Personal computing discussed

Moderators: renee, Flying Fox, morphine

 
Concupiscence
Gerbil Elite
Topic Author
Posts: 705
Joined: Tue Sep 25, 2012 7:58 am
Location: Dallas area, Texas, USA
Contact:

Is Skylake going to be a legitimately big deal for desktops?

Wed Apr 08, 2015 12:38 pm

I keep flirting with the idea of building a Skylake i5-ish rig to celebrate graduating with my Master's this fall, but outside of Intel's marketing I haven't seen much to definitively support the idea that it's going to be a material advancement over Haswell. What say you, gerbils?
Work: Core i9 7940x, 32 gigs RAM, RX Vega 56 + GTX Titan Xm, Win10 Pro + Xubuntu 18.04
Media: Ryzen 7 1700, 16 gigs RAM, Radeon R9 Fury, Win10 Pro
Tinker: Xeon E5-4640, 32 gigs ECC RAM, GTX 750 Ti, Slackware 14.2

Read me at https://www.wallabyjones.com/
 
Dysthymia
Gerbil
Posts: 49
Joined: Thu May 01, 2008 4:54 pm

Re: Is Skylake going to be a legitimately big deal for deskt

Wed Apr 08, 2015 12:48 pm

My guess is it'll be within 5% of the jump from Ivy Bridge to Haswell. Continued focus on power efficiency and lower TDPs, with only modest performance gains on the desktop.
 
Takeshi7
Gerbil First Class
Posts: 122
Joined: Tue Nov 21, 2006 12:50 am

Re: Is Skylake going to be a legitimately big deal for deskt

Wed Apr 08, 2015 12:52 pm

The main advantage of Skylake that I was excited about was additional PCIe 3.0 lanes.
 
The Egg
Minister of Gerbil Affairs
Posts: 2938
Joined: Sun Apr 06, 2008 4:46 pm

Re: Is Skylake going to be a legitimately big deal for deskt

Wed Apr 08, 2015 1:22 pm

Besides the PCIe/SATA/Thunderbolt enhancements, you've got DDR4, an additional instruction set or two, and some chips sporting an additional 64-128MB eDRAM cache. It probably won't be huge, but I'd expect a little more than what was gained with Haswell.
 
Flatland_Spider
Graphmaster Gerbil
Posts: 1324
Joined: Mon Sep 13, 2004 8:33 pm

Re: Is Skylake going to be a legitimately big deal for deskt

Wed Apr 08, 2015 1:52 pm

If you already have a Haswell proc, I wouldn't worry about Skylake. DDR4 will be nice, but I haven't seen anything that looks incredibly better then Haswell.

Ivy Bridge era proc is a maybe, Sandy Bridge possibly, and anything earlier, or AMD, is at upgrade.
 
sweatshopking
Graphmaster Gerbil
Posts: 1464
Joined: Fri Aug 15, 2008 10:37 am

Re: Is Skylake going to be a legitimately big deal for deskt

Wed Apr 08, 2015 1:53 pm

no.
 
Airmantharp
Emperor Gerbilius I
Posts: 6192
Joined: Fri Oct 15, 2004 10:41 pm

Re: Is Skylake going to be a legitimately big deal for deskt

Wed Apr 08, 2015 4:22 pm

I'm less worried- a 2500k at 4.5GHz is still 'fast enough' to not warrant spending hundreds on an upgrade.

For me, I'd need an application that could truly use the juice- and likely a few more cores- for Skylake to be worthwhile.

(I dabbled with editing video from my cameras, and *that* could use the juice!)
 
chuckula
Gold subscriber
Minister of Gerbil Affairs
Posts: 2109
Joined: Wed Jan 23, 2008 9:18 pm
Location: Probably where I don't belong.

Re: Is Skylake going to be a legitimately big deal for deskt

Wed Apr 08, 2015 4:26 pm

sweatshopking wrote:
no.


