Intel X299 expected before year's end

[Lord Nicoll]

10 hours ago, nerdslayer1 said:

really excited hope the new chip is a real improvement rather than a small clock speed boost. 

That's a good question, I'd say it's really hard to say, Skylake from Broadwell was a decent jump in IPC, and clock speeds, but then the enthusiast platform has been rather neglected as of late, and judging by Kaby lake, I wouldn't bet on anymore than Broadwell to Skylake got, but if Zen's halfway decent they might take their thumb out of their ass

 

[DrMikeNZ]

1 hour ago, NumLock21 said:

It make sense going from 9 to 10. Because there is the H110, B150, H170, and Z170.

It would have been better with Z107, not Z170 though. That way 11 could follow 10 with the Z117, instead it is like they were counting ...8,9,10,20.

Having zeros on the end also makes it impossible to have prime numbers.

[Coaxialgamer]

12 hours ago, deXxterlab97 said:

Rip ryzen. You beat broadwell_e but you will not beat kaby lake-e? 

At best skylake x and kaby lake x have a 5% ipc bump over broadwell. 

Plus  4, 6,8 and 10 core chips are already available on the market.  Sure you might get higher clocks,  but looking at how x79 and x99 chips typically clock lower than their mainstream counterparts,  doesn't seem much.  Plus,  broawell e can already clock fairly high. .

I'm not trying to hype ryzen,  but skylake x and kabylake x are hardly anything revolutionary. 

[AnonymousGuy]

I'll probably upgrade my 1650v3 to the v5.  I like that the DDR speeds are bumping up every generation.  2600 should be enough to carry for the next several years.

[suits]

Let's home Ryzen is good competition and it forces intel to give us vastly improved x299 chips as opposed to what kaby lake was. 

 

I'm curious what they'll call it as they already have cpu's in the 7800 and has had an x7900.

[niofalpha]

21 hours ago, dragoon20005 said:

their Kabylake shows how lazy Intel are already

 

so i am not holding my breath on the X299 chipset and CPU to be ground breaking

 

10 cores CPU for 1.8K is bat shit crazy

I don't think it's "Lazy". I think it's them intentionally not trying to make AMD go Bankrupt so they don't get labeled a Monopoly...

[dragoon20005]

20 minutes ago, niofalpha said:

I don't think it's "Lazy". I think it's them intentionally not trying to make AMD go Bankrupt so they don't get labeled a Monopoly...

their CPU performance for each gen is like 5 to 10%

 

Their CPU performance have since stagnant from the Sandy Bridge era

 

Recent CPU benchmarks with games show the i5 2500K still a decent CPU for gaming even with a GTX 1070 installed

[TroubleKlef]

I can't believe X299 is coming out already. Seems like yesterday when I was choosing between z97 and the newly released x99.

[PianoPlayer88Key]

On 1/23/2017 at 10:36 PM, dragoon20005 said:

i do need one 16 core but i not going to pay extreme prices for the Xeon and the server grade board

 

I need to test VMs and study some ransomware virus behavior using a VM

 

My customers have been hit by a few ransomware programs but we lack the knowledge on tackling them

 

And my workload is a mixture of Adobe Creative Suite and Handbrake

You use Adobe CS?  I've been wanting to know something.  How much faster is it at encoding 4K H.264 video than, say, free solutions?  I feel like my 4790K and GTX 1060, or 6700K and GTX 970M, are both dreadfully slow, typically getting in the 5-7 fps range or so with DaVinci Resolve or AviDemux (which I'm not using much anymore).  I'd really like 30fps (realtime for my videos) or better, especially with all intra-frame compression.- Does CS speed things up dramatically like that, or is it a hardware limitation that only dual or quad Broadwell-EX Xeons and SLI Pascal Quadros would solve?   I haven't done enough with Handbrake yet to know much about using it.

