Is it really over for Intel®?

[Phen8210]

BTW, I am biased against Windows on ARM64, let's see if I am wrong in a few years...

It's s not really about your prediction, complicated software run so slow on windows arm, emulation has issue, and some that do not work at all. This is just the surface issue, the next issue is the unsupported plugins and extension in your software, and missing features etc.

The software application ecosystem for arm64 is light years behind x86.

How it cant even use google drive for desktop is already a major deal breaker, if large companies dont show interest in the transition, it will not happen..

For mac you can just stop selling mac with x86 and cuck everyone in your ecosystem, for windows pc it has always been about interoperability, it's not the same.

 

[Outramkia]

Why am5 mobo only mostly have 4 m.2 slots? Z890 has 5 or even 6. At 4k gaming does 3d cache matter?

 

[matique]

I honestly think it's quite over for gaming, since stock latency at 6400c30 is close to 70ns, and when tuned, just under 60ns. L3 is like waiting for the train during smrt breakdown, too slow. On direct die OC there's a 200mhz uplift for ST and 300mhz uplift for MT, but it can't make up for latency being poor.

That's too bad, the boards look great especially the MSI edge itx. 4x m.2 in ITX is crazy, as is the IO for the board.

 

[matique]

Why am5 mobo only mostly have 4 m.2 slots? Z890 has 5 or even 6. At 4k gaming does 3d cache matter?

Yeah, at 4K the cache still helps to keep computations quick. With the current 4090, there can still be CPU bottleneck in some games at 4K, especially when you add RTX or path tracing computation to the mix.

 

[Outramkia]

Yeah, at 4K the cache still helps to keep computations quick. With the current 4090, there can still be CPU bottleneck in some games at 4K, especially when you add RTX or path tracing computation to the mix

Not much difference in 4k. Within 10 frames

 

[Outramkia]

I'd hoped for more m.2 slots in am5 midrange mobo. Most maxed out at 4 except x670e pro rs from asrock.

Maybe I'm just obsessed with the thought of having all ssd storage. I once disconnected my spinning drive and the sound of silence was a revelation to me.

 

[watzup_ken]

I'd hoped for more m.2 slots in am5 midrange mobo. Most maxed out at 4 except x670e pro rs from asrock.

Maybe I'm just obsessed with the thought of having all ssd storage. I once disconnected my spinning drive and the sound of silence was a revelation to me.

The problem is the real estate on motherboards can be limited. You can add "daughter boards" to stack it higher, but you will eventually run into other issues.

 

[watzup_ken]

https://www.chiphell.com/thread-2641878-2-1.html

Some nice ITXs this round though.
Quad M.2 if you don't have clearance issues with the daughterboard.
ASRock's US$299 board already gives you triple M.2 with dual TB4.0, though 5.0 is appreciated but not coming this gen.

I was previously thinking of getting an ITX board for my CPU, but I decided against it because these daugther board can really limit your cooler choices and makes it very difficult to take anything out. There is simply not enough space on an ITX board to comfortably fit so many features. Plus, they cost a bomb even if the specs are really not as good as those high end ATX board.

 

[watzup_ken]

Personally I am AMD/Intel neutral, Windows/macOS/Linux neutral. I tend to think Intel is too big to fail for US government. So Intel will survive in the end. But it probably will be a different Intel after all the problems.

BTW, I am biased against Windows on ARM64, let's see if I am wrong in a few years...

My take, Intel was "too big to fail" only because of the fact that they are the only western foundry which can produce cutting edge nodes. So if anything, the government will want to protect this chip producing capability. I said "was" because now that Intel have spun off the foundry business on some shape or form, the chip designing part of Intel is actually quite vulnerable. If they fail to keep up with competition, there is no guarantee they will survive. Any bailout will likely be for the foundry business only. I may be wrong, but this is my opinion.

The chip business is probably not that important because there are plenty of chip designing companies in US.

 

[Outramkia]

The problem is the real estate on motherboards can be limited. You can add "daughter boards" to stack it higher, but you will eventually run into other issues.

Some platform allow pcie addin card but need bios bifurcation support. Else need to get cards with plex chips.

I think right now the z890 strix gaming-e has the most m.2 nvme slot. It has 7 without the need for addin card

 

[kimsix]

No, must buyed 9800X3D on first day, reserves with simlim if you must, top up a bit more if need!

remember this famous slide?

siao liao, the 285K was specs with 6400c32 @ 250W, vs 14900K 5600c28 @ 253W, vs 9950X only 5600c28 @ 170W stock!

Intel 13/14 gen advantage was able to run P-core sustained at 5.8-6Ghz. Tsmc Amd ryzen cores are alway throttled from rated max boost clocks. 5.7Ghz specs usually around only 5.45Ghz on gaming load. Only time you see 5.7Ghz is maybe open edmw webpage 1s burst

Now Intel lost the clocks, lost the ddr5 advantage. All tsmc fault, good for efficiency but less hardy than Intel fabbed chips

I wouldn't do that if i were you

I honestly think it's quite over for gaming, since stock latency at 6400c30 is close to 70ns, and when tuned, just under 60ns. L3 is like waiting for the train during smrt breakdown, too slow. On direct die OC there's a 200mhz uplift for ST and 300mhz uplift for MT, but it can't make up for latency being poor.

