What things can Intel and AMD do to make the desktop a better value?

[VirtualLarry]

Is there a solution integrated in a hard drive controller to pair a SSD with a regular HDD to make a hybrid drive kind of deal yet? I remember reading about something that would automatically profile the loading usage of your files and cache the relevant files from your HDD into the SSD which would translate into 99% of the benefits of a SSD with all the cheap capacity of regular HDDs.

I'd like to see that being a standard feature on motherboards one day...

Some mobos with Marvell SATA6G controllers support their "HyperDuo" feature, which does what you want, I think,

 

[cbn]

After reading the following article at Phoronix, I'll request better OPEN GL driver support from AMD in the OP.

http://www.phoronix.com/scan.php?page=article&item=linux_gpus_high514&num=2

While our comparison was limited on the AMD side due to the lack of AMD's cooperation, overall, it seems the high-end NVIDIA GeForce graphics cards continue to smash the AMD Radeon graphics cards under Linux when it comes to performance with the proprietary drivers. The AMD Catalyst Linux driver has been improving since Valve's been pushing Linux gaming, but still the NVIDIA proprietary Linux graphics driver is delivering top-notch OpenGL performance and continues to be the leading recommendation by game developers and gamers alike.

P.S. At some point, I am sure it will be very interesting to see how an affordable APU would compete against either Athlon x4 750K/760K/860K + discrete video card or some Intel processor + discrete video card in a Steam OS or Steam on Ubuntu FPS per system dollar spent scenario.

 

[cbn]

After reading the Anandtech Broadwell preview (specifically the iGPU section---> http://www.anandtech.com/show/8355/intel-broadwell-architecture-preview/3 ), I updated the OP with my seventh request:

"GT2 or greater iGPUs on at least some of Intel Broadwell based Celeron and Pentium desktop processors."

According the Anandtech preview it sounds like the GT2 iGPUs will receive a good boost in performance.

Intel has greatly increased the ratio of L1 cache and samplers relative to EUs. There is now 25% more sampling throughput per EU, with a total increase in sampler throughput (at identical clockspeeds) of 50%. By PC GPU standards increases in the ratio of samplers to EUs is very rare, with most designs decreasing that ratio over the years. The fact that Intel is increasing this ratio is a strong sign that Haswell’s balance may have been suboptimal for modern workloads, lacking enough sampler throughput to keep up with its shaders.

Moving on, along with the sub-slices front end and common slice are also receiving their own improvements. The common slice – responsible for housing the ROPs, rasterizer, and a port for the L3 cache – is receiving some microarchitecture improvements to further increase pixel and Z fill rates. Meanwhile the front end’s geometry units are also being beefed up to increase geometry throughput at that end.

Much like overall CPU performance, Intel isn’t talking about overall GPU performance at this time. Between the 20% increase in shading resources and 50% increase in sampling resources Broadwell’s GPU should deliver some strong performance gains, though it seems unlikely that it will be on the order of a full generational gain (e.g. catching up to Haswell GT3).

P.S. I was glad to find out Intel updated the video out options on the Broadwell iGPU to include display port.

 

[cbn]

So it sounds like the Broadwell GT2 might be pretty good. Maybe not up to the level of a Haswell HD5000, but pretty darn good.

It would be a shame if we had to step up to a Core i3 processor to get that level of iGPU performance.

The next question that crosses my mind is what kind of Linux graphics performance can we expect for Intel broadwell? (re: Steam Linux HTPC, etc)

 

[Enigmoid]

So it sounds like the Broadwell GT2 might be pretty good. Maybe not up to the level of a Haswell HD5000, but pretty darn good.

It would be a shame if we had to step up to a Core i3 processor to get that level of iGPU performance.

The next question that crosses my mind is what kind of Linux graphics performance can we expect for Intel broadwell? (re: Steam Linux HTPC, etc)

I'm guessing it will be pretty good. Note that HD 5000 is pretty weak, substantially weaker than HD 4600 due to TDP limitations. Intel's drivers have shown steady improvements as well and they are on top of features.