 

And speaking of VMs, might that be a way to run old DOS 3.3 software I recently found on an old hard drive, under Windows 10?  (We also have some 5.25" floppies, but Idk if we got rid of our 5.25" drive yet (we still had it several years ago), and my board only supports PCI-E and SATA.  I'm pretty sure we no longer have our 40MB MFM hard drive. )

 

 

Also, for me, 8 vs 16 cores, I think, isn't so much of an issue, compared to only supporting 64GB RAM on the Ryzen non-server platform. I too would run a lot of VMs.  Also, other things I do take up a lot of RAM.  (I wish I could edit full-length uncompressed 4K movies entirely in RAM, but I don't think we're there yet.)

 

 

On 1/23/2017 at 10:55 PM, DrMikeNZ said:

I just have multiple desktops. I have different rendering, encoding and computing tasks running on the systems concurrently. If you need multiple VMs for testing, there is no reason why you need to have them all running on a single system.

 

On 1/23/2017 at 10:58 PM, dragoon20005 said:

budget man...

 

On 1/23/2017 at 11:16 PM, DrMikeNZ said:

For $ of 16C you could get multiple PCs...

 

On 1/23/2017 at 11:22 PM, dragoon20005 said:

which is why i am waiting for the AMD version that could rival the Intel Xeons.

 

Another consideration, at least for me, is physical space taken up by the PCs.

 

I went on pcpartpicker & put a few builds together of various sizes.  Some of them are well beyond what my budget would be for a system like that (the E5-2660 V4 CPU by itself would far exceed my entire budget including peripherals/software if I was shopping now), but they can give an idea how much is packed into a small footprint.

 

https://pcpartpicker.com/list/kwCKkT - i7-7700, mini ITX, Silverstone ML09B case (433.7 in³)

 

https://pcpartpicker.com/list/NWKMm8 - E5-2660 V4, mini ITX, Fractal Design Core 500 case (1136.1 in³)

 

https://pcpartpicker.com/list/xsjhGf - 2x E5-2670, ATX, Fractal Design Node 605 case (1541 in³)

 

I wonder if even smaller cases could be used to fit the components in those parts lists?  Also, if someone could fit an LGA2011 socket (with a 1U-height heatsink) in the new ASRock DeskMini GTX/RX (or use a case small enough so the board takes up the entire inner surface area), that'd be interesting.

 

For now, I was unable to come up with a (relatively) SFF quad-socket build.   The only thing I've come up with is the SuperMicro SuperServer 8017R-7FT+, which I'm sure is not muy barato.

 

OTOH, if I was only building one system, I'd be fine with using a larger case, like a Fractal Design Define XL R2, Rosewill B2 Spirit, etc.

 

 

 

 

17 hours ago, NumLock21 said:

x199 makes no sense

E.g.

Druken master 

Druken master 2

Druken master 3

 

Not

Druken master

Druken master 1

Druken master 2

Druken master 3

 

So Druken master is Druken Master 0? No one names something with 0.

 

X99, X299.

Not X099, X199, X299

 

Yeah.  Also look at the Z series, like someone else touched on.  Z77 > Z87 > Z97 > no, not Z070, Z170 > Z270.  I too wonder why Intel didn't go Z97 > Z107 > Z117, etc.

That reminds me of something related to Nvidia GTX names.  GTX 480 > 580 > 680 > 780 > (skipped 880) 980 > 1080 > ???.  Rumor has it that 2080 would be next, then 3080, not 1180 then 1280.

Personally, I'd prefer the larger the jump in number, the larger the leap in performance.  For example, a GT 2010, imo, should demolish SLI GTX 1080 Tis.  But, I expect that won't happen - I'd expect more like, a GTX 11(?)60 would compete with a GTX 1080.

Actually, in CPUs AND GPUs ... I would love to see large performance gains per generation - for example, new ULV/ULP budget CPU/GPU beats multiple preceeding-generation server/workstation flagship CPU/GPUs in performance.  Or, at minimum, catch up, on CPUs, to where we would be if the ~2x IPC per generation trend from 8086 > 286 had continued to this day.  I'd be okay with waiting a bit longer between generations, like 3-5 years, for that to happen, although an incremental 25-40% bump at the 2 year mark might be nice.