YOU STOLE MY POST! THEREFORE I MUST CONTRADICT YOU: yes

A little more seriously: Apparently there are some interesting things going into Skylake. The question is: Will those interesting things translate into your existing applications getting a big speed boost? Or... is this technology that requires the right kind of software to be written in order to actually take advantage of the hardware?
4770K @ 4.7 GHz; 32GB DDR3-2133; Officially RX-560... that's right AMD you shills!; 512GB 840 Pro (2x); Fractal Define XL-R2; NZXT Kraken-X60
--Many thanks to the TR Forum for advice in getting it built.
 
the
Gerbil Elite
Posts: 941
Joined: Tue Jun 29, 2010 2:26 am

Re: Is Skylake going to be a legitimately big deal for deskt

Wed Apr 08, 2015 5:05 pm

There have been some interesting rumors going around that Intel has radically altered its microarchitecture for SkyLake. Much of it is predicting a Westmere -> Sandy Bridge-like jump in per clock performance. Not bad but Intel would have to ship at clock speeds similar to Haswell to make it attractive on the desktop. If there is a reduction and/or low overclocking headroom, it'll be another let down as aggregate performance gains will again be marginal.

First off is AVX 3 which widens the vector unit to 512 bits. There are also a few new scatter/gather instructions. I'd expect a similar jump from SSE to AVX when binaries start making use of it. This is also the same AVX extensions being used in the next Xeon Phi. The only ISA extension SkyLake that Xeon Phi does not is TSX. This opens the door for Intel to try a big.LITTLE like configuration in their mobile SoCs.

I'm also predicting a new on-chip interconnect. The current ring bus isn't scaling that well in the Xeon line as core counts go up. They've already announced that the next Xeon Phi is going to be using a 2D grid topology and it wouldn't surprise me the bigger Core series get it too. If Intel tries to emulate big.LITTLE, a common on chip interconnect is a must. This should also shave off a few miliwatts of power even at low core counts.

4-way SMT has been rumored, though this feature maybe a Core i7 feature only like Hyperthreading is today.

There is also a chance that consumer desktops could see a 6 core chip. Intel's previous indications several years ago were that it'd take awhile for additional cores to be worthwhile on the desktop and we're just now hitting that time frame. Luckily things like DX12 are coming that can genuinely scale with thread count so Intel would have perfect timing here. Though these extra cores could also be in a big.LITTLE like configuration.
Dual Opteron 6376, 96 GB DDR3, Asus KGPE-D16, GTX 970
Mac Pro Dual Xeon E5645, 48 GB DDR3, GTX 770
Core i7 3930K@4.2 Ghz, 32 GB DDR3, GA-X79-UP5-Wifi
Core i7 2600K@4.4 Ghz, 16 GB DDR3, GTX 970, GA-X68XP-UD4
 
HorseIicious
Gerbil First Class
Posts: 128
Joined: Tue Aug 16, 2011 7:52 pm
Location: Florida

Re: Is Skylake going to be a legitimately big deal for deskt

Wed Apr 08, 2015 6:13 pm

I've read in a few places that Prescott to Conroe level jump is within the realm of possibility. If that's the case, then obviously it will be a legitimately big deal. May have to wait until late '16 for unlocked K versions though...

Here is the most recent source I've seen these rumours on:

http://wccftech.com/intels-broadwell-sk ... tt-conroe/
 
sweatshopking
Graphmaster Gerbil
Posts: 1464
Joined: Fri Aug 15, 2008 10:37 am

Re: Is Skylake going to be a legitimately big deal for deskt

Wed Apr 08, 2015 6:54 pm

chuckula wrote:
sweatshopking wrote:
no.


YOU STOLE MY POST! THEREFORE I MUST CONTRADICT YOU: yes

A little more seriously: Apparently there are some interesting things going into Skylake. The question is: Will those interesting things translate into your existing applications getting a big speed boost? Or... is this technology that requires the right kind of software to be written in order to actually take advantage of the hardware?

stop pretending. you know i had the right answer.
 
UnfriendlyFire
Gerbil Team Leader
Posts: 285
Joined: Sat Aug 03, 2013 7:28 am

Re: Is Skylake going to be a legitimately big deal for deskt

Wed Apr 08, 2015 7:09 pm

Since Skylake is coming out later this year, I'd say wait for the reviews.

What's your current computer? If you're using Sandy Bridge or above, I'd say just get a SSD if you don't have one already.
 