 

[Revolution 11]

1.) More options for more affordable primary storage included on the chipset (eMMC 5.0, UFS 2.0, BGA SSD, etc)

2.) More video outputs for processors with iGPU. (Display Port, HDMI and at least two DVI-I would be a big improvement IMO over the current state of affairs: HDMI, DVI-D, VGA)

These suggestions so much. What keeps the desktop from achieving complete home media/theater versatility is the gimped set of I/O options and display outputs. :'(

Not having to mess with converters and less-than-ideal storage would be very nice.

 

[cbn]

Note that HD 5000 is pretty weak, substantially weaker than HD 4600 due to TDP limitations.

I couldn't find the TDP spec for the iGPUs, only the processors they come integrated with. HD 5000 is included on 15 watt Haswell ULV dual core and HD 4600 comes integrated on 84 watt Haswell quad core.

Here are some specs from this source:

HD 5000 has 40 EUs with frequency range of 200 MHz to 1100 MHz.

HD 4600 has 20 EUs with frequency range of 200 MHz to 1350 MHz.

 

[Insert_Nickname]

After reading the Anandtech Broadwell preview (specifically the iGPU section---> http://www.anandtech.com/show/8355/intel-broadwell-architecture-preview/3 ), I updated the OP with my seventh request:

"GT2 or greater iGPUs on at least some of Intel Broadwell based Celeron and Pentium desktop processors."

According the Anandtech preview it sounds like the GT2 iGPUs will receive a good boost in performance.

I'd expect a bit above (desktop) Ivy HD4000 level for Broadwell GT2. Anything to improve the baseline PC graphics capabilities are fine with me.

P.S. I was glad to find out Intel updated the video out options on the Broadwell iGPU to include display port.

They've had that since Sandy Bridge. Finding mainboard with it hasn't been easy though.

On a side note, its good that Pentium/Celerons finally get 4K support, a feature they've been lacking.

 

[witeken]

Broadwell GT2 actually has 24EUs.

 

[Insert_Nickname]

Broadwell GT2 actually has 24EUs.

Misread. You're right, its 3x8 not 2x8 for the GT2. Fixed

 

[cbn]

Yes, with 24EUs the Broadwell GT2 should be a good performer on the desktop.

The question is how much opportunity does Intel see with this iGPU? Are they going to keep it reserved to the relatively expensive Core i3s? Or open it up to the lower SKUs as well?

If Intel can release Broadwell GT2 to certain Celerons and Pentiums I just wonder if they have the opportunity to create another strong category of x86 desktop PC beyond Enthusiast Gaming Tower, AIO and NUC?

The category I am thinking about is the Gaming HTPC/PC gaming console (with Steam OS or an ubuntu based distro running Steam and/or some other future Linux based gaming client). If the hardware Intel (and AMD) releases is affordable and capable enough we should easily be able to build something much cheaper than a PS4 or Xbox One. This while still having a good "FPS to system dollar spent metric" (assuming the other components can be sourced at reasonable enough prices).


P.S. Regarding the performance of Celeron or Pentium in a SteamOS console (or Linux based Gaming HTPC), I have found my own Pentium G3258 more than capable of handling Battlefield 3 on 64 player with the processor overclocked to 4.5 GHz. So assuming an unlocked multiplier is provided on these GT2 Celerons and Pentiums I don't see adequate cpu performance being a problem.

 

[cbn]

Intel's drivers have shown steady improvements as well and they are on top of features.

Assuming Intel does (at some point, hopefully soon) release a low end enthusiast chip for Linux based PC console/Gaming HTPC, I just wonder how its performance in Linux games (Civilization V, Metro Last Light, Witcher 2, Left for Dead 2, etc) would compare to the Windows equivalent?

Would the linux drivers be good enough that a cheap Linux console PC/Linux Gaming HTPC (using an inexpensive unlocked multiplier Intel chip) could beat one of the cheaper Windows Pre-built desktops? In absolute performance? In value?

EDIT: With that mentioned, maybe a good comparison would be an unlocked Broadwell Celeron with GT2 (running Steam Linux) vs. one of the GT2 processors I expect to be found in Windows pre-built desktops, a Broadwell Core i3. Maybe the unlocked Linux Celeron GT2 loses in absolute gaming performance against the Windows Core i3, but still makes a very compelling argument from a value standpoint?

