• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Computex 2024 Preview: The Hardware Side

erek

erek

8=D
2FA
Joined
Dec 19, 2005
Messages
18,248
"
  1. Computex 2024 — opening keynote by AMD CEO Lisa Su
    Su's speech is set for the morning of June 3, 2024, shortly before the formal start of the show..."


"This year, we expect to learn a lot more about the future of local AI processing on PCs. A number of vendors will show off new systems based on the new Snapdragon X series and high-end or even next-gen processors from AMD and Intel. Here’s what we expect from Computex 2024:


  • New chips from Intel and AMD: Both of these companies are holding major keynote events and we’d hazard to guess that they will show off new products. After the recent announcement that only Snapdragon X-powered laptops would be able to run Windows 11’s new AI features, the two x86 manufacturers need to answer back. We expect them to show off new processors with built-in NPUs that can exceed the 40 TOPS minimum Microsoft requires to classify a laptop as a Copilot+ PC.

  • Something interesting from Nvidia: Nvidia has made a lot of headlines for its datacenter and professional GPUs, but the company hasn’t done much to impress consumers of late. Its flagship RTX 4090 GPU is nearly two years old at this point, which leaves us wondering: what’s next? CEO Jensen Huang has a major keynote on Sunday evening, and we would not be surprised if he talked about next-gen graphics for end users and gaming. But since this is an industry event, we also expect him to tout datacenter wins. The most interesting question: Will Nvidia make its own processor for Windows on Arm (WoA) laptops and compete with Qualcomm?

  • More Copilot+ PCs: Yes, we saw a slew of new Copilot+ laptops at Microsoft’s Build press conference the other day, but we expect to see even more break cover at Computex.

  • Next-Gen AMD and Intel Laptops: Since Intel’s Lunar Lake CPUs are due in Q3, there’s a great chance that we’ll see Lunar Lake-powered laptops at Computex. Laptops based on AMD’s next-gen chips may also make an appearance.

  • Surprises from Qualcomm: Qualcomm has a major keynote happening on Monday, and while the company could just use the time to hype its Snapdragon X chips, those are kind of old news as they’re about to appear in shipping laptops starting in June. We’re more curious to see if the company talks about what comes after Snapdragon X — and whether it’s going to target desktops or servers for Windows on Arm.

  • More rear-connector cases and components: One of the biggest PC building trends of the past year has been the growing popularity of systems that let you plug all of your power cables, including the GPU power on some systems, into the back of the motherboard. MSI, Asus and Gigabyte are all leaders in this space, although Maingear holds a patent on some of the technology. We expect to see a lot more motherboards and cases that support rear-connector boards and components.

  • Unique peripherals and components: It’s fun when you see a unique keyboard or a mouse with a brand new feature, or even an automated desk. We’re always on the lookout for the interesting stuff, which is sometimes good, sometimes bad, but always noteworthy. Last year, we were impressed by Cooler Master’s MasterHUB tech, a modular version of the Steam Deck, but we still haven’t seen it come out.

Keynote Schedule​

The following four major keynotes are taking place during Computex. We’ll be reporting live from each of them and posting a streaming link if one becomes available.

Note that the times below are local Taipei times (UTC+8). So, if you live in the eastern U.S., Taiwan is 12 hours ahead of you. If you live in the Pacific time zone, it’s 15 hours ahead of you. And, if you live in the UK, it’s 7 hours ahead of you."

Source: https://hardforum.com/forums/hardforum-tech-news.116/post-thread
 
So please riddle me this:

If everything AI and Copilot is all about cloud processing and hoovering up as much data as possible...

Why do we need local NPU power? Isn't all of this stuff unavailable offline, and requires a Microsoft Account connected to Miscrosoft's servers?
 
I saw Alphacool hyping a new version of their AIO pump/block. Apparently a DDC-style pump is replacing their DC-LT in it. I'm curious if it will be strong enough to run a full Cpu/gpu loop.

Hopefully computex will have more info on the product.
 
KazeoHin said:
So please riddle me this:

If everything AI and Copilot is all about cloud processing and hoovering up as much data as possible...