NovusBogus
Graphmaster Gerbil
Posts: 1408
Joined: Sun Jan 06, 2013 12:37 am

Re: Is Skylake going to be a legitimately big deal for deskt

Wed Apr 08, 2015 10:47 pm

I doubt it will mean much to someone who already has a Haswell i5. DDR4 will be a nice feature, but probably not that big of a deal as memory speed hasn't been a bottleneck for quite some time.

I have a Sandy Bridge 2500K and have been targeting Skylake for a major overhaul, but at this point I'm not sure if the gain will be worth it.
 
Kougar
Minister of Gerbil Affairs
Posts: 2305
Joined: Tue Dec 02, 2008 2:12 am
Location: Texas

Re: Is Skylake going to be a legitimately big deal for deskt

Thu Apr 09, 2015 1:54 am

HorseIicious wrote:
I've read in a few places that Prescott to Conroe level jump is within the realm of possibility. If that's the case, then obviously it will be a legitimately big deal. May have to wait until late '16 for unlocked K versions though...

Here is the most recent source I've seen these rumours on:

http://wccftech.com/intels-broadwell-sk ... tt-conroe/


I wish, but I find that hard to believe without some real evidence behind it. The removal of the FIVR is probably a good thing for enthusiasts and helps explain the socket change if so.

Skylake is supposed to offer AVX-512 (doubling Haswell's AVX-256) capabilities, but that's something software will have to be re-compiled to take advantage of so it won't matter on launch day. Beyond that I'm not aware of anything groundbreaking... keep in mind Conroe was a very poorly kept secret even a full year in advance. Enthusiasts working at various tech jobs were so excited that OEM samples were leaking lots of results, and it was pretty clear Conroe was going to be big early on. If Skylake was even close to a "Conroe"-sized improvement I would expect more indications of it. :(
 
Concupiscence
Gerbil Elite
Topic Author
Posts: 705
Joined: Tue Sep 25, 2012 7:58 am
Location: Dallas area, Texas, USA
Contact:

Re: Is Skylake going to be a legitimately big deal for deskt

Thu Apr 09, 2015 9:07 am

UnfriendlyFire wrote:
Since Skylake is coming out later this year, I'd say wait for the reviews.

What's your current computer? If you're using Sandy Bridge or above, I'd say just get a SSD if you don't have one already.


Current workhorse / workstation is an FX-8320 with 16 gigs of RAM and a GTX 750 Ti. It's well-suited to my workloads and stable (and has a decent SSD with a 2 terabyte drive for /home), but I am a little tired of individual threads running at speeds slightly under what a six year old i5 could muster... Whenever I get around to the new build it'll still be pretty nicely overqualified as a server or computational node.

Kougar wrote:
HorseIicious wrote:
I've read in a few places that Prescott to Conroe level jump is within the realm of possibility. If that's the case, then obviously it will be a legitimately big deal. May have to wait until late '16 for unlocked K versions though...

Here is the most recent source I've seen these rumours on:

http://wccftech.com/intels-broadwell-sk ... tt-conroe/


I wish, but I find that hard to believe without some real evidence behind it. The removal of the FIVR is probably a good thing for enthusiasts and helps explain the socket change if so.

Skylake is supposed to offer AVX-512 (doubling Haswell's AVX-256) capabilities, but that's something software will have to be re-compiled to take advantage of so it won't matter on launch day. Beyond that I'm not aware of anything groundbreaking... keep in mind Conroe was a very poorly kept secret even a full year in advance. Enthusiasts working at various tech jobs were so excited that OEM samples were leaking lots of results, and it was pretty clear Conroe was going to be big early on. If Skylake was even close to a "Conroe"-sized improvement I would expect more indications of it. :(


I'll be a little surprised if it's the kind of per-clock jump that happened between Nehalem and Sandy Bridge, though clock speeds might nudge up a little without an integrated VRM's thermal contribution. I'm looking forward to it, but will vigorously kick myself if this winds up as another installment in Intel's ongoing No Big Deal on the desktop...
Work: Core i9 7940x, 32 gigs RAM, RX Vega 56 + GTX Titan Xm, Win10 Pro + Xubuntu 18.04
Media: Ryzen 7 1700, 16 gigs RAM, Radeon R9 Fury, Win10 Pro
Tinker: Xeon E5-4640, 32 gigs ECC RAM, GTX 750 Ti, Slackware 14.2

Read me at https://www.wallabyjones.com/
 
the
Gerbil Elite
Posts: 941
Joined: Tue Jun 29, 2010 2:26 am

Re: Is Skylake going to be a legitimately big deal for deskt

Thu Apr 09, 2015 9:31 am

Kougar wrote:
HorseIicious wrote:
I've read in a few places that Prescott to Conroe level jump is within the realm of possibility. If that's the case, then obviously it will be a legitimately big deal. May have to wait until late '16 for unlocked K versions though...

