X499 or X299G? Intel's Fall 2019 HEDT Update Heralds a New Chipset?

[btarunr]

Our readers spotted an interesting visual detail that missed us during our coverage of GIGABYTE's three new socket LGA2066 motherboards unveiled at Computex 2019. One of the three boards, the X299S Designare 10G, has the hard-marking "X499" on its CPU VRM heatsink. Another detail that strikes us is that none of the three new boards we pictured has "X299" printed on the PCB anywhere. The Designare 10G has a sticker below the printed GIGABYTE logo that reads "X299G Designare 10G." The purported X299G Aorus Master has another interesting detail: right above the "Aorus Master" print, there's a tiny sticker marked "X299G," positioned as if it's covering up a printed marking on the PCB itself.

All these details lead us to wonder if GIGABYTE tried to cover up that these boards are in fact based on the unannounced X499 Express chipset, and made to appear like they are X299. We only have paper stickers and the booth placards that indicate "X299," while a metal embossing on the Designare 10G's VRM heatsink reads X499. Intel in its Computex 2019 keynote announced that it will introduce new Core X HEDT processors. It's been over 2 years since the first Core X "Skylake-X" processors launched in Q2-2017. Intel refreshed the lineup in 2018 with 9th generation branding and soldered TIM, with a few specification improvements across the product-stack, but a largely unchanged silicon. It's likely that the Fall 2019 release could see new chips with increased core counts, perhaps even the fabled 22-core die, and some hardware mitigation against recent security vulnerabilities.



View at TechPowerUp Main Site

 

[Xzibit]

Hardware Unboxed talked to Gigabyte and they said that X499 was a place holder name before they settled on X299G

 

[efikkan]

I assume Intel decided on the "X299G" name fairly late in the process.
They do already have the 400/495 series chipsets lined up for "next year", perhaps they want to use the "X499" naming at some point later?

 

[LiveOrDie]

I assume Intel decided on the "X299G" name fairly late in the process.
They do already have the 400/495 series chipsets lined up for "next year", perhaps they want to use the "X499" naming at some point later?

X299G does make more sense seeing its the same chipset.

 

[dj-electric]

Don't care about the chipset name, just care that Intel will finally get CPU prices that make sense on this platform. SKL-X 9000 refused, maybe next refresh won't. Maybe the pricing of and viability of this socket will keep going out in flames. Maybe we will get another 16 core CPU that cost twice as much as the competition by the end of this year.

 

[HwGeek]

If X499 brings nothing new, it will be DOA and just make AMD to introduce the X599 TR 3000 sooner, just by looking at TR 2920X 12C@180W vs 3900X 12C@105W that has higher clocks and IPC, with this logic the new 32C will be just ~150W vs 250W 2990WX, now how AMD gonna use that extra 100W headroom looks interesting, they can clocks go much higher so no current/ next 2066 Refresh can match it.

Also if X570 boards gone crazy with VRM quality and power delivery, then the new TR4 boards will be even better and will run the 48C~64C at high clocks, Intel needs new gen, not another refresh .

 

[efikkan]

Don't care about the chipset name, just care that Intel will finally get CPU prices that make sense on this platform. SKL-X 9000 refused, maybe next refresh won't. Maybe the pricing of and viability of this socket will keep going out in flames. Maybe we will get another 16 core CPU that cost twice as much as the competition by the end of this year.

Comparing prices of 16-core vs. 16-core is unfair and pointless. Workstations are built for a purpose, and the metric that matters is performance, if it takes 10 or 100 cores to achieve that performance level is less important. Skylake-X's 10-12 cores performs very well against AMD's 16-core.

I do hope prices are adjusted though, not because the market demands it until Threadripper 3 arrives, but because it's been two years since the last adjustment.

If X499 brings nothing new, it will be DOA and just make AMD to introduce the X599 TR 3000 sooner, just by looking at TR 2920X 12C@180W vs 3900X 12C@105W that has higher clocks and IPC, with this logic the new 32C will be just ~150W vs 250W 2990WX, now how AMD gonna use that extra 100W headroom looks interesting, they can clocks go much higher so no current/ next 2066 Refresh can match it.

Cascade Lake-X will be a refresh, featuring tweaks, cache adjustments, AVX additions and higher clocks. So there is nothing to get very excited about, but that doesn't make it less of a solid product, for the time being it will be the best performing HEDT lineup.

Threadripper 3 will certainly be interesting when it arrives, but don't expect it to crush Intel in IPC based on Cinebench scores. The viability of Threadripper 3 as a good HEDT platform will also depend on how well the new chiplet design deals with latency. For workstations, real world performance is what matters, not fancy specs.

 

[Deleted member 178884]

The real question, is it yet another socket meaning current gen x299 boards won't support the new CPUs? Lga 2066v2?

 

[Tomgang]

The more i see what intel has to offer this round, the more i am getting dissapointed. Are they really only maybe give a few more cores and thats it.

Same chipsæt, same socket, still 14 nm, more or less same core count.


So far i am really gonna believe that amd is really gonna win this round. Because as i am a intel man, but lets face it, amd ryzen 3000 looks far more interesting this time than intels offering for not to mention the price difference. If amd new chips can have single core performance like intels 9000 cpu gen and can overclock to around 4.7-4-8 ghz with the price range they are at. Amd is really stading strong.