 

[cbn]

If yields on 14nm improve to the point where certain budget LGA chips can have the GT3 iGPU then so much the better. (re: budget gamers would more likely to use a GT3 iGPU than someone buying an expensive quad core desktop chip with GT3 iGPU)

So in an ideal world maybe we could have something like the following:

unlocked LGA Celeron GT2
unlocked LGA Pentium GT3

unlocked LGA Celeron and Pentium GT1 (for people wanting to use a discrete video card)
unlocked BGA Celeron GT1 (a cheap platform for someone wanting really good desktop performance)

 

[Lorne]

thoughts

AMD to drop the module and just go back to big cores.

Both AMD and Intel.
Dynamic HT.
Dynamic PCIe lanes, BIOS set or auto direct by connection. (supporting odd lane count)
How about double sided clamp soc design that feeds the iGPU (or more cores) through to the otherside for another cooler.

 

[TuxDave]

thoughts

AMD to drop the module and just go back to big cores.

Both AMD and Intel.
Dynamic HT.
Dynamic PCIe lanes, BIOS set or auto direct by connection. (supporting odd lane count)
How about double sided clamp soc design that feeds the iGPU (or more cores) through to the otherside for another cooler.

Please elaborate.

 

[cbn]

Regarding my request #6 in the OP: "Better Open GL support from AMD"

I really do hope the following helps out:

http://techreport.com/news/26922/amd-hopes-to-put-a-little-mantle-in-opengl-next

At Siggraph yesterday, we met with Richard Huddy, AMD's newly appointed Gaming Scientist. Huddy gave us some dirt on the incoming wave of FreeSync monitors, but that wasn't all. He also provided an update on Mantle, and he revealed some interesting details about AMD's role in the development of the next-gen OpenGL API.

Announced earlier this week, the next-gen version of OpenGL is a complete, from-the-ground-up rewrite aimed at slashing overhead and giving developers more control over the hardware. In that sense, the upcoming API, which Huddy calls "OpenGL Next," will follow in the footsteps of Mantle and DirectX 12.

Huddy told us AMD has done a "great deal of work" with the Khronos Group, the stewards of the OpenGL spec, on OpenGL Next. AMD has given the organization unfettered access to Mantle and told them, in so many words, "This is how we do it. If you want to take the same approach, go ahead." Khronos is free to take as many pages as it wants out of the Mantle playbook, and AMD will impose no restrictions, nor will it charge any licensing fees.

While Huddy didn't say how closely OpenGL Next might mirror Mantle, he repeated the contention that Mantle shaped DirectX 12's development. We expressed some doubts about that contention when we addressed it earlier this year, but Huddy was adamant. Development on DirectX 12's new features may have begun before Mantle, he said, but the "real impetus" for DX12's high-throughput layer came from the AMD API.

Speaking of which, Huddy told us 75 developers are now working on Mantle titles in the consumer realm. Enthusiasm, he added, "seems unbridled." Some of those developers see Mantle as a stepping stone to DirectX 12, while others view the API as an "opportunity to differentiate themselves."

Huddy expects developers to keep using Mantle after DX12 arrives, too, for two reasons. First, AMD can add support for new GPU features very quickly—much quicker than Microsoft, which rolls out major DirectX updates only every 4-5 years. Second, Mantle has full support for Windows 7, which DX12 may not. This second point is somewhat at odds with what Microsoft hinted at GDC, but we're still awaiting an official statement about DX12 backward-compatibility. The API is expected to ship with the next Windows release, according to Huddy.

Finally, Mantle is heading to the workstation space. AMD hasn't released a public Mantle driver for its FirePro graphics cards yet, but a beta program is underway. A number of software vendors, some of which are "not small," are already on board. Huddy was reluctant to name names, but he sounded very pleased. Enthusiasm about Mantle among workstation software vendors, he said, is "very high." But because things move slower in the workstation space, he doesn't expect the first Mantle-enabled workstation apps to come out until next year.