 

[dragoon20005]

37 minutes ago, PianoPlayer88Key said:

You use Adobe CS?  I've been wanting to know something.  How much faster is it at encoding 4K H.264 video than, say, free solutions?  I feel like my 4790K and GTX 1060, or 6700K and GTX 970M, are both dreadfully slow, typically getting in the 5-7 fps range or so with DaVinci Resolve or AviDemux (which I'm not using much anymore).  I'd really like 30fps (realtime for my videos) or better, especially with all intra-frame compression.- Does CS speed things up dramatically like that, or is it a hardware limitation that only dual or quad Broadwell-EX Xeons and SLI Pascal Quadros would solve?   I haven't done enough with Handbrake yet to know much about using it.

 

And speaking of VMs, might that be a way to run old DOS 3.3 software I recently found on an old hard drive, under Windows 10?  (We also have some 5.25" floppies, but Idk if we got rid of our 5.25" drive yet (we still had it several years ago), and my board only supports PCI-E and SATA.  I'm pretty sure we no longer have our 40MB MFM hard drive. )

 

 

Also, for me, 8 vs 16 cores, I think, isn't so much of an issue, compared to only supporting 64GB RAM on the Ryzen non-server platform. I too would run a lot of VMs.  Also, other things I do take up a lot of RAM.  (I wish I could edit full-length uncompressed 4K movies entirely in RAM, but I don't think we're there yet.)

 

 

 

 

 

 

Another consideration, at least for me, is physical space taken up by the PCs.

 

I went on pcpartpicker & put a few builds together of various sizes.  Some of them are well beyond what my budget would be for a system like that (the E5-2660 V4 CPU by itself would far exceed my entire budget including peripherals/software if I was shopping now), but they can give an idea how much is packed into a small footprint.

 

https://pcpartpicker.com/list/kwCKkT - i7-7700, mini ITX, Silverstone ML09B case (433.7 in³)

 

https://pcpartpicker.com/list/NWKMm8 - E5-2660 V4, mini ITX, Fractal Design Core 500 case (1136.1 in³)

 

https://pcpartpicker.com/list/xsjhGf - 2x E5-2670, ATX, Fractal Design Node 605 case (1541 in³)

 

I wonder if even smaller cases could be used to fit the components in those parts lists?  Also, if someone could fit an LGA2011 socket (with a 1U-height heatsink) in the new ASRock DeskMini GTX/RX (or use a case small enough so the board takes up the entire inner surface area), that'd be interesting.

 

For now, I was unable to come up with a (relatively) SFF quad-socket build.   The only thing I've come up with is the SuperMicro SuperServer 8017R-7FT+, which I'm sure is not muy barato.

 

OTOH, if I was only building one system, I'd be fine with using a larger case, like a Fractal Design Define XL R2, Rosewill B2 Spirit, etc.

 

 

 

 

 

Yeah.  Also look at the Z series, like someone else touched on.  Z77 > Z87 > Z97 > no, not Z070, Z170 > Z270.  I too wonder why Intel didn't go Z97 > Z107 > Z117, etc.

That reminds me of something related to Nvidia GTX names.  GTX 480 > 580 > 680 > 780 > (skipped 880) 980 > 1080 > ???.  Rumor has it that 2080 would be next, then 3080, not 1180 then 1280.

Personally, I'd prefer the larger the jump in number, the larger the leap in performance.  For example, a GT 2010, imo, should demolish SLI GTX 1080 Tis.  But, I expect that won't happen - I'd expect more like, a GTX 11(?)60 would compete with a GTX 1080.

Actually, in CPUs AND GPUs ... I would love to see large performance gains per generation - for example, new ULV/ULP budget CPU/GPU beats multiple preceeding-generation server/workstation flagship CPU/GPUs in performance.  Or, at minimum, catch up, on CPUs, to where we would be if the ~2x IPC per generation trend from 8086 > 286 had continued to this day.  I'd be okay with waiting a bit longer between generations, like 3-5 years, for that to happen, although an incremental 25-40% bump at the 2 year mark might be nice.

 

This has been hotly debated but you will be surprised that the bulk of the encoding is still heavily rely on the CPU, RAM and the speed of your scratch disks. The GPU has very little effect to the encoding it can only assist in certain situations but the CPU is the workhorse during encoding.

