much power. To be fair, that is with a GTX 580 at idle, but a Piledriver overlocked chip isn't exactly power friendly
 
my friends i5-3570k system idles at 77 with a geforce 580gtx 1.5gb, while running 3dmark firestrike it was around 310, which is higher then any of my systems... those 580s drink the juice dont they :)
 
my friends i5-3570k system idles at 77 with a geforce 580gtx 1.5gb, while running 3dmark firestrike it was around 310, which is higher then any of my systems... those 580s drink the juice dont they :)

Yup, my FX-8350 (slightly overclocked) and GTX 580 break 400w at load easy
 
Wow they had to work to get it stable at stock and the highest they could overclock was 5.0GHz. It's almost like these are just FX-8350's that they tag an extra $600 on to pay for all the RMAing people will have to do to get a chip that can actually run at advertised speeds.

I hope the people that plan to buy these realize they're just getting a $200 CPU and the other $600 is just a charitable donation to AMD (that you don't even get a tax write off for).
 
Looks like is a fairly mediocre processor compared to I7s
EXTENSIVE full review here
http://www.kitguru.net/components/cpu/zardon/amd-fx9590-5ghz-review-w-gigabyte-990fxa-ud5/

Did not look that good until the game benchmarks came up at the end. I am not sure what to make of those since the 9590 looked very competitive. I would have liked to see more CPU comparisons in those than just an overclocked i7 3960x to see if these games were even CPU bound (I guess I could google for that).
 
Did not look that good until the game benchmarks came up at the end. I am not sure what to make of those since the 9590 looked very competitive. I would have liked to see more CPU comparisons in those than just an overclocked i7 3960x to see if these games were even CPU bound (I guess I could google for that).
It would appear that the reason that games were only between the i7-3960X and the FX-9590 was because that site put the ASUS GTX780 Direct CU II OC in the FX test box and the only other review they had done with that video card was the actual review of the video card I linked just there and the CPU they used for that test (presumably to preclude the possibility of being CPU bound) was the i7-3960X, so basically they flipped around that GPU review into a CPU review and added the FX-9590. But I thought it was fairly common knowledge in these parts that very few games are close to CPU bound which is why so few actually run intel 2011 based boards unless they have an actual need for some other feature of the platform other than CPU power, such as extra RAM, extra PCIe bandwidth for super high rez/multi monitor displays, or, probably most predominantly, extra e-peen.

Another issue, he was not using a board correct for that CPU.
Wait so you're saying that $170 board he used wasn't correct? I thought you could slap this thing on any old $50 AM3+ board and call it good, that's the whole point of the AMD "platform cost" advantage right?
 
The UD5 series isn't top of the line for Gigabyte and the cpu is not listed as supported. The latest bios listed is from January of this year, before the cpu was officially announced. This probably accounts for the wrong multiplier being used, etc by the board.
The review says the Sabertooth and Crosshair were validated. Why not use one of those for the review so the cpu is supported, working correctly, and the results valid? Seems like this reviewer was in a hurry and only had the UD5 available.
 
Wait so you're saying that $170 board he used wasn't correct? I thought you could slap this thing on any old $50 AM3+ board and call it good, that's the whole point of the AMD "platform cost" advantage right?

It is very well known that overclocked systems function better with 8+2 phase CPU VRM power design for AMD high-TDP CPU.
You can do it on other boards but that would actually not work for the long run.

But by your logic overclockers just buy the cheapest AM3+ and ruin their mainboard just because that is what overclockers do ....
 
It is very well known that overclocked systems function better with 8+2 phase CPU VRM power design for AMD high-TDP CPU.
You can do it on other boards but that would actually not work for the long run.

But by your logic overclockers just buy the cheapest AM3+ and ruin their mainboard just because that is what overclockers do ....
It's not my logic, it's the logic of all the fanboys on this forum who every time someone mentions that the $250 cheaper i7-3930k will dump all over the FX-9590 at stock and take a messy shit all over it when overclocked parrot the old "platform cost" bullshit as if you could use a sub $100 board to compensate for the fact that a decent 2011 board is $200+ (ignoring the fact that that's only a $100-150 difference that doesn't even cover the $250 spread on the CPU cost much less the fact that even if the total system cost were the same the FX-9590 still loses handily at everything), yet here we have a $170 board that people are saying isn't "top end" enough for this CPU, and that you really need an even more expensive board to properly support these CPUs, which starts making these FX-9590 boards in the same price range as 2011 boards, so you have an equally expensive MB and a slower but much more expensive CPU.