That's too bad, the boards look great especially the MSI edge itx. 4x m.2 in ITX is crazy, as is the IO for the board.

 

[Phen8210]

No, must buyed 9800X3D on first day, reserves with simlim if you must, top up a bit more if need!

remember this famous slide?

siao liao, the 285K was specs with 6400c32 @ 250W, vs 14900K 5600c28 @ 253W, vs 9950X only 5600c28 @ 170W stock!

Intel 13/14 gen advantage was able to run P-core sustained at 5.8-6Ghz. Tsmc Amd ryzen cores are alway throttled from rated max boost clocks. 5.7Ghz specs usually around only 5.45Ghz on gaming load. Only time you see 5.7Ghz is maybe open edmw webpage 1s burst

Now Intel lost the clocks, lost the ddr5 advantage. All tsmc fault, good for efficiency but less hardy than Intel fabbed chips

If they went for intel fab and produce faulty chips again.. i think INTC will drop to $0.. they going for the safe route and while ironing out the problems in their fab . If made by intel fab right after the incident, who dare to buy .

 

[kimsix]

Some platform allow pcie addin card but need bios bifurcation support. Else need to get cards with plex chips.

I think right now the z890 strix gaming-e has the most m.2 nvme slot. It has 7 without the need for addin card

Amd sux, they droped the ball with x870e.

z890 is better for connectivity


Intel Z890 PCIe connectivity
16X gen5 PEG
4X gen5 nvme
4X gen4 nvme or motherboard vendor choice
4X gen4 for integrated TB4
8X gen4 downstream to chipset

AMD X870E PCIe connectivity
16X gen5 PEG
4X gen5 nvme
4X gen4 integrated TB4 ASM4242
4X gen4 downstream to chipset.

 

[elmariachi]

No, must buyed 9800X3D on first day, reserves with simlim if you must, top up a bit more if need!

remember this famous slide?

siao liao, the 285K was specs with 6400c32 @ 250W, vs 14900K 5600c28 @ 253W, vs 9950X only 5600c28 @ 170W stock!

Intel 13/14 gen advantage was able to run P-core sustained at 5.8-6Ghz. Tsmc Amd ryzen cores are alway throttled from rated max boost clocks. 5.7Ghz specs usually around only 5.45Ghz on gaming load. Only time you see 5.7Ghz is maybe open edmw webpage 1s burst

Now Intel lost the clocks, lost the ddr5 advantage. All tsmc fault, good for efficiency but less hardy than Intel fabbed chips

I've heard from those who tested that you're lucky to get even 200mhz overclock on the new CPUs without hitting the thermal wall. I think we will see overclocks slowly being irrelevant as time passes as CPUs will probably be maxed out from factory.

Those hardcore will probably delid and aim for the moon. The new Intels will be probably be much more efficient compared to 13/14 gen but I do not think it's materially going to be faster than the 14th gen at least. Probably new CUDIMM rams will make it fun to play but it may not necessarily be enough for a sizeable uplift.

I actually think Intel Bartlett may end up being much more interesting with just a 12 p-core CPU which will breathe some life into LGA1700.

https://www.tweaktown.com/news/9935...cket-up-to-8-6-hybrid-12-core-only/index.html

 

[xiaofan]

My take, Intel was "too big to fail" only because of the fact that they are the only western foundry which can produce cutting edge nodes. So if anything, the government will want to protect this chip producing capability. I said "was" because now that Intel have spun off the foundry business on some shape or form, the chip designing part of Intel is actually quite vulnerable. If they fail to keep up with competition, there is no guarantee they will survive. Any bailout will likely be for the foundry business only. I may be wrong, but this is my opinion.

The chip business is probably not that important because there are plenty of chip designing companies in US.

Actually x86_64 is still important so that Intel/AMD join forces along with some other companies to form the x86 Alliance.
https://www.extremetech.com/computing/intel-and-amd-form-x86-alliance-with-industry-partners

Even NVIDIA CEO Jensen Huang says that it Is necessary to keep the x86 Architecture alive.
https://wccftech.com/nvidia-ceo-jensen-huang-says-intel-amd-x86-alliance-says-necessary/

After all, it takes time for ARM CPU to catch up with x86_64 in PCs, Workstations and Datacenters.

 

[matique]

No, must buyed 9800X3D on first day, reserves with simlim if you must, top up a bit more if need!

remember this famous slide?

siao liao, the 285K was specs with 6400c32 @ 250W, vs 14900K 5600c28 @ 253W, vs 9950X only 5600c28 @ 170W stock!

Intel 13/14 gen advantage was able to run P-core sustained at 5.8-6Ghz. Tsmc Amd ryzen cores are alway throttled from rated max boost clocks. 5.7Ghz specs usually around only 5.45Ghz on gaming load. Only time you see 5.7Ghz is maybe open edmw webpage 1s burst

Now Intel lost the clocks, lost the ddr5 advantage. All tsmc fault, good for efficiency but less hardy than Intel fabbed chips

OC folks still tuning their chips, but can't say much before the 24th haha.