Here is the most recent source I've seen these rumours on:

http://wccftech.com/intels-broadwell-sk ... tt-conroe/


I wish, but I find that hard to believe without some real evidence behind it. The removal of the FIVR is probably a good thing for enthusiasts and helps explain the socket change if so.

Skylake is supposed to offer AVX-512 (doubling Haswell's AVX-256) capabilities, but that's something software will have to be re-compiled to take advantage of so it won't matter on launch day. Beyond that I'm not aware of anything groundbreaking... keep in mind Conroe was a very poorly kept secret even a full year in advance. Enthusiasts working at various tech jobs were so excited that OEM samples were leaking lots of results, and it was pretty clear Conroe was going to be big early on. If Skylake was even close to a "Conroe"-sized improvement I would expect more indications of it. :(


The thing with Conroe is that Intel wanted some leaks to defend against AMD at the time whose Athlon X2 line up was very strong. Then there was the fact that Conroe used the same socket 775 infrastructure as its Pentium 4 predecessor. Now Intel reigns supreme on the desktop and every other generation of chips brings a new socket. Intel doesn't want leaks as they want consumers to buy their products now. If you're lucky enough to be in a position to get a SkyLake engineering sample, you'd also need a prototype motherboard to mate with it.

Unless there is a break through in extracting raw parallelism from serial code, performance gains will come from a multitude of smaller tweaks that provide a decent aggregate gain. The other factor is that Intel is operating on a self imposed design rule where a 2% performance gain is only allowed to consume 1% more power (net efficiency will increase). Intel is likely sitting on a pile of smaller minor tweaks that they'll never use in the current design evolution due to the power increase that it would impose. Perhaps a genuinely new microarchitecture from the ground up could implement it without the same power penalty.

As for what Intel could do, I know of a couple of ideas but I'm unsure if they'd be permitted by Intel's design rule. Out of order execution was first introduced in the Pentium Pro nearly 20 years ago. While the actual instructions maybe executed out of order, logically they're put back in order afterward to maintain logically serial stream of results. Sun experimented with out of order instruction retirement (OoOR?) where the results may not be put into proper order based upon dependencies. No reason why Intel could not attempt the same feat. One technique Intel introduced with Conroe is micro-op fusion where two simple micro-ops are replaced by a more complex micro-op that performs both functions. Intel could extend the capabilities of this further by fusing separate multiply and add instructions into a single FMA operation. This could save few cycles on instruction execution and save precious space in the micro-op cache. Intel currently uses a small micro-op cache to enable the instruction decoders to enter a sleep state for a few moments. Intel could simply make the entire L1 instruction cache contain decoded micro-ops and drop the L0 micro-op cache. This wouldn't necessarily improve raw performance but it could potentially save more power by keep the decoders in a sleep state for longer. A structure of decoded instructions could also be beneficial for OoOR or more aggressive micro-op fusion. Intel could implement a multi-pass OoOE algorithm that looks through the instruction/decode caches to find more optimizations. Current OoOE designs are rather simplistic as they are designed to quickly find parallelism in a few cycles just prior to execution. This would be similar to how code morphing works but without the translation of one ISA to another. Being able to go through the process repeatedly to find a more optimal arrangement would certainly improve performance but also at the cost of power.
Dual Opteron 6376, 96 GB DDR3, Asus KGPE-D16, GTX 970
Mac Pro Dual Xeon E5645, 48 GB DDR3, GTX 770
Core i7 3930K@4.2 Ghz, 32 GB DDR3, GA-X79-UP5-Wifi
Core i7 2600K@4.4 Ghz, 16 GB DDR3, GTX 970, GA-X68XP-UD4
 
Kougar
Minister of Gerbil Affairs
Posts: 2305
Joined: Tue Dec 02, 2008 2:12 am
Location: Texas

Re: Is Skylake going to be a legitimately big deal for deskt

Sat Apr 11, 2015 2:45 am

the wrote:
The thing with Conroe is that Intel wanted some leaks to defend against AMD at the time whose Athlon X2 line up was very strong. Then there was the fact that Conroe used the same socket 775 infrastructure as its Pentium 4 predecessor. Now Intel reigns supreme on the desktop and every other generation of chips brings a new socket.