This whole FX-9xxx ploy is so retarded that for the same money you could probably build two whole FX-8350 systems and, assuming your workload could run across multiple machines, get a setup that was almost 2x faster than the single FX-9xxx system for the same price.

I don't care if people buy AMD but FFS buy a FX-8350, I find it horrifying that people here are actually falling for this 5GHz crap (esp since that's not even the normally advertised base speed but a turbo speed!)
 
It's not my logic, it's the logic of all the fanboys on this forum who every time someone mentions that the $250 cheaper i7-3930k will dump all over the FX-9590 at stock and take a messy shit all over it when overclocked parrot the old "platform cost" bullshit as if you could use a sub $100 board to compensate for the fact that a decent 2011 board is $200+ (ignoring the fact that that's only a $100-150 difference that doesn't even cover the $250 spread on the CPU cost much less the fact that even if the total system cost were the same the FX-9590 still loses handily at everything), yet here we have a $170 board that people are saying isn't "top end" enough for this CPU, and that you really need an even more expensive board to properly support these CPUs, which starts making these FX-9590 boards in the same price range as 2011 boards, so you have an equally expensive MB and a slower but much more expensive CPU.

This whole FX-9xxx ploy is so retarded that for the same money you could probably build two whole FX-8350 systems and, assuming your workload could run across multiple machines, get a setup that was almost 2x faster than the single FX-9xxx system for the same price.

I don't care if people buy AMD but FFS buy a FX-8350, I find it horrifying that people here are actually falling for this 5GHz crap (esp since that's not even the normally advertised base speed but a turbo speed!)

You need to find a way to be less caustic with your posts. Couldn't make it through to your point... whatever it was... due to the terrible way you choose to express yourself.
 
Not sure what review you guys have been reading but it states
The 990FXA-UD5 has received AMD Validation
and this
We got hold of their 990FXA-UD5 and flashed to the latest bios

Also there was nothing really questionable. Everything was explained easy enough and shows this CPU really isn't offering anything you can't get from a i7 3770k overclocked to 4.4GHz at a 1/3rd of the cost.

If you google i7 4770k vs i7 3960x then the 4770k trumps it gaming wise on almost every game tested. Oh here we go:

http://www.bit-tech.net/hardware/2013/06/01/intel-core-i7-4770k-cpu-review/6
 
Last edited:
It's not my logic, it's the logic of all the fanboys on this forum who every time someone mentions that the $250 cheaper i7-3930k will dump all over the FX-9590 at stock and take a messy shit all over it when overclocked parrot the old "platform cost" bullshit as if you could use a sub $100 board to compensate for the fact that a decent 2011 board is $200+ (ignoring the fact that that's only a $100-150 difference that doesn't even cover the $250 spread on the CPU cost much less the fact that even if the total system cost were the same the FX-9590 still loses handily at everything), yet here we have a $170 board that people are saying isn't "top end" enough for this CPU, and that you really need an even more expensive board to properly support these CPUs, which starts making these FX-9590 boards in the same price range as 2011 boards, so you have an equally expensive MB and a slower but much more expensive CPU.

This whole FX-9xxx ploy is so retarded that for the same money you could probably build two whole FX-8350 systems and, assuming your workload could run across multiple machines, get a setup that was almost 2x faster than the single FX-9xxx system for the same price.

I don't care if people buy AMD but FFS buy a FX-8350, I find it horrifying that people here are actually falling for this 5GHz crap (esp since that's not even the normally advertised base speed but a turbo speed!)

I'm not sure why you're getting so huffy... I haven't seen one person on this board claim this is a good buy. I'm sure people have talked about the FX-8350 and a high range AMD board, but not these two new chips.

Anyhoo, this thread is about the benchmarks, not fanboyism, except for the misguided hate that just spewed here.
 
Not sure what review you guys have been reading but it states and this

Also there was nothing really questionable. Everything was explained easy enough and shows this CPU really isn't offering anything you can't get from a i7 3770k overclocked to 4.4GHz at a 1/3rd of the cost.

If you google i7 4770k vs i7 3960x then the 4770k trumps it gaming wise on almost every game tested. Oh here we go:

http://www.bit-tech.net/hardware/2013/06/01/intel-core-i7-4770k-cpu-review/6

Which rev. of the Gigabyte motherboard, 1.x or 3.x? I see your point but I would have definitely used the top of the line UD7 for a review with that expensive of a chip.
 
Back
Top