AMD ''Vishera'' FX-Series CPU Specifications Confirmed

[mamailo]

[citation][nom]the3dsgeek[/nom]if only they could release Phenom X6 1100t @4ghz, that was the bomb[/citation]

Unfortunately; all the hand made fine tunning of the stars core don´t allow it (on normal environments). It will be possible in CAD derived Piledriver/Steamroller if AMD last long enuff to make it until the next fab node.CAD has advantages.

It will not unless fix that horrible x86 decoder.Stars cores get 1 more instruction per clock than BD/PD.That alone has enable PII hold their ground Vs FX line.The sad part is that any elementary schooler cant tell you 3 is better than 2, chief engineers really fuck up here.

Meanwhile the only option is fooling around with threading as has been explained above and even then is suboptimal since 2 instructions from same thread can not be decoded.What they where thinking?.
Well; is better than nothing.

 

[tomhuang03]

Nice!

 

[blazorthon]

[citation][nom]mamailo[/nom]Unfortunately; all the hand made fine tunning of the stars core don´t allow it (on normal environments). It will be possible in CAD derived Piledriver/Steamroller if AMD last long enuff to make it until the next fab node.CAD has advantages.It will not unless fix that horrible x86 decoder.Stars cores get 1 more instruction per clock than BD/PD.That alone has enable PII hold their ground Vs FX line.The sad part is that any elementary schooler cant tell you 3 is better than 2, chief engineers really *** up here.Meanwhile the only option is fooling around with threading as has been explained above and even then is suboptimal since 2 instructions from same thread can not be decoded.What they where thinking?.Well; is better than nothing.[/citation]

To be fair, Phenom II has a lot of advantages, such as 50% more decoders per core, but after all of that, it still only has something like 20% higher performance at the same clock frequency. Piledriver seems to have comparable performance per Hz to Phenom II and I think that it has the same front end configuration as Bulldozer. AMD does seem to have done a fairly decent job of knowing just what parts they should cut when they made Bulldozer and improving the Bulldozer design with Piledriver (AM3+) might make it roughly on-par with Phenom II despite the apparent architectural cuts and compromises.

 

[SessouXFX]

I'm not the Geek that many of you are. I'm also not a huge fan of one make over another (unless it's XFX....), although I tend to root for the underdog more times than not. Having said that much, I'm still not expecting AMD to trump Intel at this point in time.

All I want from AMD is for them to find their "Way". Every company has a certain something that makes customers come back to them. I been with AMD for awhile now, after having left Intel due to their dirty tactics against competitors. I respect those who do it the right way and straight forward. No dirt bag, cheap tricks just to get over. Cheating is a good way to tel the customer, we'll do anything for the money. It also tells me, they'll cheat me as well, because what would I know of it anyhow?

So when I read these articles going on about Vishera, a chip that I've been waiting for, after having learned from AMD's past with Phenom, I knew waiting for Vishera was the right thing to do. If I was going to see anything improve after Phenom II, it would come after the 2nd wave of the FX series.

Vishera doesn't have to destroy i3-i7 Sandy or Ivy Bridge procs for me to buy it. However, It MUST be competitive to keep me coming back to AMD. I won't think of it as settling for whatever, they just need to show that they're doing their very best to be relevant in this industry. I know they don't have Intel's never ending war chest backing them. But show me some pride and a solid foundation at least. You don't have to be perfect, but at least be worthy enough to support with my money and I do believe that's what AMD is trying to do as of late.

Now, about that 'driver' they released on the 23rd.... *,..,*

 

[mlscrow]

It's sad to see so many people setting their hearts on Vishera performing well, because the truth is it won't. It will perform slightly better than Bulldozer, but as we all know, Bulldozer performed like doo-doo, so we'll have slightly less smelly doo-doo with Vish. Why do you think AMD has skipped Vishera advertising and went strait to Steamroller advertising? Because they know that Vishera performance will be disappointing and they want people to focus on what will be their biggest performance improvement for the architecture, which is Steamroller. Don't get me wrong, maybe it'll beat the old Phenom II's (maybe?), but it won't touch Sandy Bridge.

 

[Guest]

AMD Wireless Display. No thanks!

 

[bernardblack]

I was hoping for more with their APU's. 8 core options and an integrated 7970 for their flagship version. I'm willing to go completely APU but they need to do better with the integrated GPU. Heck, the PS4 is rumored to be using a 7970 APU from AMD.

 

[bernardblack]

[citation][nom]nebun[/nom]isn't 2012 almost over?[/citation]
[citation][nom]antilycus[/nom]Are you kidding? My FX-8150 CRUSHES any and all intel equivalent when I compare then on Linux. Now since I run everything on Debian Linux (including Windows via VIRT) it still CRUSHES the charts compared to Intel. You don't have to like it, but it's true. The problem is MS can't write stable, multi-threaded kernels. Linux / Debian already do. Up to 256 cores and near perfect memory management. I run 8 Virtual Servers from the FX-8150 and the kernel never has a swap file...EVER and I only have 8gb ram total.[/citation]


This is true, what many people don't realize, is AMD is the most innovative CPU company but nothing is optimized to use it efficiently, including and especially Windows.