Why do we need local NPU power? Isn't all of this stuff unavailable offline, and requires a Microsoft Account connected to Miscrosoft's servers?
Click to expand...
They want to move more things to the local system. Probably not everything, some of the LLMs are real heavy on that first "L" (large), but some stuff. There's sort of two sides to the reasons they would do this:

1) From a consumer standpoint, local processing is faster and works even if there are internet problems. So it'll feel better if they can offload as much to the local system as possible to get a more responsive setup.

2) From a corporate standpoint, because this shit costs them a lot of money and they'd rather you spent your resources on it instead of theirs.
 
Sycraft said:
They want to move more things to the local system. Probably not everything, some of the LLMs are real heavy on that first "L" (large), but some stuff. There's sort of two sides to the reasons they would do this:

1) From a consumer standpoint, local processing is faster and works even if there are internet problems. So it'll feel better if they can offload as much to the local system as possible to get a more responsive setup.

2) From a corporate standpoint, because this shit costs them a lot of money and they'd rather you spent your resources on it instead of theirs.
Click to expand...
I agree with everything you say.

But will ANY of the Copilot+ features be available offline? Isn't the whole profit incentive of Copilot+ to get as much data as possible to sell to advertisers? How does that influence their profits if it can be done offline?
 
KazeoHin said:
So please riddle me this:

If everything AI and Copilot is all about cloud processing and hoovering up as much data as possible...

Why do we need local NPU power? Isn't all of this stuff unavailable offline, and requires a Microsoft Account connected to Miscrosoft's servers?
Click to expand...
Because Microsoft invested Billions into AI and they are trying to find a use for it. The better question is, why do we need NPU's when the GPU is faster? I get that NPU's are more power efficient but how often does Microsoft think they'll have a use for AI in Windows? The gamer in me would rather see laptops with more power GPU's than NPU's. Everyone went nuts for AI, and Microsoft is now trying to justify that investment with FOMO.

Another good question is why is the tech news media going nuts for Qualcomms Snapdragon Elite X/Plus? Apple going ARM hasn't done any good for them so why mimic Apple now? Microsoft has the majority market share even on AMD's and Intel's CPU's, so why go ARM for a 3rd time? I get that Microsoft is a publicly trade company and must always seek growth, but this seems like a great way to lose people.
 
Sycraft said:
They want to move more things to the local system. Probably not everything, some of the LLMs are real heavy on that first "L" (large), but some stuff. There's sort of two sides to the reasons they would do this:

1) From a consumer standpoint, local processing is faster and works even if there are internet problems. So it'll feel better if they can offload as much to the local system as possible to get a more responsive setup.

2) From a corporate standpoint, because this shit costs them a lot of money and they'd rather you spent your resources on it instead of theirs.
Click to expand...
This will have to see, maybe for video, but server side compute can be quite faster (because it make economic sense being use all the time by different customer versus just sometime) and the text send or receive on modern broadband internet, image as well with how their compressed it for how those ai look at them... gpt4-o is much faster than anything I ever tried to run local on a 3070 or cpu I think.

GPT 4-0 seem to be able to respond to an audio request in less than 250ms, video of people using grog for example, that probably already close to close enough in terms of latency for a lot of stuff, people can play video game on a server after all, for DLSS, blur on a video phone call, running local can be good for AI, but for a lot of stuff it is not necessary.

View: https://youtu.be/pRUddK6sxDg?t=233

KazeoHin said:
If everything AI and Copilot is all about cloud processing and hoovering up as much data as possible...

Why do we need local NPU power? Isn't all of this stuff unavailable offline, and requires a Microsoft Account connected to Miscrosoft's servers?
Click to expand...
it is obviously not all about hoovering up as much data and one of the reason for local NPU power, there like 40 local different LLM type ai that come on those laptop.

Many people would not like something like recall to call home (and we can imagine handling all of that for free on their side would cost them way too much) and those companies would like to pass some of the cost to run the model down to customer, it is extremely cheap to have those 25 or so NPU part of a soc, almost a non brainer.
 
Last edited:
DukenukemX said:
Apple going ARM hasn't done any good for them so why mimic Apple now?
Click to expand...
Maybe you would think people that mimic know more about this than random online people ?

