Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Locked thread
The Lord Bude
May 23, 2007

ASK ME ABOUT MY SHITTY, BOUGIE INTERIOR DECORATING ADVICE

Beautiful Ninja posted:

Haven't a lot, if not all the major motherboard manufacturers released BIOS updates recently that allowed overclocking on the cheaper platforms? I see a lot of combo deals on like Newegg that pair a Pentium Anniversary Edition with an H81 motherboard for example and the reviews indicate that overclocking works fine with that Pentium. My understanding is support is limited to Haswell Refresh and the Pentium in these situations, but you can save even more money without having to buy a Z97 mobo. I usually see the Pentium + mobo paired for 75-80 dollars.

Intel doesn't allow it, but yes some motherboard manufacturers have figured out how to get around that. I tend to view it as a half arsed hack though, and I've always viewed it dubiously.

Lord Windy posted:

I was about to ask "how do you know this will be better" but even in just GHz, the mobo Bude suggested beats the poo poo out of practically everything AMD offers.

Last question, where do I go to learn more about overclocking?

Edit:

I know Ghz != speed, but I think I am safe in thinking that Intel Ghz > AMD Ghz

Pretty much. We have an overclocking thread; I'm sure you can get the information you need there.

Edit: to clarify; that is the cheapest z97 motherboard you can get. It cuts corners on just about everything that isn't related to the overclocking, such as audio, and the quality of the NIC. If you could afford to step it up a notch to something like a z97m-pro4 or an MSI z97m-G43 you probably should, especially since we don't have special deals in Australia that pair one of those pentiums with an Anniversary motherboard for extra cheap.

Second edit: looks like the z97m-pro4 is only $6 more, definitely get it, since it's a major step up. the Anniversary is grossly overpriced in Australia, it needs to be much cheaper than it is in order to make any kind of sense purchasing it.

The Lord Bude fucked around with this message at 15:39 on Sep 4, 2014

Adbot
ADBOT LOVES YOU

The Lord Bude
May 23, 2007

ASK ME ABOUT MY SHITTY, BOUGIE INTERIOR DECORATING ADVICE

Bob Morales posted:



I secretly enjoy seeing the Phenom X6 post higher numbers than any other AMD CPU in the last 4 years.

I used to have one of those :3:

1gnoirents
Jun 28, 2014

hello :)

Lord Windy posted:

I was about to ask "how do you know this will be better" but even in just GHz, the mobo Bude suggested beats the poo poo out of practically everything AMD offers.

Last question, where do I go to learn more about overclocking?

Edit:

I know Ghz != speed, but I think I am safe in thinking that Intel Ghz > AMD Ghz

Yes

I feel funny *requiring* people to learn about overclocking but that does make that bundle a great value. I'd just google it, but the overclocking thread will have the answer in the last few pages.

This is probably the fastest and messiest way to do it but

1. Update bios
2. Set vcore to 1.299
3. Raise cpu multiplier one by one (starting at 40) and test thoroughly on each level with IBT, P95 small fft, and watch temps
4. repeat until crashing or temps too high (85+ for IBT small FFT)

JawnV6
Jul 4, 2004

So hot ...

tijag posted:

ARM's ISA is already RISC so I'm not sure that the two ideas are comparable, but I'm just looking for someone to explain in a way I can understand what exactly nvidia thinks it's gaining by doing this transmetaish implementation of the ARM v8 ISA.

You say "transmetaish" like it's a bad thing? :confused:

Look at it like a worse-is-better to The Right Way of VLIW. It'd be great if compilers were smart enough to schedule arbitrary computation amongst execution units, but it turns out information at runtime makes those decisions more efficient. I can't speak with much more fidelity without diving into details (IO vs OoO) but Transmeta had the right approach and that's where computation should eventually end up.

Bob Morales
Aug 18, 2006


Just wear the fucking mask, Bob

I don't care how many people I probably infected with COVID-19 while refusing to wear a mask, my comfort is far more important than the health and safety of everyone around me!

The Lord Bude posted:

I used to have one of those :3:

I remember when MicroCenter had the 1090T for $199 with a free motherboard, forgot why I didn't buy one.