 

a CPU with the max number of treads is desirable, plus fast RAM with good speed and low CAS latency is a must

 

SSD preferably in a RAID 0 array is a must

 

 

Yea i almost forget about space. There is no way i can fit 5 desktops in my office at one time. SO i can only depend on one very large PC.

 

I used the VMs on a WIndows Server 2012 system and the VMs are running WIndows 7 and 10

 

If Ryzen can support 128GB max, that will be nice

 

64GB will be more than enough since each VM at most uses 4GB to simulate a typical office desktop.

 

My boss can still afford the Xeons but I need to justify the cost. If I were to suggest Ryzen CPUs which can do the same task for lesser price. I wont need a long essay to explain the use case of a multi core desktop.

[cj09beira]

3 hours ago, dragoon20005 said:

This has been hotly debated but you will be surprised that the bulk of the encoding is still heavily rely on the CPU, RAM and the speed of your scratch disks. The GPU has very little effect to the encoding it can only assist in certain situations but the CPU is the workhorse during encoding.

 

a CPU with the max number of treads is desirable, plus fast RAM with good speed and low CAS latency is a must

 

SSD preferably in a RAID 0 array is a must

 

 

Yea i almost forget about space. There is no way i can fit 5 desktops in my office at one time. SO i can only depend on one very large PC.

 

I used the VMs on a WIndows Server 2012 system and the VMs are running WIndows 7 and 10

 

If Ryzen can support 128GB max, that will be nice

 

64GB will be more than enough since each VM at most uses 4GB to simulate a typical office desktop.

 

My boss can still afford the Xeons but I need to justify the cost. If I were to suggest Ryzen CPUs which can do the same task for lesser price. I wont need a long essay to explain the use case of a multi core desktop.

why not by a cheap used 2011 socket server on ebay? they are probably good enough for what you need.

[Coaxialgamer]

13 hours ago, CoolJ S.A.S. said:

But if there's a 4 core on x299, and people more cores adds a lot more heat, so it won't suffer from heating problems on that end at least. The cores could also be a lot bigger too. 

What? A broadwell e core is as fast as a broadwell core,  and skylake x will as fast as skylake per core.  They use the same exact cores,  just more of them. 

And looking at previous enthusiast quad cores like the 4820k and 3820,  it doesn't seem kaby lake x will clock as high as it's desktoo counterpart 

[CoolMarquis97]

Do new generations of CPUs matter? I thought we had reached a point where all we were getting was slightly increased clock speeds for every new generation.

[PianoPlayer88Key]

10 hours ago, dragoon20005 said:

This has been hotly debated but you will be surprised that the bulk of the encoding is still heavily rely on the CPU, RAM and the speed of your scratch disks. The GPU has very little effect to the encoding it can only assist in certain situations but the CPU is the workhorse during encoding.

 

a CPU with the max number of treads is desirable, plus fast RAM with good speed and low CAS latency is a must

 

SSD preferably in a RAID 0 array is a must

 

I wish the GPU helped with encoding.  What would it take, for a GPU to be as much faster than a CPU at video editing, as it is faster than an iGPU (not counting Iris Pro) at gaming?

 

Linus made a video a while back, that shows a concern I also have, about a mega powerful CPU not being THAT much better, in some encoding situations.

 

 

@LinusTech, what was the playtime of the videos you were working with in that clip?  So far I've been unable to figure out which 4K encoding/transcoding tasks, if any, were faster than realtime.

 

You don't think stock-clocked RAM (like DDR3-1600 CAS 9, or DDR4-2133 CAS 15) is good enough?  What if you're using a platform with ECC RAM and can't really clock it any higher?

 

I'm a bit wary of SSDs, or anything, in RAID.  I do like the idea of parity in Raid 5 and 6, to compensate for drive failures, but I don't like what could happen if the controller itself dies.  