DukenukemX said:
why do we need NPU's when the GPU is faster?
Click to expand...
Is anyone trying to sales NPU for people that have a GPU and do not need ultra low power for task ? I am not sure we do, but they seem to be really cheap in cost, power, die space.
On Apple M1 there npu were like 0.2/0.25 mm, would not surprise me if we can consider it nearly "free", if game start to use the NPU for some physic simulation, text to speech, speech to emotion, could be well worth it even for gamer.
There probably very little lost in term of the GPU on a gaming laptop.

DukenukemX said:
how often does Microsoft think they'll have a use for AI in Windows?
Click to expand...
One cool new trick windows do with AI, if you ctrl-c something when you ctrl-windows-V (or shift, something more than ctrl-v) you can paste in intelligent way (that you can type in regular english to do), if it is a screenshot of a table you can have it pasted back in markdown, csv, etc... if you copied an mp3 you can paste the transcript in notepad.

you can ask in plain english for complex terminal commands line (that make your wsl Linux bash inside your windows quite powerful for the novice)
 
Last edited:
KazeoHin said:
So please riddle me this:

If everything AI and Copilot is all about cloud processing and hoovering up as much data as possible...

Why do we need local NPU power? Isn't all of this stuff unavailable offline, and requires a Microsoft Account connected to Miscrosoft's servers?
Click to expand...
Copilot doesn’t train on user data and it runs locally, they train it in the cloud and push the updated algorithms down. Microsoft trains on the Data collected by ChatGPT, but Copilot data isn’t shared back. It’s how Microsoft differentiates between the two services.
Microsoft has essentially adopted Apples strategy they use for the Neural Engine.
 
LukeTbk said:
One could new trick windows do with AI, if you ctrl-c something when you ctrl-windows-V (or shift, something more than ctrl-v) you can paste in intelligent way (that you can type in regular english to do), if it is a screenshot of a table you can have it pasted back in markdown, csv, etc... if you copied an mp3 you can paste the transcript in notepad.

you can ask in plain english for complex terminal commands line (that make your wsl Linux bash inside your windows quite powerful for the novice)
Click to expand...
I had it build me different CSV files for Laptops, Desktops, iOS, MacOS, and ChroleOS devices by automatically parsing through 4 years of invoices and putting it into a format I could import into the inventory system that’s been pushed on me for the upcoming year.
It collected and populated well over 900 devices in an afternoon and the only thing I needed to correct was a few MAC addresses. For whatever reason it removed leading and trailing 0’s in the addresses, but there were less than 50 that needed manual verification when done.
 
LukeTbk said:
On Apple M1 there npu were like 0.2/0.25 mm, would not surprise me if we can consider it nearly "free", if game start to use the NPU for some physic simulation, text to speech, speech to emotion, could be well worth it even for gamer.
There probably very little lost in term of the GPU on a gaming laptop.
Click to expand...
The possibilities the NPU opens up for NPC pathing and decision trees is very interesting.
NPC AI is very CPU thread heavy and one of those jobs that is very difficult to spread across threads.
It could also be awesome at accelerating physics effects, almost all the popular libraries are CPU driven, Bullet, Havok, even PhysX. There are significant advantages there made available by AVX512 but the NPU could potentially do it even better without reliance on a somewhat niche instruction set.
 
LukeTbk said:
Maybe you would think people that mimic know more about this than random online people ?
Click to expand...
Have you seen what Microsoft has done lately? Have you seen the names Microsoft has given to their Xbox consoles? If Microsoft announced that the company is now run by a bunch of monkeys in a room, then I'd have higher hopes. Assuming they don't write Shakespeare first.
wDjgzOS3TpoPvLj-Y92cLXpyhGVJQkB3HRBDw=w400-h300-no.gif