 

[blazorthon]

[citation][nom]bernardblack[/nom]I was hoping for more with their APU's. 8 core options and an integrated 7970 for their flagship version. I'm willing to go completely APU but they need to do better with the integrated GPU. Heck, the PS4 is rumored to be using a 7970 APU from AMD.[/citation]

The 7970 die is already huge compared to a CPU die. An APU with it and say 8 cores and the full 8MiB of L3 cache that the FX die has would suck more than 300W of power. That's about the maximum of some of the best coolers and at full GPU+CPU load, this monstrous APU could easily suck 350-400W. Even water cooling would struggle with it. Yield issues would be worse than that of the Big Nvidia GPUs. No offense, but what you ask of AMD is so far beyond reason that it's not even theoretically practical to attempt. That PS4 rumor is BS and even then, it clearly indicated a 7870 or 7950-like APU, which although still not reasonable, is at least theoretically practical.

Besides, even if it could be done practically, it would be incredibly expensive and would need expensive and complex motherboards. The socket would need to be huge to account for the huge memory interface (a whole eight 64 bit memory controllers, six for the GPU's GDDR5 memory and two for the CPU's DDR3 memory, would take up a lot of space) and the display outputs and much more. The motherboards would need the full 3GiB of memory soldered onto the board and placed so as to not be in the way of the huge amount of VRM and the CPU's memory DIMM slots.

[citation][nom]bernardblack[/nom]This is true, what many people don't realize, is AMD is the most innovative CPU company but nothing is optimized to use it efficiently, including and especially Windows.[/citation]

They can be very innovative, but AMD makes far too many stupid mistakes for that to matter. AMD can make design some incredible and differentiated hardware in some ways, in that sense they are innovative, but stupid mistakes such as not making enough chipsets for FM1 motherboards for the amount of Llano APUs that they sell or taking FX and leaving the L3 cache at the same frequencies that it's been for several years, not prioritizing the second core of each module themselves so that they don't have to rely on MS to do their scheduling work for them. There's a setting in PS Check (AMD's P state and such settings modification program for the FX CPUs) to improve prioritization of thread scheduling, why didn't AMD have simply it set it up properly by default?

I can take an FX-81xx CPU and bring the CPU/NB frequency up to 3GHz or so and then fix the priority settings on the cores and that CPU can now easily compete with the Sandy Bridge i7s in performance per core and highly threaded performance. AMD could literally be competing with Intel at any given price point on the LGA 1155 socket platforms if they simply set up their CPUs properly like this.

Then there was their graphics cards. It took AMD about six months after launch to get excellent drivers. The hardware was great in pretty much every way, better than anything that AMD and Nvidia had made in a huge variety of ways and to varying degrees, but there were too many situations where they had driver issues. There were a lot of situations that didn't have issues, but there were simply too many that did have issues.

EDIT: There's also the Enduro issue with AMD's GCN GPU-based Radeon 7000 mobile graphics cards. Even if it gets fixed in October like it's supposed too, it would also have already had about six months to piss off many users of what would otherwise be some of the best mobile graphics cards to date.

 

[bernardblack]

I'm happy with my FX8150 at 4.66ghz and memory at 1960mhz 😉 If I move over to a Crosshair Formula V mainboard and bigger PSU, I should be able to get 5ghz out of it, stable. I worry these new ones will not be so good at overclocking. Looking at their default turbocore speeds (only +200mhz?) and I bet the 8350 still has to shut of 4 cores to do that.

 

[army_ant7]

@blaz
I see what you were talking (in another thread) about what you learned from palladin9479. Hehehe...

@palladin9479
Thanks for sharing the PSCheck trick. I vaguely remember hearing about it a while back elsewhere but didn't dig into it much. Now blaz learned something new, which led to me learning something new. :lol: It's pretty clever! Also thanks for the lesson on how Windows' thread scheduler works. Based on one of your comments, you actually sound like a server hardware specialist.

Anyone who serious thinks Microsoft is good at programming has never seen the C# garbage they invented. I have never seen such an awful, unhelpful programming language in my life. Microsoft will personally employ around 3-7 programming languages to do a single task, there is absolutely no consistency with the programming, it is a hack job at best. Many of them outright despise C++, calling it garbage, which is absolute nonsense.

Can you elaborate here more and give examples?
My brother would probably beg to differ with you since he loves Visual C# (including the XNA and .NET frameworks). He learned how to use them easily. He was able to develop a 3D game (not necessarily a good one, but a functional one with quite an amount of Trigonometry involved) in the span of maybe 6 months (for his college thesis and this happened along with other school subjects and his on-the-job training (He had to cram a lot of stuff to graduate that last year of his.)).

I, personally, would want to stick with C++ and OpenGL, but still, I'm not sure how you can say that about C# or MS.