I seem to recall Intel providing at least two official "leaks" on its own prior to launch, but Intel has never ever been one to tolerate ES chips running amok. One of them involved inviting certain tech sites over to visit and run benchmarks on a Conroe system while under Intel's watchful eye several months before launch day.

Regarding sockets, that's a natural consequence of moving everything on-die. For all its performance jump Conroe was a very simple design compared to Haswell. It didn't have onboard memory controllers, it didn't integrate PCIe lanes, and it didn't integrate any MCH or ICH functionality. And because motherboards VRMs were built to survive Prescott they were overkill for Conroe chips. The only way to stop sockets from changing as frequently as they do today would be to stop integrating new IO as well as stop expanding on already integrated IO. Those additional PCIe lanes everyone wants (and the chip badly needs) are likely going to require new pinouts, because PCIe is a point-to-point serial bus, which I'm guessing required a new socket just as much as removing the FIVR probably would've anyway.
 
the
Gerbil Elite
Posts: 941
Joined: Tue Jun 29, 2010 2:26 am

Re: Is Skylake going to be a legitimately big deal for deskt

Tue Apr 14, 2015 9:01 am

Kougar wrote:
the wrote:
The thing with Conroe is that Intel wanted some leaks to defend against AMD at the time whose Athlon X2 line up was very strong. Then there was the fact that Conroe used the same socket 775 infrastructure as its Pentium 4 predecessor. Now Intel reigns supreme on the desktop and every other generation of chips brings a new socket.


I seem to recall Intel providing at least two official "leaks" on its own prior to launch, but Intel has never ever been one to tolerate ES chips running amok. One of them involved inviting certain tech sites over to visit and run benchmarks on a Conroe system while under Intel's watchful eye several months before launch day.


Yeah, Intel did do permit some testing on pre-release systems but in a rather controlled in environment. Not the best comparison.

I seem to recall Anandtech getting full unfettered access to a Conroe system months before anyone else due to a Taiwan motherboard manufacturer contact. This to me was far more indicative of what real world performance and compatibility would be like. This turned out to be an excellent scoop.

Kougar wrote:
Regarding sockets, that's a natural consequence of moving everything on-die. For all its performance jump Conroe was a very simple design compared to Haswell. It didn't have onboard memory controllers, it didn't integrate PCIe lanes, and it didn't integrate any MCH or ICH functionality. And because motherboards VRMs were built to survive Prescott they were overkill for Conroe chips. The only way to stop sockets from changing as frequently as they do today would be to stop integrating new IO as well as stop expanding on already integrated IO. Those additional PCIe lanes everyone wants (and the chip badly needs) are likely going to require new pinouts, because PCIe is a point-to-point serial bus, which I'm guessing required a new socket just as much as removing the FIVR probably would've anyway.


True. Skylake will be moving toward DDR4 and more PCIe lanes, all good things. My beef with Intel is that they've been stuck with the basic dual channel DDR3 and 16 PCIe lanes schema from the CPU socket for the past six years now. OK, during that time frame integrated GPUs took off so Intel could have gotten away with a socket change for Sandybridge but I'm left scratching my head as to why Haswell needed a new socket on the desktop. Sure IVR would have mattered but then why not just change the packaging to ignore the IVR from the motherboard? Or you know put some forethought into socket design to accommodate both platforms easily? I'm pretty sure Intel could have had a common socket for the DDR3 era if they wanted to.
Last edited by the on Tue Apr 21, 2015 1:10 pm, edited 1 time in total.
Dual Opteron 6376, 96 GB DDR3, Asus KGPE-D16, GTX 970
Mac Pro Dual Xeon E5645, 48 GB DDR3, GTX 770
Core i7 3930K@4.2 Ghz, 32 GB DDR3, GA-X79-UP5-Wifi
Core i7 2600K@4.4 Ghz, 16 GB DDR3, GTX 970, GA-X68XP-UD4
 
