AMD CPU speculation... and expert conjecture

[colinp]

(edit: don't know why it says drinvis above, as it was clearly gamerk I was quoting...)

But that's my point; HDMI was the replacement for DVI/VGA. It was out first. It did everything Displayport does (minus the rarely used Ethernet layer). It has market acceptance. You aren't going to see Displayport make its way into TV's or Receivers, its limited to Computer Monitors, unlike HDMI. HDMI is NOT going away, so the only cost issue is the added on cost of including the Displayport connector and related circuitry.

No, HDMI is not "the" replacement for DVI / VGA. It is a proprietary technology that incurs a licence. VGA, DVI and Displayport are all VESA standards, royalty free. You were wrong on both counts, I know it's hard to admit it, gamerk316.

Displayport is NOT going away. There, I used caps just like you, so it must be true.

By the way, if you look on NewEgg, roughly a third of monitors have HDMI. A similar amount have Displayport. The latest generation of GPUs from both Nvidia and AMD mostly have both. Of course they would, since you can't daisy chain monitors together with HDMI, but you can with Displayport.

 

[8350rocks]

Net statistics for OS share!!! LOL This link was already given by the same guy in the past. His 1% myth link was refuted by a pair of links explaining how they obtain the statistics, why their numbers are incorrect, and which is the linux share in desktops: 8--13%.

I must have missed those links. Would you be able to dig them up for us, please?

No problem! The link to toms 2013 article reporting how today there are two using linux users in the world for each windows user was given a pair of posts ago.

The links debunking the 1% myth for desktop are below:


Their numbers, 8--12%, agree very well with ones that I can measure. I also shared my own statistics. My server reports >10% of my visitors use linux.

http://broadcast.oreilly.com/2010/09/debunking-the-1-myth.html

http://www.thepowerbase.com/2012/01/the-linux-marketshare-myth/

http://en.wikipedia.org/wiki/Linux_adoption

In my line of work, I have to always challenge figures like this to test their validity.

So, the Tom's link is for all platforms, not just for desktop, so we can quickly count that out for your claim about 8-13% for the desktop.

The Caitlyn Moran link... Well it's from 2010 and bases much of its figures on Linux netbook sales, which I'm sure were a factor in 2009. But capturing 6% of new sales isn't the same as being 6% of the whole market. People don't replace their laptops every year. Furthermore, netbooks are not as popular now as they were then - tablets and cheap laptops killed them.

The next link for thepowerbase does nothing to contradict the 1% desktop figure.

The 8-12% figure in your last link originates from Aaron Seigo in a talk he gave in 2009. He's a big figure in Linux/KDE... but his figure not based on anything but gut feel.

None of these figures factor in how many people installed a different operating system on their device (Windows or Linux) or dual boot. So basically it's a bit of a mess to try to estimate market share.

It's also perfectly valid to ask questions of the netmarketshare number, e.g. is there an underlying bias in the numbers? If so, how big is the bias? But it is a hard stat on current usage across 40,000 global websites, unlike any of the other reported figures. If the true Linux base is 100% higher than the figure they report, that would still be less than 3% of market share. It would be a stretch to say that the vast majority of the Linux desktop devices in the world don't access the popular websites in the same way as Windows users.

Ouch!

R9 290x performance: 1473
R9 290x performance doubled: 1473 * 2 = 2946
R9 295X2 performance: 2933

How's that result unexpected?

Well, it seems now confirmed that those Thief and BF4 slides were pure lie

"Up to". Which is exactly what AMD did not that long ago, yet you still hold on to those numbers as the absolute gospel. And yes, I noted the "up to" when i referenced the original article.

Actually, R9-295X2 run at higher clock speed maximum on both GPUs, and uses water cooling to reduce throttling. In all the tests that TH ran on the new card, there was never a single time that the 295X2 did not out pace 2x290X cards. There were a few situations where 780Tis in SLI out ran the card @1440p, but at 4K it killed everything NVidia everytime...usually by a LARGE margin too.

 

[palladin9479]

On a completely off topic tangent, Fujitsu's Sparc X+ CPU's are now out and they look insane.