Civil
Apr 21, 2003

Do you see this? This means "Have a nice day".

Bob Morales posted:



I secretly enjoy seeing the Phenom X6 post higher numbers than any other AMD CPU in the last 4 years.

That's hilarious. I handed down my parents my old 1090t when I moved to intel and they're very happy with it. But honestly, CPUs haven't moved that far ahead terms of performance over the past few years.

http://www.anandtech.com/show/7003/the-haswell-review-intel-core-i74770k-i54560k-tested/6

Sidesaddle Cavalry
Mar 15, 2013

Oh Boy Desert Map
Ladies and gentlemen, it's time to get your futile hopes up again.

http://wccftech.com/breaking-amds-gen-x86-high-performance-core-code-named-zen-debut-k12/

Beware the possibly terrible WCCF interpretation, but the original German stuff from Rory Read is kosher. Early 2016, gotta love wishful planning!

Shooting for a duality between x86 (Zen) and ARM (K12) sounds like a bit of a compromise. Okay, more like a huge compromise. An F-35 Lightning II kind of compromise.

e: oh god also avoid the comment section of that article, help

Sidesaddle Cavalry fucked around with this message at 03:31 on Sep 11, 2014

Rastor
Jun 2, 2001

Pretty sure AMD is talking about making 2 chips, not one chip that does both instruction sets.

Sidesaddle Cavalry
Mar 15, 2013

Oh Boy Desert Map
I think I used the wrong association. I'm simply more concerned about the ambition of managing two different projects to meet two different goals in the same timeframe. Goodness knows they could do better with a single success than with two not-so-successes.

Rastor
Jun 2, 2001

Well it's fine as long as it's two separate design teams. The guy spoke awkwardly in that quote, listing the chips and then listing the chip designers, but I suspect there is a 1:1 designer:chip setup over there. He was just trying to drum up some excitement, while at the same time trying not to trigger the Osborne effect. AMD still has to try to sell their 2015 products before beginning the full push for Zen / K12.

Factory Factory
Mar 19, 2010

This is what
Arcane Velocity was like.
Building six AMD APU systems and want the stock coolers, but don't need all of those manuals? Save up to $3 per system with a six-pack.

I'm sure this makes business sense, somehow, but it seems a bit silly.

Rastor
Jun 2, 2001

It's not about the stock coolers so much as the free games / other promos, which are not bundled when you buy chips by the tray.

Factory Factory
Mar 19, 2010

This is what
Arcane Velocity was like.
AMD keeps slinging the 8-cores. Now it's the FX-8310, 95W TDP, 3.4 GHz, no turbo clock, $125. Multiplier is unlocked, though.

AMD really needs to get off Piledriver and do something new if they want to keep selling in the desktop non-APU space.

Bright side: TSMC announced that they'll fab AMD CPUs at their 16nm FinFET node.

Rastor
Jun 2, 2001

Technically that isn't a TSMC announcement, it's an Italian rumor. I also saw the one that said they will use Samsung's fabs.

Whichever fab they end up choosing, the one thing the various rumor mills agree on is we're not seeing a new architecture until 2016, and not early in the year either.

orange juche
Mar 14, 2012



Rastor posted:

Technically that isn't a TSMC announcement, it's an Italian rumor. I also saw the one that said they will use Samsung's fabs.

Whichever fab they end up choosing, the one thing the various rumor mills agree on is we're not seeing a new architecture until 2016, and not early in the year either.

Oh god, 2 more years without an architectural refresh is really gonna put the hurt on AMD.

Agreed
Dec 30, 2003

The price of meat has just gone up, and your old lady has just gone down

Smaller, cooler-running, modestly more powerful turds? I'll admit that moving to 16nm/14nm FinFET would be great for anyone who isn't there already and wants to be, but you can't really sell slightly better crap when the competition has a rather large performance lead. Theoretical maximum improvements top out too low for AMD's big chips to suddenly leap into being competitive, but anything to make them more so is better than nothing, and working with a new process node is far from being asleep at the wheel.