Concupiscence
Gerbil Elite
Topic Author
Posts: 705
Joined: Tue Sep 25, 2012 7:58 am
Location: Dallas area, Texas, USA
Contact:

Re: Is Skylake going to be a legitimately big deal for deskt

Tue Apr 21, 2015 10:44 am

the wrote:
There have been some interesting rumors going around that Intel has radically altered its microarchitecture for SkyLake. Much of it is predicting a Westmere -> Sandy Bridge-like jump in per clock performance. Not bad but Intel would have to ship at clock speeds similar to Haswell to make it attractive on the desktop. If there is a reduction and/or low overclocking headroom, it'll be another let down as aggregate performance gains will again be marginal.

First off is AVX 3 which widens the vector unit to 512 bits. There are also a few new scatter/gather instructions. I'd expect a similar jump from SSE to AVX when binaries start making use of it. This is also the same AVX extensions being used in the next Xeon Phi. The only ISA extension SkyLake that Xeon Phi does not is TSX. This opens the door for Intel to try a big.LITTLE like configuration in their mobile SoCs.

I'm also predicting a new on-chip interconnect. The current ring bus isn't scaling that well in the Xeon line as core counts go up. They've already announced that the next Xeon Phi is going to be using a 2D grid topology and it wouldn't surprise me the bigger Core series get it too. If Intel tries to emulate big.LITTLE, a common on chip interconnect is a must. This should also shave off a few miliwatts of power even at low core counts.

4-way SMT has been rumored, though this feature maybe a Core i7 feature only like Hyperthreading is today.

There is also a chance that consumer desktops could see a 6 core chip. Intel's previous indications several years ago were that it'd take awhile for additional cores to be worthwhile on the desktop and we're just now hitting that time frame. Luckily things like DX12 are coming that can genuinely scale with thread count so Intel would have perfect timing here. Though these extra cores could also be in a big.LITTLE like configuration.


And now it looks like AVX3 is being reserved solely for Xeon parts. Two steps forward, one step back, over and over again in the name of product segmentation...
Last edited by Concupiscence on Tue Apr 21, 2015 11:59 am, edited 2 times in total.
Work: Core i9 7940x, 32 gigs RAM, RX Vega 56 + GTX Titan Xm, Win10 Pro + Xubuntu 18.04
Media: Ryzen 7 1700, 16 gigs RAM, Radeon R9 Fury, Win10 Pro
Tinker: Xeon E5-4640, 32 gigs ECC RAM, GTX 750 Ti, Slackware 14.2

Read me at https://www.wallabyjones.com/
 
homerdog
Gerbil First Class
Posts: 193
Joined: Wed Jul 09, 2008 9:34 am
Contact:

Re: Is Skylake going to be a legitimately big deal for deskt

Tue Apr 21, 2015 11:42 am

Concupiscence wrote:
And now it looks like AVX3 is being reserved solely for Xeon parts. Two steps forward, one step back, over and over again in the name of product segmentation...

And this is what a world with no AMD looks like :(

I'm still ticked that my 3770K doesn't support basic virtualization features that the i3-2328M in my laptop has. At least Intel came to its senses with Devil's Canyon in that regard.
Antec 300Two + i7-3770K + Gigabyte Z77-D3H + 16GB 1866MHz + GTX970 + SeaSonic S12II 520W + 180GB Intel 330 + 240GB Intel 530
CM Elite 120 + i5-3550 + Gigabyte H77N-WIFI + 16GB 1600MHz + HD7950 + SilverStone ST45SF 450W + 250GB Crucial MX100
 
just brew it!
Administrator
Posts: 53928
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: Is Skylake going to be a legitimately big deal for deskt

Tue Apr 21, 2015 12:06 pm

homerdog wrote:
Concupiscence wrote:
And now it looks like AVX3 is being reserved solely for Xeon parts. Two steps forward, one step back, over and over again in the name of product segmentation...

And this is what a world with no AMD looks like :(

I'm still ticked that my 3770K doesn't support basic virtualization features that the i3-2328M in my laptop has. At least Intel came to its senses with Devil's Canyon in that regard.