 

 

My budget for data recovery is limited to the cost of replacing the failed drive.  And, sometimes it's necessary to save up a while before I'm able to do that.  I've had situations like that, in which I pulled the problem drive out of the PC (or in that case, USB enclosure) and didn't access it until I had an opportunity to try to recover.  In that situation, it was a format / partition resize gone awry.  It happened so fast I thought there was a good chance it didn't actually overwrite the data, but at the time I didn't have a PC with a second hard drive bay to try to do a recovery.  I've since been able to recover much stuff off it, using 2 different recovery methods.  One got a lot more than the other, but the one that got a lot more didn't preserve filenames, directory structure, timestamps, etc, although it did preserve filetypes.

 

 

 

Quote

 

 

Yea i almost forget about space. There is no way i can fit 5 desktops in my office at one time. SO i can only depend on one very large PC.

I hear ya.  And where my PC is right now, putting 5 desktops there would make it a bit crowded.

 

https://goo.gl/photos/tkaJgSgUgWaXZV1B6 - where my PC is now

 

https://goo.gl/photos/nmceRuBTgR8yvMyL7 - where I'd put my case if I could, at least until I move and hopefully have room for an actual computer desk and a better case with good WC support (280-360mm rads), 3.5" HDD support (8-12 or more, plus another 10 or more in hot-swap cages), etc.

 

 

Quote

 

I used the VMs on a WIndows Server 2012 system and the VMs are running WIndows 7 and 10

 

If Ryzen can support 128GB max, that will be nice

128GB would be my bare minimum.  I don't need that now, but I'm quite sure I'll need more before I'm ready to let someone pry my motherboard out of my cold dead hands.

Only reason I'm even remotely considering replacing my 4790K and ASRock Z97 Extreme 6 this year, instead of, say, 2022 (when my AX760 warranty expires), is the possiblity of Ryzen being a significant bump in performance for a reasonable price.  (That is, if the $150/250/350/500 CPU price tier rumors hold true.)  That, and the possibility of being able to upgrade CPUs a few years later (at least until DDR5 comes out) without having to change the board.  (In that case, I'd anticipate upgrading to the last SR7 generation to be supported on AM4, then skip a generation until I upgrade again, to AM6 or whatever they call it, with DDR6 and PCI-E 5.

 

Otherwise, I don't like replacing a board, unless I can get something that's at least 4-5x faster at the same price I'd paid for the outgoing system, or after 5, 7, or even 10 years if possible.

 

My last system supported 4 GB RAM, although only 3 GB was usable due to 32-bit Windows XP limits.  My current desktop supports 32 GB.  Going up by the same factor on the next upgrade, I'd really like Ryzen to support 256 GB, or even 512 GB.  (4*8=32, 32*8=256.)  Then, after that, I'd be looking at ... (256*8=...) 2 TB or more, or, if bought after, say, 2025-2027, maybe 4 TB RAM support.

 

Actually, though, what I want to see even more than support for more RAM ... is software get more efficient, so it doesn't NEED as much RAM, CPU, disk space, etc.   I don't know that GTA 7, Witcher 5, Tomb Raider 2018, Windows 10 build compiled & released in 2018, Adobe CS 2018, etc, or whatever they're called, would run on the Xerox Alto, original Mac, or an 8086-based PC  but having them run well (at max settings) on 10-15-year-old mobile hardware that was ~$100-200 when new would be a dream come true.

 

 

Quote

 

64GB will be more than enough since each VM at most uses 4GB to simulate a typical office desktop.

 

My boss can still afford the Xeons but I need to justify the cost. If I were to suggest Ryzen CPUs which can do the same task for lesser price. I wont need a long essay to explain the use case of a multi core desktop.

https://goo.gl/photos/6B6PSatNiLadDgcN7 - is what can happen when I only have 32GB RAM.

 

For a while I was considering a Xeon system to build a FreeNAS box for backing up my data (meaning making a bit-for-bit clone of all these, plus the few 5TB & other drives that aren't pictured), but 1 - I would only be able to budget maybe $300-500 for the entire system including the storage, and 2 - I think I'll go a different route on backing up.  (Idk yet what I'll do, but I do know it doesn't involve using my internet connection to back up.  Maybe if I had 1+ G-TB/s up and uncapped, I'd consider it.)