I acknowledge that they are trying to do the appropriate things. I just think that even with their (in my opinion somewhat lenient) write-offs, their operating revenue isn't where it needs to be for them to fight a battle on all fronts. I'd be really disappointed if they just laid down and died, though; here's hoping they come up with enough added performance and/or power savings to get some Travises on board with it or something.

I know, that's not a great plan, but the heck are they supposed to do instead... The underlying architecture is just balls, and I don't know how they intend to recover it, but I don't think shrinking a bad u-arch is gonna help them as much as they'd like when the competition is shrinking a better u-arch and is doing a great job at almost everything related to designing and pushing chipsets and CPUs.

Twerk from Home
Jan 17, 2009

This avatar brought to you by the 'save our dead gay forums' foundation.

orange juche posted:

Oh god, 2 more years without an architectural refresh is really gonna put the hurt on AMD.

They have enough revenue to continue their existence as a company from the console deal, but they're going to have a hell of a hard time trying to claw their way back into the server market with their next generation uarch.

I don't see them breaking back into the consumer PC market anytime soon, and that's a shrinking market at this point anyway. Intel's getting involved in a race to the bottom with Haswell Celerons and Atoms in the chromebook / tablet market. I'd expect all of AMD's focus on getting Opterons that can undercut Intel's multi-socket Xeons and somehow clawing their way back into servers. The problem is that will take a new architecture that's far more power efficient than their current one.

Rosoboronexport
Jun 14, 2006

Get in the bath, baby!
Ramrod XTreme

Rosoboronexport posted:

HP is bringing out a Chromebook competitor powered by A4 Micro-6400T with $200 price tag. The model in picture is equipped with 14" 1366x768 display, but when they release a model with 9" or 10.1" screen(and hopefully same resolution) with 64 gb storage, I'm going to snag it ASAP. I currently use a EeePC 1015BX which would otherwise be acceptable but it has 1 GB RAM(and the GPU reserves 275 MB of it) so it is painful to use.

And scratch that, HP is only offering AMD on the 14" model, other models feature Intel Bay Trail Celerons or Atoms. Either Intel is throwing Bay Trails to OEMs for almost free or AMD has problems getting the performance or thermals for these things.

I've now ordered the replacement for 1015BX, The Transformer T100 refresh with Atom Z3775. It almost triples available RAM and GPU performance should be about the same and it will run circles around the Ontario CPU.

Rastor
Jun 2, 2001

Twerk from Home posted:

They have enough revenue to continue their existence as a company from the console deal, but they're going to have a hell of a hard time trying to claw their way back into the server market with their next generation uarch.

I don't see them breaking back into the consumer PC market anytime soon, and that's a shrinking market at this point anyway. Intel's getting involved in a race to the bottom with Haswell Celerons and Atoms in the chromebook / tablet market. I'd expect all of AMD's focus on getting Opterons that can undercut Intel's multi-socket Xeons and somehow clawing their way back into servers. The problem is that will take a new architecture that's far more power efficient than their current one.

Hmm. An architecture that's far more power efficient than their current one. You mean like ARM?

orange juche
Mar 14, 2012



Rastor posted:

Hmm. An architecture that's far more power efficient than their current one. You mean like ARM?

Unless something crazy happens when you no longer constrain an an ARM architecture processor within the thermal limits of a tablet, ARM gets walked all over by Intel's Core architecture. Sure ARM is extremely power and thermal efficient, but it also doesn't have much oomph under the hood at comparable thermal limits when compared with Core anything, due to race to sleep and stuff that Intel has been doing with its ultra low power chips.

SwissArmyDruid
Feb 14, 2014

by sebmojo

orange juche posted:

Oh god, 2 more years without an architectural refresh is really gonna put the hurt on AMD.

As if not having products they can sell NOW isn't putting the hurt on them already. I have already replaced one AMD system with an Intel, and will be doing the same with another very shortly in order to resolve a seasonal thermal issue with the one I'm typing on as we speak.

One of these machines will not be updated or replaced again within the next five years. The other remains to be seen.