Hmm... while I also find Intel's segmentation strategies infuriating, unless you're running on an Intel motherboard I don't think you can blame them for this one. The 3770K supports VT-x; if it is not available on your system it is a limitation of the motherboard/BIOS.
Nostalgia isn't what it used to be.
 
derFunkenstein
Gerbil God
Posts: 25284
Joined: Fri Feb 21, 2003 9:13 pm
Location: Comin' to you directly from the Mothership

Re: Is Skylake going to be a legitimately big deal for deskt

Tue Apr 21, 2015 12:06 pm

Your 3770 supports hardware virtualization. VT-x is fine on my 3570K. It's the virtualized I/O (VT-d) that isn't there, and unless you're trying to run a server full of VMs on your 3770K you're not missing out on anything.
I do not understand what I do. For what I want to do I do not do, but what I hate I do.
Twittering away the day at @TVsBen
 
Flying Fox
Gerbil God
Posts: 25554
Joined: Mon May 24, 2004 2:19 am
Contact:

Re: Is Skylake going to be a legitimately big deal for deskt

Tue Apr 21, 2015 12:13 pm

I have no idea what he was talking about. According to the source of truth the i3-2328M also does not support VT-d. So no idea what "basic virtualization features" he was talking about. :roll:
The Model M is not for the faint of heart. You either like them or hate them.

Gerbils unite! Fold for UnitedGerbilNation, team 2630.
 
just brew it!
Administrator
Posts: 53928
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: Is Skylake going to be a legitimately big deal for deskt

Tue Apr 21, 2015 12:19 pm

Flying Fox wrote:
I have no idea what he was talking about. According to the source of truth the i3-2328M also does not support VT-d. So no idea what "basic virtualization features" he was talking about. :roll:

VT-d is not a "basic" virtualization feature. It allows hardware devices to be passed into the guest OS natively (completely bypassing the host OS's I/O subsystem), and needs support from additional hardware (IOMMU) on the motherboard to function. The one you really care about at the consumer level is VT-x, which is supported by both the i3-2328M and the 3770K. As noted in my previous post, it is possible his motherboard's BIOS does not provide an option to enable it though.
Nostalgia isn't what it used to be.
 
Takeshi7
Gerbil First Class
Posts: 122
Joined: Tue Nov 21, 2006 12:50 am

Re: Is Skylake going to be a legitimately big deal for deskt

Tue Apr 21, 2015 12:43 pm

HorseIicious wrote:
I've read in a few places that Prescott to Conroe level jump is within the realm of possibility./


There will never be as big of a jump as when Conroe and the 8800GT came out. The companies are smarter than to release awesome products that instantly and completely make 100% of their older products obsolete.

Now Intel, AMD, and nVidia release products that are only incrementally better, because it still leaves room in the market for their older products.
 
Concupiscence
Gerbil Elite
Topic Author
Posts: 705
Joined: Tue Sep 25, 2012 7:58 am
Location: Dallas area, Texas, USA
Contact:

Re: Is Skylake going to be a legitimately big deal for deskt

Tue Apr 21, 2015 1:25 pm

Takeshi7 wrote:
HorseIicious wrote:
I've read in a few places that Prescott to Conroe level jump is within the realm of possibility./


There will never be as big of a jump as when Conroe and the 8800GT came out. The companies are smarter than to release awesome products that instantly and completely make 100% of their older products obsolete.

Now Intel, AMD, and nVidia release products that are only incrementally better, because it still leaves room in the market for their older products.


Pfft. Nvidia would love to make another G80, and Intel would love to strike gold with another Conroe. In many ways the biggest impediments to modern computer hardware manufacturer profits are stocks of product that haven't moved yet. Release a new class of parts that's less than 50% faster than your prior flagship, and watch as people snag deals from vendors eager to clear their existing inventory to accommodate new shipments. A quick check around tech forum classified sections shows that prices on used hardware are much lower than new retail prices. There's a lot of stagnation - people don't buy desktops like they used to, head-turning product advances don't emerge every 6 to 8 months, and it's entirely possible to build a new computer and expect it to work in a basic capacity for years at a time.
Work: Core i9 7940x, 32 gigs RAM, RX Vega 56 + GTX Titan Xm, Win10 Pro + Xubuntu 18.04
Media: Ryzen 7 1700, 16 gigs RAM, Radeon R9 Fury, Win10 Pro
Tinker: Xeon E5-4640, 32 gigs ECC RAM, GTX 750 Ti, Slackware 14.2