600mm^2 size at TSMC 28nm
16 cores, each core can do two threads.
Up to four sockets without glue circuitry.
Two DDR3 memory channels per chip
22~24MB of L2 cache per chip
Up to 512GB of memory per CPU
1500 pin socket.
And ECC support for the CPU and I/O bus's

http://www.tomsitpro.com/articles/fujitsu-m10-servers-sparc64-processor,1-1857.html
http://www.fujitsu.com/global/services/computing/server/sparc/products/m10-1/spec/index.html
http://www.fujitsu.com/global/services/computing/server/sparc/products/m10-4/spec/index.html

Interesting thing about the m10-4s is that you can link them together within the appropriate rack system and they will act as a single system while being fully hot plugable. Also the m10-4 has a built in closed loop watercooling system for the CPU and memory / I/O chips.

That is an insane amount of computing power for modeling.

 

[Cazalan]

On a completely off topic tangent, Fujitsu's Sparc X+ CPU's are now out and they look insane.

.

Yeah they seem to have accelerated their roadmap recently. Did you see the one slated for next year? Could be taken as a sign of confidence in TSMC 20nm process.

4.5GHz+, 24core, 96thread

http://www.fujitsu.com/downloads/TC/sc13/fujitsu-m10.pdf

 

[cobra2]

if people are just going to delete my valid comments for no reason... i don't know why i even bothered commenting on this site

 

[palladin9479]

On a completely off topic tangent, Fujitsu's Sparc X+ CPU's are now out and they look insane.

.

Yeah they seem to have accelerated their roadmap recently. Did you see the one slated for next year? Could be taken as a sign of confidence in TSMC 20nm process.

4.5GHz+, 24core, 96thread

http://www.fujitsu.com/downloads/TC/sc13/fujitsu-m10.pdf

They took it one step further. Oracle and Fujitsu have been working on embedding code into the CPU's themselves that accelerates common DB and OWLS instructions. This lets Oracle RDBMS, OWLS and some of their fusion middle-ware run even faster. I work on both x86 and Sparc systems and whenever I have to do anyone on the x86 systems I end up really missing ILOM.

 

[juanrga]

Ouch!

R9 290x performance: 1473
R9 290x performance doubled: 1473 * 2 = 2946
R9 295X2 performance: 2933

How's that result unexpected?

Well, it seems now confirmed that those Thief and BF4 slides were pure lie

"Up to". Which is exactly what AMD did not that long ago, yet you still hold on to those numbers as the absolute gospel. And yes, I noted the "up to" when i referenced the original article.

R9 295X2 != 2 R9 290x because 5>0

The problem on the slide was not about reporting only "up to" numbers, but that the Nvidia numbers were artificially increased with stuff like "our new driver boosts performance by 71%… if you also throw in a second graphics card!".

Check the link again. This is why the modified slide contains new (corrected) "up to" numbers.

 

[jimmysmitty]

if people are just going to delete my valid comments for no reason... i don't know why i even bothered commenting on this site

I did because the Linux vs Windows is way too off topic now. If you have a problem with it, well that is just too bad.

Let us all just agree to disagree. Some like Linux and its customization, others prefer Windows and then some strangely prefer OSX.

Now back to AMD as they seem to be giving us news for once again.

 

[gamerk316]

(edit: don't know why it says drinvis above, as it was clearly gamerk I was quoting...)

But that's my point; HDMI was the replacement for DVI/VGA. It was out first. It did everything Displayport does (minus the rarely used Ethernet layer). It has market acceptance. You aren't going to see Displayport make its way into TV's or Receivers, its limited to Computer Monitors, unlike HDMI. HDMI is NOT going away, so the only cost issue is the added on cost of including the Displayport connector and related circuitry.

No, HDMI is not "the" replacement for DVI / VGA. It is a proprietary technology that incurs a licence. VGA, DVI and Displayport are all VESA standards, royalty free. You were wrong on both counts, I know it's hard to admit it, gamerk316.

Displayport is NOT going away. There, I used caps just like you, so it must be true.

By the way, if you look on NewEgg, roughly a third of monitors have HDMI. A similar amount have Displayport. The latest generation of GPUs from both Nvidia and AMD mostly have both. Of course they would, since you can't daisy chain monitors together with HDMI, but you can with Displayport.

The failure here, is thinking monitors, rather then TV's, drives the market. They don't. And when you look at TVs, its 90%+ HDMI.

 

[colinp]

The failure here, is thinking monitors, rather then TV's, drives the market. They don't. And when you look at TVs, its 90%+ HDMI.