SwissArmyDruid fucked around with this message at 20:58 on Oct 6, 2014

GrizzlyCow
May 30, 2011
AMD already seem to be readying themselves to release a new non-APU desktop architecture codenamed Zen. As per the WCCFTech link, AMD seems to have transitioned to 20nm or 16nm FinFET already, so maybe their next "high performance" microarchitecture will be decent.

orange juche
Mar 14, 2012



There's still a year and change between now and Zen, which means several Intel iterations and improvements on their tick/tock scaling. I won't hold my breath for AMD to even get close to passing that momentum.


E: VVVV I can't believe I didn't see that when I was typing it.

orange juche fucked around with this message at 00:55 on Oct 7, 2014

SwissArmyDruid
Feb 14, 2014

by sebmojo

orange juche posted:

There's still a year and change between now and Zen

☜(゚ヮ゚☜)

No Gravitas
Jun 12, 2013

by FactsAreUseless

orange juche posted:

There's still a year and change between now and Zen, which means several Intel iterations and improvements on their tick/tock scaling. I won't hold my breath for AMD to even get close to passing that momentum.

Intel has a bit of a mess with their "Broadlake" thing happening... AMD could make some progress till then.

I presume someone realised that Bulldozer is poo poo and started work on fixing the said poo poo a while back, not just now.

Maybe, just maybe this thing has been in the works for a long while now?

Factory Factory
Mar 19, 2010

This is what
Arcane Velocity was like.
IIRC it takes about six years from starting a design to sipping product. But this suggests that they had no idea how bad Bulldozer was until it first missed it's release date.

Lord Windy
Mar 26, 2010

orange juche posted:

Unless something crazy happens when you no longer constrain an an ARM architecture processor within the thermal limits of a tablet, ARM gets walked all over by Intel's Core architecture. Sure ARM is extremely power and thermal efficient, but it also doesn't have much oomph under the hood at comparable thermal limits when compared with Core anything, due to race to sleep and stuff that Intel has been doing with its ultra low power chips.

If ARM is no good, what about going back to POWER or MIPS? I'm legit interested because I would love to see MIPS used more. It's assembly is fun.

orange juche
Mar 14, 2012



Lord Windy posted:

If ARM is no good, what about going back to POWER or MIPS? I'm legit interested because I would love to see MIPS used more. It's assembly is fun.

POWER is extremely potent, magnitudes more powerful than Intel's stuff, but getting good benchmarks on them is pretty difficult because they do things that Intel couldn't begin to deal with. They also are pretty much exclusively used by big iron setups e.g. mainframes/supercomputers/VM farms (1000+ per system), due to them costing massive amounts of money.

ARM is good, it can probably compete with Intel for web server farms, because most webservers don't need a lot of horsepower, and power efficiency equals less money spent on your infrastructure for cooling. Though with VM's being a thing, you could just use a dual CPU top end Xeon and host lots of webservers on VMs, saving space certainly and possibly thermals because you are only cooling one hot server instead of 12 cooler ones.

~Coxy
Dec 9, 2003

R.I.P. Inter-OS Sass - b.2000AD d.2003AD

Rosoboronexport posted:

Either Intel is throwing Bay Trails to OEMs for almost free or AMD has problems getting the performance or thermals for these things.

The former wouldn't surprise me given that there's a Bay Trail Windows 8.1 Chinese tablet now for $150.

Lord Windy
Mar 26, 2010

orange juche posted:

POWER is extremely potent, magnitudes more powerful than Intel's stuff, but getting good benchmarks on them is pretty difficult because they do things that Intel couldn't begin to deal with. They also are pretty much exclusively used by big iron setups e.g. mainframes/supercomputers/VM farms (1000+ per system), due to them costing massive amounts of money.

ARM is good, it can probably compete with Intel for web server farms, because most webservers don't need a lot of horsepower, and power efficiency equals less money spent on your infrastructure for cooling. Though with VM's being a thing, you could just use a dual CPU top end Xeon and host lots of webservers on VMs, saving space certainly and possibly thermals because you are only cooling one hot server instead of 12 cooler ones.