Is anyone trying to sales NPU for people that have a GPU and do not need ultra low power for task ? I am not sure we do, but they seem to be really cheap in cost, power, die space.
On Apple M1 there npu were like 0.2/0.25 mm, would not surprise me if we can consider it nearly "free", if game start to use the NPU for some physic simulation, text to speech, speech to emotion, could be well worth it even for gamer.
There probably very little lost in term of the GPU on a gaming laptop.
Click to expand...
Apple's M1 NPU is no where near the TOPS that Micrsosoft requires, which is 40 TOPS. Is it "free" silicon when I'm still stumbling on what I can do with it? All the things you mentioned we've already had, except maybe with an NPU it's more power efficient?
One cool new trick windows do with AI, if you ctrl-c something when you ctrl-windows-V (or shift, something more than ctrl-v) you can paste in intelligent way (that you can type in regular english to do), if it is a screenshot of a table you can have it pasted back in markdown, csv, etc... if you copied an mp3 you can paste the transcript in notepad.
Click to expand...
These sound like really cool ideas that I will never use. If Microsoft is supplying these programs with Windows then that's cool, because what you're describing is something that already exists in software, just not software included with Windows.
you can ask in plain english for complex terminal commands line (that make your wsl Linux bash inside your windows quite powerful for the novice)
Click to expand...
I can type faster than I can speak. I also don't want to sit in front of my computer to dictate commands. This isn't a new thing either, as I hate using voice command sometimes because I feel like an idiot. Especially when the voice command wasn't correct the first time, so I have to repeat louder and clearer so it works correctly the second time.
 
DukenukemX said:
Have you seen what Microsoft has done lately?
Click to expand...
1b758eb16266f6f416c1347495dac5af.png


Change Xbox failure into Playstation "success" story, not sure if we would see it on that graph.

DukenukemX said:
Apple's M1 NPU is no where near the TOPS that Micrsosoft requires, which is 40 TOPS
Click to expand...
The small A17 pro in the Iphone has an NPU with 35 TOPS, the second you have hardware support for Int8, that amount isreally not that high to reach, we are talking about adding a small part of phone battery powered chips to laptop. It always compute with budget but not with the 120-300mm discrete gaming GPU of a gaming laptop, more a bit more cache on the cpu.

DukenukemX said:
except maybe with an NPU it's more power efficient?
Click to expand...
And price-die size efficient, they are extremelly cheap, this is using the easiest-fastest-simplest compute operation (int 8 bits, 4 bits even possible):
https://www.eetimes.eu/putting-ai-into-the-edge-is-a-no-brainer-heres-why/
Exynos’s edge AI neural processing unit (NPU) represents roughly 5% of the chip’s total cost. That translates to about US$3.50 each.

$2 USD of chips, will buy you a lot of AI inference compute and will do a lot of things with an extremely small power budget (which from my understanding is the big attraction, the power efficient, they can do interesting things with half a watt)

DukenukemX said:
These sound like really cool ideas that I will never use. If Microsoft is supplying these programs with Windows then that's cool, because what you're describing is something that already exists in software, just not software included with Windows.
Click to expand...
It is already there
https://learn.microsoft.com/en-us/windows/powertoys/advanced-paste
It is open source:
https://github.com/microsoft/PowerToys/tree/main/src/modules/AdvancedPaste/AdvancedPaste

DukenukemX said:
I can type faster than I can speak.
Click to expand...
If you know what to type obviously (well some command can get quite long), for someone that never used bash, typing (or saying, you can obviously still type in plain english), change the end of lines of all the files in all the subfolder to be of Unix type can be faster to finding how to do it.
 
Last edited:
LukeTbk said:
View attachment 656861

Change Xbox failure into Playstation "success" story, not sure if we would see it on that graph.


The small A17 pro in the Iphone has an NPU with 35 TOPS, the second you have hardware support for Int8, that amount isreally not that high to reach, we are talking about adding a small part of phone battery powered chips to laptop. It always compute with budget but not with the 120-300mm discrete gaming GPU of a gaming laptop, more a bit more cache on the cpu.


And price-die size efficient, they are extremelly cheap, this is using the easiest-fastest-simplest compute operation (int 8 bits, 4 bits even possible):
https://www.eetimes.eu/putting-ai-into-the-edge-is-a-no-brainer-heres-why/
Exynos’s edge AI neural processing unit (NPU) represents roughly 5% of the chip’s total cost. That translates to about US$3.50 each.