 

 

18 hours ago, CoolJ S.A.S. said:

But if there's a 4 core on x299, and people more cores adds a lot more heat, so it won't suffer from heating problems on that end at least. The cores could also be a lot bigger too. 

 

5 hours ago, Coaxialgamer said:

What? A broadwell e core is as fast as a broadwell core,  and skylake x will as fast as skylake per core.  They use the same exact cores,  just more of them. 

And looking at previous enthusiast quad cores like the 4820k and 3820,  it doesn't seem kaby lake x will clock as high as it's desktoo counterpart 

 

I would have figured, that with a quad-core on LGA 2066 ... there'd be more room for heat dissipation, or something, allowing the cores to clock higher.  I was imagining 5 to 5.5 GHz base clocks, for example, maybe OC'ing to 6, 6.5 or even 7 on air, like a 212 Evo or C7, not an NH-D15.  And if that's not quite attainable, I'd at least want them to clock somewhat higher than Kaby Lake i7-7700K at stock, and run cooler at stock. (An example of "run cooler" would be, Kaby Lake-X get the same temps in Prime95 28.7 Small FFT at stock clock, as the i7-7700K gets in Aida64 (not counting FPU test) or Cinebench at stock, assuming both use the same cooler.)

[Coaxialgamer]

50 minutes ago, PianoPlayer88Key said:

 

 

 

I would have figured, that with a quad-core on LGA 2066 ... there'd be more room for heat dissipation, or something, allowing the cores to clock higher.  I was imagining 5 to 5.5 GHz base clocks, for example, maybe OC'ing to 6, 6.5 or even 7 on air, like a 212 Evo or C7, not an NH-D15.  And if that's not quite attainable, I'd at least want them to clock somewhat higher than Kaby Lake i7-7700K at stock, and run cooler at stock. (An example of "run cooler" would be, Kaby Lake-X get the same temps in Prime95 28.7 Small FFT at stock clock, as the i7-7700K gets in Aida64 (not counting FPU test) or Cinebench at stock, assuming both use the same cooler.)

Are you nuts? No amount of extra die space will allow you to dissipate the extreme heat of those kinds of clocks.  Maybe on ln2,  but that's only because the low temperature reduces the power consumption to lower levels.  No silicon chip will ever reach 6ghz on air or water,  let alone on a 212...

Kaby lake x will use only one quad core die,  so it's likely to be quite small,  and similar in size to the existing 7700k.

And even older large chips such as the fx 8350 ( 316mm2) ,  the 2600k (216mm2) and 3820 (416mm2) could reach 5ghz at best despite their size (especially hard on the 3820). 

You're lucky if you hit 7ghz on ln2,  and that's assuming you get a good chip

 

Power just runs away after a certain point.  Higher clocked chips use exponentially more power,  whuch means more heat,  which makes your chip run hotter,  increasing internal resistance,  making problems worse ( this becomes a runaway problem after 5ghz) 

I mean we've been stucknin the low single digit ghz since the pentium 4 in 2000...

[niofalpha]

19 hours ago, dragoon20005 said:

their CPU performance for each gen is like 5 to 10%

 

Their CPU performance have since stagnant from the Sandy Bridge era

 

Recent CPU benchmarks with games show the i5 2500K still a decent CPU for gaming even with a GTX 1070 installed

 Replace Joke/ Reference with "the point".

 

https://en.wikipedia.org/wiki/United_States_antitrust_law

Quote

 These Acts, first, restrict the formation of cartels and prohibit other collusive practices regarded as being in restraint of trade. Second, they restrict the mergers and acquisitions of organizations that could substantially lessen competition. Third, they prohibit the creation of a monopoly and the abuse of monopoly power.

They have next to no competition, they A. Have no incentives to advance further, and B. Can't overlap AMD so much that they go bankrupt, and close. Otherwise Intel would be fractured by the FTC. 

[Misanthrope]

What happened to x199, sacrificed to the God of shitty naming conventions? 

[Trik'Stari]

Excellent.....

finger pyramid of evil contemplation

[TheCherryKing]

I assume that the C622 chipset will arrive around the same time. It is also likely that X299 will use the LGA 3647 Socket.