You seem to know a lot about this, what about the other setups like SPARC, Itanium and others which I don't know? Can you speak more on POWER as well because I always just assumed that it was a failure of a system hence why Apple dropped it for Intel.

No Gravitas
Jun 12, 2013

by FactsAreUseless
How does POWER work for single-threaded performance compared to Intel?

Factory Factory
Mar 19, 2010

This is what
Arcane Velocity was like.

Rosoboronexport posted:

And scratch that, HP is only offering AMD on the 14" model, other models feature Intel Bay Trail Celerons or Atoms. Either Intel is throwing Bay Trails to OEMs for almost free or AMD has problems getting the performance or thermals for these things.

Intel's taking a loss on Bay Trail. The SoC itself is priced at cost, and then they'll help you engineer an ARM design into a Bay Trail design and provide extra support chips for free. They're in the "market grab" phase right now trying to catch up to ARM, and they're banking that the investment means less subsidy will be necessary from here on out.

orange juche posted:

Unless something crazy happens when you no longer constrain an an ARM architecture processor within the thermal limits of a tablet, ARM gets walked all over by Intel's Core architecture. Sure ARM is extremely power and thermal efficient, but it also doesn't have much oomph under the hood at comparable thermal limits when compared with Core anything, due to race to sleep and stuff that Intel has been doing with its ultra low power chips.

Most ARM cores today are architected for low power, and Core is very much a high-power core in comparison. Today's ARM cores are much closer to apples-to-apples vs. the current Atom, which is x86 like Core but very ARM-like in terms of power and performance.

I think the biggest ARM-ISA core right now is Apple's Swift, and while it's pretty buff, it's not up to Core in buffness and certainly not up to Core in attempts to scale it up to desktop-type thermals and core counts.

orange juche
Mar 14, 2012



Lord Windy posted:

You seem to know a lot about this, what about the other setups like SPARC, Itanium and others which I don't know? Can you speak more on POWER as well because I always just assumed that it was a failure of a system hence why Apple dropped it for Intel.

I am just armchair quarterbacking this, compiling what I can get out of design docs and spec sheets from benchmarks like specPERF and other synthetic benchmarks along with crawling through a bunch of google searches for info, I'm probably not the best source of information.

I do know that POWER as it exists now is much different from what it was when Apple dropped POWER for Intel. They have since gone into multiple cores per die and multiple chips per board, as well as massive SMT that Intel could only dream of, the POWER8 has a maximum of 12 cores, each capable of working on 8 simultaneous threads for 96 threads per chip, and they usually have 2 processors in a standard 2U chassis.

Granted a single server of that type would cost an insane amount of money and they're not even close to the same price bracket.:rolleyes:

(The fully enabled 24-core Power8 pre-built AIX or Linux servers cost $65,291.00, and they offer custom-build versions that are even more powerful but you would have to contact them for a quote)

orange juche fucked around with this message at 05:31 on Oct 7, 2014

Gwaihir
Dec 8, 2009
Hair Elf

orange juche posted:

I am just armchair quarterbacking this, compiling what I can get out of design docs and spec sheets from benchmarks like specPERF and other synthetic benchmarks along with crawling through a bunch of google searches for info, I'm probably not the best source of information.

I do know that POWER as it exists now is much different from what it was when Apple dropped POWER for Intel. They have since gone into multiple cores per die and multiple chips per board, as well as massive SMT that Intel could only dream of, the POWER8 has a maximum of 12 cores, each capable of working on 8 simultaneous threads for 96 threads per chip, and they usually have 2 processors in a standard 2U chassis.

Granted a single server of that type would cost an insane amount of money and they're not even close to the same price bracket.:rolleyes:

(The fully enabled 24-core Power8 pre-built AIX or Linux servers cost $65,291.00, and they offer custom-build versions that are even more powerful but you would have to contact them for a quote)

We've got a Power770 box at work (running i/OS, not Linux) with (iirc) dual 8 core Power7+ chips and 256 gigs of ram in it that probably ran $700k something, although a shitload of that is certainly in the 48 SSD array it uses for storage :v:

It supports about 1200 simultaneous users doing terminal casework and barely breaks 25% CPU usage.