No, the failure here is the assertion that Displayport is intended to replace HDMI. It's not, it's intended to replace (and is replacing) DSUB and DVI.

Modern graphics cards have simply swapped out the DSUB for Displayport. And that costs nothing, since contrary to your incorrect assertion, Displayport is royalty free.

 

[juanrga]

Nvidia being caught inventing about the performance of their new drivers ("our new driver boosts performance by 71%… if you also throw in a second graphics card!") makes sense with news as this:

3D Mark’s Hall of Fame is one of the most prestigious proving grounds for Overclockers all over the world and recently it has been dominated by Nvidia products. However the pros over at OcUK have landed the top spot with AMD’s R9 290X.

http://wccftech.com/r9-290x-reclaims-top-position-top-categories-3dmark-hall-fame/

 

[Embra]

http://www.tomshardware.com/news/amd-freesync-nvidia-g-sync-vesa,26483.html
VESA Accepts AMD FreeSync, the G-Sync Alternative

 

[de5_Roy]

http://www.tomshardware.com/news/amd-freesync-nvidia-g-sync-vesa,26483.html
VESA Accepts AMD FreeSync, the G-Sync Alternative

that's good. never liked gsync-enabled monitors' high prices.

 

[juanrga]

http://www.tomshardware.com/news/amd-freesync-nvidia-g-sync-vesa,26483.html
VESA Accepts AMD FreeSync, the G-Sync Alternative

News given before in this same page

http://www.tomshardware.co.uk/forum/352312-28-steamroller-speculation-expert-conjecture/page-250#13069034

 

[gamerk316]

The failure here, is thinking monitors, rather then TV's, drives the market. They don't. And when you look at TVs, its 90%+ HDMI.

No, the failure here is the assertion that Displayport is intended to replace HDMI. It's not, it's intended to replace (and is replacing) DSUB and DVI.

Modern graphics cards have simply swapped out the DSUB for Displayport. And that costs nothing, since contrary to your incorrect assertion, Displayport is royalty free.

Its NOT free; you forget the cost to add in the circuitry alongside HDMI. And personally? Having a DSUB is preferable in case I never need to hook up to an old, analog display without needing to use a ton of conversion cables to get the job done. No reason to support two digital standards that do the same, exact, thing.

 

[colinp]

Its NOT free; you forget the cost to add in the circuitry alongside HDMI. And personally? Having a DSUB is preferable in case I never need to hook up to an old, analog display without needing to use a ton of conversion cables to get the job done. No reason to support two digital standards that do the same, exact, thing.

Gamerk, there's an old saying that goes, "If a hole is in the wrong place, no amount of digging will put it in the right place."

You said specifically that there was a licence fee associated with Displayport. That was incorrect. Period.

There is a cost associated with Displayport due to the parts, just as there is for all the other ports. Well, duh. It'll be less than $1 per port, I wager.

And if you need Dsub, then you'd better not be in the market for a current generation graphics card, as none of them have it any more. Or alternatively, you could use one of the readily available converter cables.

Really, it's starting to sound like you are trying to argue the moon is made from cheese. We all know how modest you are when you claim to have been proven right about something. Well, on this occasion you are wrong. Just admit it or drop the argument, and then we can move onto discussing the merits of Freesync on the facts rather than falsehoods.

So, to keep things on topic, I can't see how it could be a bad thing that an open, royalty free standard has had an optional feature added to it.

 

[vmN]

Well, the specs of GTX 880 have been released.


20 nm GM204 silicon
7.9 billion transistors
3,200 CUDA cores
200 TMUs
32 ROPs
5.7 TFLOP/s single-precision floating-point throughput
256-bit wide GDDR5 memory interface
4 GB standard memory amount
238 GB/s memory bandwidth
Clock speeds of 900 MHz core, 950 MHz GPU Boost
7.40 GHz memory
230W Minimum Power

TSMC is on the 20nm wagon.
I don't get nVidia sometimes, like when it only have a 256-bit bus.

Atleast lets hope they get it out before broadwell, just to make fun of Intel.

Hm, Excavador surely will be out in 20nm too, gonna be interesting and with AVX2 support we will see great SIMD improvements. But skylake will have AVX512 at that point, so AMD stays one step behind.