 

[dj-electric]

Comparing prices of 16-core vs. 16-core is unfair and pointless. Workstations are built for a purpose, and the metric that matters is performance, if it takes 10 or 100 cores to achieve that performance level is less important. Skylake-X's 10-12 cores performs very well against AMD's 16-core.

I do hope prices are adjusted though, not because the market demands it until Threadripper 3 arrives, but because it's been two years since the last adjustment.


Cascade Lake-X will be a refresh, featuring tweaks, cache adjustments, AVX additions and higher clocks. So there is nothing to get very excited about, but that doesn't make it less of a solid product, for the time being it will be the best performing HEDT lineup.

Threadripper 3 will certainly be interesting when it arrives, but don't expect it to crush Intel in IPC based on Cinebench scores. The viability of Threadripper 3 as a good HEDT platform will also depend on how well the new chiplet design deals with latency. For workstations, real world performance is what matters, not fancy specs.

Zen2 might change all of that regarding skylake-X as competition. People who need fast cores, and multiple, would possibly prefer going 16 core X570 instead of 16 core X299
X399 will just amplify that two-fold, and possibly further
Setting up compute clusters will even further enhance the choice for pay-less get-more.
Intel needs to wake up and stop selling 8-10 core HEDT CPUs for 600-900$

 

[ValenOne]

Comparing prices of 16-core vs. 16-core is unfair and pointless. Workstations are built for a purpose, and the metric that matters is performance, if it takes 10 or 100 cores to achieve that performance level is less important. Skylake-X's 10-12 cores performs very well against AMD's 16-core.

I do hope prices are adjusted though, not because the market demands it until Threadripper 3 arrives, but because it's been two years since the last adjustment.


Cascade Lake-X will be a refresh, featuring tweaks, cache adjustments, AVX additions and higher clocks. So there is nothing to get very excited about, but that doesn't make it less of a solid product, for the time being it will be the best performing HEDT lineup.

Threadripper 3 will certainly be interesting when it arrives, but don't expect it to crush Intel in IPC based on Cinebench scores. The viability of Threadripper 3 as a good HEDT platform will also depend on how well the new chiplet design deals with latency. For workstations, real world performance is what matters, not fancy specs.

Zen 2 has dual 256 bit wide AVX units improvements which rivals non-512 bt AVX Intel competition.

Intel needs to push AVX-512 optimizations.

 

[nemesis.ie]

At this rate, the next TR boards are going to be stunning.

 

[EarthDog]

Hardware Unboxed talked to Gigabyte and they said that X499 was a place holder name before they settled on X299G

Investigation instead of guessing and speculation FTW.

 

[Vya Domus]

Intel needs to push AVX-512 optimizations.

Why though ? No mainstream piece of software that I know of uses AVX-512, few even use AVX/AVX2. On the contrary, this is a waste of silicon and TDP.

 

[ValenOne]

Why though ? No mainstream piece of software that I know of uses AVX-512, few even use AVX/AVX2. On the contrary, this is a waste of silicon and TDP.

Your argument was used for Ryzen V1's introduction while Intel pushed dual 256 bit AVX v2 benchmarks.

Intel's new Icelake supports AVX 512.

 

[Vya Domus]

No, AVX 512 really is a waste of silicon, not because of performance considerations but for other practical reasons. It turns out that even 4 wide fp32 SIMD (SSE) is enough for a lot of stuff (eg. ray tracing) and going beyond that doesn't bring a whole lot, AVX2 is already pushing the limits of how much speed up can be gained from widening vector instructions.

Having more 256 bit wide instructions is way more useful right now that anything else. Remember that these things aren't interchangeable, the performance of 2x256bit cannot be matched by 1x512bit. More narrower vector instruction are more robust. It seems to me that whoever gets to decide these things at Intel seems out of touch with the real world, there so many anomalies such as desktop Pentiums not having AVX support but low power M Core products do.

 

[kapone32]

At this rate, the next TR boards are going to be stunning.

The X399 boards are already stunning it will the 3rd gen of TR4 chips that will be an absolute game changer.

 

[efikkan]

No, AVX 512 really is a waste of silicon, not because of performance considerations but for other practical reasons.

For buyers not running custom software, AVX-512 have no use at present, but that will gradually change over the next years.

But this is the chicken and the egg problem again.

It turns out that even 4 wide fp32 SIMD (SSE) is enough for a lot of stuff (eg. ray tracing) and going beyond that doesn't bring a whole lot, AVX2 is already pushing the limits of how much speed up can be gained from widening vector instructions.

Not really.
While SSE is of course much faster than single FPU operations, vector workloads scale very well on AVX2 and AVX-512. But the software have to be readjusted and recompiled to support it. Luckily, changing code from AVX -> AVX2 -> AVX-512 is nearly trivial.

SSE is BTW on it's way to being deprecated.

Having more 256 bit wide instructions is way more useful right now that anything else. Remember that these things aren't interchangeable, the performance of 2x256bit cannot be matched by 1x512bit.

Intel have 2x512-bit on Skylake-X/SP.