Maybe if OPEN GL-Next takes enough from Mantle the gap between AMD and Nvidia proprietary Linux graphics drivers will be largely eliminated?

In any event, I would like to see AAA titles on Linux playing at the same frame rate as the Windows version of the same title.

 

[Conker10k]

Make a motherboard with a high end gpu onboard. Make the cases smaller. Make the screens that come with these computers more bigger. 285x on the motherboard. To appeal to the minimum income people make the computers dirt cheap. ITX motherboard with everything onboard but with lots of memory around 8GB+. Make sure hardware acceleration is there and it works or else its all garbage. Maybe a ps3 apu chip inside. Make a computer the size of the Wii. People like small things these days. Thinner, smaller, tiny'er.

 

[TidusZ]

make a processor faster than a 2500k and I'll call that value

Otherwise my ivy bridge will be my main pc til it burns out and I replace it with an ivy bridge

 

[Enigmoid]

Something that would cost intel almost nothing? Activate all the L3 cache on its dies. 8MB for i5 or 4 MB for i3.

 

[MiddleOfTheRoad]

How about an i3 that is unlocked? I still don't feel an i5 is required for most games -- I think a Dual Core with HT is good enough for 90% of what is out there.

 

[DrMrLordX]

How about an i3 that is unlocked? I still don't feel an i5 is required for most games -- I think a Dual Core with HT is good enough for 90% of what is out there.

That is precisely the reason why Intel would be hesitant to release an i3 with unlocked multipliers. It would eat into their i5 sales unless they set the unlocked i3 at an unacceptably high price point.

 

[MongGrel]

Is there a solution integrated in a hard drive controller to pair a SSD with a regular HDD to make a hybrid drive kind of deal yet? I remember reading about something that would automatically profile the loading usage of your files and cache the relevant files from your HDD into the SSD which would translate into 99% of the benefits of a SSD with all the cheap capacity of regular HDDs.

I'd like to see that being a standard feature on motherboards one day...

Some mobos with Marvell SATA6G controllers support their "HyperDuo" feature, which does what you want, I think,

I have one of these I'm using for a couple of Samsung EVO SSD's at the moment in RAID0.

Vantec 4 Channel 6-Port SATA 6 Gb/s PCIe RAID Host Card Model UGT-ST644R with HyperDuo

http://www.newegg.com/Product/Produc...82E16815287028

I had a WD Black 500G in with it for awhile in HyperDuo and it work really well actually, but I all ready have a hardware RAID setup in here with 4X1Tb WD RE3 HDD's, so I moved the Black to another when I built a HTPC.

Magician doesn't work with them that way though, but I don't really miss that myself.

Bought it just to get Sata III on the old P6T7, if works well here.

 

[jpiniero]

Is there a solution integrated in a hard drive controller to pair a SSD with a regular HDD to make a hybrid drive kind of deal yet? I remember reading about something that would automatically profile the loading usage of your files and cache the relevant files from your HDD into the SSD which would translate into 99% of the benefits of a SSD with all the cheap capacity of regular HDDs.

OSX's Fusion does something similar like this, but that's done in software.

BTW, The real upgrade would be non-volatile memory. If you could get that small enough to have a meaningful size on package, that would really kill two birds with one stone (memory and disk), while saving a ton of power.

 

[cbn]

Updated OP with my eight request:

" Make the desktop easier to build for beginning DIYers."

(Not sure how they would go about this, but something like NUC is easy to build. Unfortunately NUC is too low on power IMO. What is needed is a true desktop class processor in an easy to assemble package. This is especially true on the low end or entry level desktop category where new DIYers are looking for maximum value)

 

[poohbear]

Updated OP with my eight request:

" Make the desktop easier to build for beginning DIYers."

(Not sure how they would go about this, but something like NUC is easy to build. Unfortunately NUC is too low on power IMO. What is needed is a true desktop class processor in an easy to assemble package. This is especially true on the low end or entry level desktop category where new DIYers are looking for maximum value)

Well building desktops & even overclocking are easier now than they've ever been! I started making computers back in the AMD Athlon days, things weren't as straight forward as they are now. How much more idiot proof do they need to make them?