Benchmarking it without Linux installed is sadly pretty impossible. I'm just as interested in seeing how that hardware really compares, but finding real comparative benches for anything at all recent is an exercise in futility.

I think IBM really really really wants to sell you Power servers running Linux though.

Professor Science
Mar 8, 2006
diplodocus + mortarboard = party
afaik Power is not a meaningful player in high-performance computing (Top500 and the like) anymore, for lots of reasons that nobody probably cares about but me. glancing through Top500, the only new IBM-based machine at all in the past two years in the top 25 is an internal IBM Power8 machine.

also this conversation about ISAs is silly because for various reasons the ISA has been proven to pretty much not matter so long as it's not completely idiotic (eg, x86 is kind of idiotic, x86-64 is not, ARMv8 is not, I don't know enough about MIPS64 to say either way). if you really wanted to build a <1W Power8 CPU with good perf/W, you probably could. if you wanted to build a 300W ARMv8 CPU with big iron style perf/W, you could do that too. the argument people always make is that x86 in particular has huge overhead for instruction decode, but things like Silvermont/Knights Corner/Knights Landing have shown that instruction decode is such a minuscule part of a modern CPU it doesn't matter. the thing that matters once you have equivalent processors with different ISAs: toolchain. x86 dominates (especially for HPC), ARM is closing fast in the not-FORTRAN oriented space, and everything else is significantly further behind.

Menacer
Nov 25, 2000
Failed Sega Accessory Ahoy!

Professor Science posted:

afaik Power is not a meaningful player in high-performance computing (Top500 and the like) anymore, for lots of reasons that nobody probably cares about but me.
Chief among them that IBM's management doesn't want to be in that market anymore because it was a prestige game for a long time, and they weren't making the profit margins they wanted (see: pulling out of Blue Waters, no more Blue Genes on the horizon, etc.)

You'll likely see Power cores back in the HPC lists once Nvidia applies for the next couple of generations of high-end supercomputers. HPC users don't particularly enjoy discrete GPU solutions (as you've made clear in the past). NV is likely banking on OpenPower to supply CPUs for their HPC SoC designs. They're aware that IBM's engineers have produced microarchitectures that perform worlds better than anyone working in the ARM world so far. This isn't insurmountable (you're correct that architecture has very little to do with microarchitecture these days), but the sheer amount of effort that goes into producing a microarchitecture that performs like POWER8, Haswell, or even (though you may not believe it) Steamroller means that NV can't expect their in-house ARM designs to compete at that level in the near term. Not to mention, IBM has put a lot of work into support infrastructure like CAPI that just doesn't exist in the ARM ecosystem yet.

One Eye Open
Sep 19, 2006
Am I awake?

Lord Windy posted:

You seem to know a lot about this, what about the other setups like SPARC, Itanium and others which I don't know? Can you speak more on POWER as well because I always just assumed that it was a failure of a system hence why Apple dropped it for Intel.

SPARC is used for supercomputers, but what I find interesting is that they released the verilog source to their T1 and T2 processors(the T2 is eight cores with eight threads per core) back when they were owned by Sun (under the GPL), so it's a good educational resource.

Lord Windy
Mar 26, 2010
So what you're saying is better software is more advantageous than the Instruction Set Architecture? That makes sense, a whole bunch of chips were better on paper at the time but it ended up being x86 that won thanks to software.

orange juche
Mar 14, 2012



That and the price point of the hardware and support for said hardware. x86 is bargain bin cheap compared to Power, and if it gets the job done "good enough" then that is what an enterprise user is going to use. Power got pushed out of the server market in a major way around the same time Intel pushed AMD into the backseat, because they didn't take Intel very seriously, and lost all of their base to Intel's rapidly improving Core architecture. Maybe they'll capture it back with Power8, maybe not, it all depends on whether they are willing to build price competitive solutions that can go neck and neck with Xeons. (they won't)

Adbot
ADBOT LOVES YOU

PerrineClostermann
Dec 15, 2012

by FactsAreUseless
So this popped up on my news feed...

Am I correct in assuming this doesn't bode well for AMD?

  • Locked thread