For what it's worth, I think when tuned, the chips wouldn't be bad, could even match a stock 9800x3d. There's some overhead, with possibility of running 9400-9600 in gear 2. Sticking with gear 2 would result in less memory latency than ADL/RPL.

The problem is tile latency. The stupid glue so slow. A tuned 9800x3d will likely spank the tuned 285K.

I have a gene x670e incoming, and more likely than not for this upcoming gen I'll be repping 9800x3d or better chips, unless it's not much faster than my current 13900K then I'll just save money by not upgrading hahaha.

Honestly, Intel is probably cooked for this and next tock gen. They need to solve tile latency issues, clock issues and voltage scaling as well. Hopefully they get to their 18a ASAP and that it functions well.

 

[watzup_ken]

Now Intel lost the clocks, lost the ddr5 advantage. All tsmc fault, good for efficiency but less hardy than Intel fabbed chips

Just curious why TSMC is at fault when Intel specially picked the foundry? There's no Intel chips produced using another foundry before to compare, but looking at AMD's chips, it seems clear that TSMC's node was never meant for high clockspeed. I feel these nodes are being designed for low power/ efficiency since Apple is the main cutting edge buyer and their chips don't run at 6 Ghz and above.

 

[kimsix]

Just curious why TSMC is at fault when Intel specially picked the foundry? There's no Intel chips produced using another foundry before to compare, but looking at AMD's chips, it seems clear that TSMC's node was never meant for high clockspeed. I feel these nodes are being designed for low power/ efficiency since Apple is the main cutting edge buyer and their chips don't run at 6 Ghz and above.

Yep Intel foundry is bestest. Wasted they lost their way and allow TSMC to steal the business. The old batch of Intel engineers must be chopped! Pat must save the day!
TSMC chips are boring, squeeze until no headroom. Good enough to lelong to third parties. They ownself hang up and keep increasing prices now! Pat must save the day!

My first Intel OC, i was blown away, from 3ghz turbo to 4ghz sustained, just punch in bios. This after i changed from Amd P2X4 940, which struggled to get from 3ghz to 3.2ghz

Intel OC headroom still can see up to 13900K but the power sucking was at its limit.

 

[elmariachi]

Yep Intel foundry is bestest. Wasted they lost their way and allow TSMC to steal the business. The old batch of Intel engineers must be chopped! Pat must save the day!
TSMC chips are boring, squeeze until no headroom. Good enough to lelong to third parties. They ownself hang up and keep increasing prices now! Pat must save the day!

My first Intel OC, i was blown away, from 3ghz turbo to 4ghz sustained, just punch in bios. This after i changed from Amd P2X4 940, which struggled to get from 3ghz to 3.2ghz

Intel OC headroom still can see up to 13900K but the power sucking was at its limit.

You are right about the overclockability of Intel chips. I remembered I got almost 700Mhz out of the earlier Intel CPUs I had paired to a DFI Fatality Motherboard. Play with DIP switches untill song. Those were the days where main memory timings all single digit and the craze over Mushkin DDR sticks.

I think moving forward we have to change our mindset on overclocking. Even if Intel moves to 18A, I personally do not think overclocking will be worth it unless you go down the extreme route. Then maybe yes. I think the new 285K will still fare pretty competitively but questions about memory latency and gaming is something we have to wait for reviews for. So far the 285k in benches seems to be pretty competitive but it still loses to a 9950X which is understandable because you have 24 cores 24 threads compared to 16 cores 32 threads. 9950X is scoring around 47k points in CBR23. 285K is around 45K points in CBR23 at 250W. Temps looks pretty decent. For reference, my de-lidded 14900KS (HT ON) scores 43,500 points at 5.9/4.6/4.8 at 350W/260A with max temps around 78 degrees. So clearly, clockspeed is not everything. Seems that with lower clock speed, 285K manages a much higher score at 100W less with lower clockspeeds so looks like there is a tangible gain in IPC performance.

I think the main issue for most of us enthusiasts is if 285K will be actually overall a better CPU in gaming and day-day tasks. I think 285k is good if you are upgrading from 10/11/12th Gen but those on 13/14th Gen with no issues may not be so inclined to upgrade. I am not depending on CUDIMM DDR5 in excess of 9000Mhz to materially change the odds in Intel's favour but it is a good start with a new architecture. I do not think Z890 will last longer than 1 gen unless Arrow Lake Refresh rumours of it being cancelled is false which will suck big time if 16th Gen requires another board change which is a huge bummer but typically Intel.

Best to wait for NDA to lift on 24th and see how the platform is in entirety all the way down to memory performance. My guess is the latency issue is a sore point as @matique mentioned.

 

[matique]

Agreed on life expectancy. 2 gens per board is already bad. Single gen on a board would be absolutely horrid. I think the new cores are very powerful, but it's moot if it doesn't translate in games.

Also don't think Intel will just die off. It's certainly AMD's turn to shine, I'm all for it. That said, hoping AMD and Intel both have much better offering the next gen. Because 9000 series and ARL has been mighty disappointing...but especially Intel.