Read me at https://www.wallabyjones.com/
 
morphine
TR Staff
Posts: 11567
Joined: Fri Dec 27, 2002 8:51 pm
Location: Portugal (that's next to Spain)

Re: Is Skylake going to be a legitimately big deal for deskt

Tue Apr 21, 2015 2:31 pm

Concupiscence wrote:
Intel would love to strike gold with another Conroe.

While I do get your point and rationale, Intel doesn't need another Conroe at this point in time (subject to change). Apart from the pretty low end, the CPU market is pretty much nailed down.
There is a fixed amount of intelligence on the planet, and the population keeps growing :(
 
Concupiscence
Gerbil Elite
Topic Author
Posts: 705
Joined: Tue Sep 25, 2012 7:58 am
Location: Dallas area, Texas, USA
Contact:

Re: Is Skylake going to be a legitimately big deal for deskt

Tue Apr 21, 2015 2:39 pm

morphine wrote:
Concupiscence wrote:
Intel would love to strike gold with another Conroe.

While I do get your point and rationale, Intel doesn't need another Conroe at this point in time (subject to change). Apart from the pretty low end, the CPU market is pretty much nailed down.


You're not wrong. The current scenario is just boring: the foreseeable future is an expanse of ever-lighter disposable mobile devices with steadily increasing efficiency. HPC applications are running into the constraints in a lot of ways, and the answer's not always "just port it to CUDA/OpenCL and run it on a fat GPU." The Xeon Phi is a partial answer to that, but I certainly didn't imagine things would turn out like this as a fledgling CS student. As a geophysicist the state of things doesn't get my juices flowing.
Last edited by Concupiscence on Tue Apr 21, 2015 2:43 pm, edited 2 times in total.
Work: Core i9 7940x, 32 gigs RAM, RX Vega 56 + GTX Titan Xm, Win10 Pro + Xubuntu 18.04
Media: Ryzen 7 1700, 16 gigs RAM, Radeon R9 Fury, Win10 Pro
Tinker: Xeon E5-4640, 32 gigs ECC RAM, GTX 750 Ti, Slackware 14.2

Read me at https://www.wallabyjones.com/
 
morphine
TR Staff
Posts: 11567
Joined: Fri Dec 27, 2002 8:51 pm
Location: Portugal (that's next to Spain)

Re: Is Skylake going to be a legitimately big deal for deskt

Tue Apr 21, 2015 2:41 pm

Yep, and if you look up "product segmentation" in the dictionary, "Intel" comes up.

I'm still angry that ECC support is reserved for the Xeon line. But I can be angry all I want, because they won't have a reason to change that anytime soon.
There is a fixed amount of intelligence on the planet, and the population keeps growing :(
 
SuperSpy
Minister of Gerbil Affairs
Posts: 2402
Joined: Thu Sep 12, 2002 9:34 pm
Location: TR Forums

Re: Is Skylake going to be a legitimately big deal for deskt

Tue Apr 21, 2015 2:54 pm

morphine wrote:
Yep, and if you look up "product segmentation" in the dictionary, "Intel" comes up.

I'm still angry that ECC support is reserved for the Xeon line. But I can be angry all I want, because they won't have a reason to change that anytime soon.

It wouldn't make me so mad if you didn't have to shell out $200 bucks for a low-end Xeon or worse another $200 for a motherboard that will support it. For data storage machines that don't need hardly any CPU power, but _need_ ECC memory, it's super frustrating.

If they made say $50 low-end dual core Xeons, and you could get a board that supported them for under $100, I'd be a lot more willing to pay the Xeon tax.
Desktop: i7-4790K @4.8 GHz | 32 GB | EVGA Gefore 1060 | Windows 10 x64
Laptop: MacBook Pro 2017 2.9GHz | 16 GB | Radeon Pro 560

Who is online

Users browsing this forum: No registered users and 3 guests
GZIP: On