 

[Kulasko]

I saw many positive reviewes about the AM1 platform, just as this one in german:
http://www.planet3dnow.de/cms/9323-kabini-fuer-sockel-am1-athlon-5350-im-test/subpage-fazit/

This one also says that the dealers experience a high demand on AM1 platforms, which might be good for AMD, even though it's a low price/low margin platform

 

[jimmysmitty]

Well, the specs of GTX 880 have been released.


20 nm GM204 silicon
7.9 billion transistors
3,200 CUDA cores
200 TMUs
32 ROPs
5.7 TFLOP/s single-precision floating-point throughput
256-bit wide GDDR5 memory interface
4 GB standard memory amount
238 GB/s memory bandwidth
Clock speeds of 900 MHz core, 950 MHz GPU Boost
7.40 GHz memory
230W Minimum Power

TSMC is on the 20nm wagon.
I don't get nVidia sometimes, like when it only have a 256-bit bus.

Atleast lets hope they get it out before broadwell, just to make fun of Intel.

Hm, Excavador surely will be out in 20nm too, gonna be interesting and with AVX2 support we will see great SIMD improvements. But skylake will have AVX512 at that point, so AMD stays one step behind.

Those are rumored specs. I guarantee that it has a 512bit buss.

As for TSMC getting 20nm before Intel gets 14nm, Intel and Micron already have and have had 20nm NAND. Not the same but I don't think Intel worries about that stuff. They will have 14nm out later this year easily.

I saw many positive reviewes about the AM1 platform, just as this one in german:
http://www.planet3dnow.de/cms/9323-kabini-fuer-sockel-am1-athlon-5350-im-test/subpage-fazit/

This one also says that the dealers experience a high demand on AM1 platforms, which might be good for AMD, even though it's a low price/low margin platform

I will have to check but when I was working for a local shop here, they didn't sell as much AMD as they did Intel and if they did it was either the super die hard AMD fans or people who wanted the cheapest possible.

As well, due to the insane price gouging the highest AMD GPU they have kept in stock is the 270 series.

We will see if this becomes high demand or just becomes another niche market like a lot of stuff has become.

 

[Cazalan]

My local Microcenter had the high end one for just $59, but they didn't have any motherboards to go with them yet.

 

[blackkstar]

gamerK man holy ***, how old are you? You whine about every new technology that comes out.

The list of technology gamerK hates

1. Alternative APIs to DirectX
2. CPUs with multiple cores
3. Replacements for DSUB
4. low IPC, high frequency designs
5. Any OS besides Windows
6. Any GPU company besides Nvidia
7. Any CPU company besides Intel
8. Any IDE that isn't Visual Studio
9. Any programming language that isn't directly supported by MS
10. Linux
11. Royalty free, open standards
12. HSA
13. Anything AMD does

Things gamerK loves

1. Propretiary Windows
2. PhysX
3. CUDA
4. G-Sync
5. Visual Studio
6. DirectX 11
7. Single core CPUs
8. Anything Intel does that resembles HSA
9. DSUB
10. Hooking really old monitors up to expensive graphic cards
11. Never admiting he's wrong.
12. Forum sliding

 

[de5_Roy]

^^ gamerk fanboy revealed?

 

[colinp]

Well, G-Sync needs DisplayPort as well, and additional circuits in the form of the G-Sync module in the monitor. It'll take some cognitive dissonance to find an upside to that vs. FreeSync.

 

[jdwii]

FreeSync>G-Sync
Anyone notice how Nvidia always finds way to lock things down and charge a lot of money for them. When i heard Nvidia shield was 200$ i was amazed i thought they would charge 400$ for it.

 

[Lessthannil]

http://www.tomshardware.com/news/amd-freesync-nvidia-g-sync-vesa,26483.html
VESA Accepts AMD FreeSync, the G-Sync Alternative

that's good. never liked gsync-enabled monitors' high prices.

FreeSync wont exactly be free, either.

1. Its an optional standard. It isn't mandatory for all of DisplayPort. Since monitors that come with DisplayPort are a minority already, monitor OEMs can close off the VRR capable ones to "Gaming" monitors that cost more.
2. FreeSync will still need an ASIC designed specfically for that monitor to work. Either AMD or monitor manufacturers will have to fund it and they can and will pass the expense onto you.
3. You are going to need a graphics card with DisplayPort 1.3

Either way, you have to buy new hardware in order to have VRR. At least NVIDIA was up front about the nature of VRR whereas AMD told us that getting FreeSync or whatever would be at little to no cost to the consumer, which just isnt true.