$2 USD of chips, will buy you a lot of AI inference compute and will do a lot of things with an extremely small power budget (which from my understanding is the big attraction, the power efficient, they can do interesting things with half a watt)


It is already there
https://learn.microsoft.com/en-us/windows/powertoys/advanced-paste
It is open source:
https://github.com/microsoft/PowerToys/tree/main/src/modules/AdvancedPaste/AdvancedPaste


If you know what to type obviously (well some command can get quite long), for someone that never used bash, typing (or saying, you can obviously still type in plain english), change the end of lines of all the files in all the subfolder to be of Unix type can be faster to finding how to do it.
Click to expand...

TechPowerUp @ Computex 2024 Preview: AI Everywhere! Next Gen Platforms and Teasers​

COMPUTEX PINNED by btarunr Today, 11:04 Discuss (9 Comments)
TechPowerUp flies to Taipei this weekend to bring you our biggest ever Live coverage of what is shaping up to be the most exhaustive edition of Computex ever, with hundreds of brands and thousands of new products on display. We have scheduled meetings with all brands from the world of hardware and gaming, so that we can get you full coverage, including hands-on with the hardware you're looking forward to. Besides the show floor, there's a lot happening at Computex, with leading hardware companies announcing their latest platforms. The running theme of course is AI for everyone, and AI everywhere. Since Computex is a mainly PC-focussed expo, the dominating device is bound to be the AI PC. This would mean a slew of core hardware and peripherals enhanced with on-device AI acceleration capabilities.

After the break, we've compiled a list of announcements that we expect from major companies like Intel, AMD, NVIDIA
 
AnandTech preview:

https://www.anandtech.com/show/21418/computex-2024-keynote-preview

Intel: Bringing AI Everywhere​

When: June 3rd, 11:00pm ET (8:00pm PT)
What: Intel CEO Pat Gelsinger will present a keynote entitled “Bringing AI Everywhere.”
What To Expect: The final of the major Computex keynotes, Intel has been the most up-front about what to expect from their hour-long presentation. The company’s Computex 2024 page already outlines their four major topics: AI PCs, Xeon 6 Processors, Gaudi AI accelerators, and Intel’s OpenVINO software ecosystem.
On the consumer hardware front, the company set the table with a significant teaser earlier this month about their forthcoming mobile PC SoC, Lunar Lake. The next generation of Core Ultra processors, Intel is touting significant energy efficiency gains for the mobile-focused chip, with new architectures driving their Performance and Efficiency CPU cores, Xe2 GPU, and a much faster 45+ TOPS (INT8) NPU. While the Lunar Lake announcement is coming relatively soon after the Meteor Lake launch, Intel has made it clear that it’s not going to hold back on shipping future products; they are looking to make up for lost time. Still, Lunar Lake devices are not expected to hit retail shelves until Q4 of this year, so this announcement is coming months in advance of the hardware itself.
On the server front, Intel has been publicly prepping for the launch of a new generation of Xeons with the Xeon 6 platform. The most notable part of this being the release of the company’s first Efficiency-core Xeon, Sierra Forest. Sierra Forest is set to be the first Xeon 6 chip out the door this year, and will offer up to 288 E cores on a single chip, allowing Intel to tap into the many (many) core CPU markets that AMD and Arm-based rivals have been unopposed at thus far.
Finally, the company has fully pivoted its server AI accelerator strategy to its Gaudi accelerators. Gaudi 3 was introduced back in April, and while it isn’t expected to go toe-to-toe with NVIDIA’s top accelerators in every workload, Intel is betting that they can beat NVIDIA on critical workloads, all while undercutting them significantly in pricing. The first Gaudi 3 parts are set to be released in the second half of this year, so hopefully we’ll be hearing a bit more about Intel’s plans as part of their keynote.


AMD: The Future of High-performance Computing in the AI Era​

When: June 2nd, 9:30pm ET (6:30pm PT)
What: AMD CEO Dr. Lisa Su will be presenting a keynote entitled “The future of high-performance computing in the AI era.”
What To Expect: The first off the official Computex keynotes, AMD’s keynote is the closest thing the show has to a “prime” keynote that kicks off the whole show. Like NVIDIA, AMD’s AI-focused description intersects virtually their entire product stack. But as AMD has been consistently executing on a well-published architecture roadmap over the last few years, we have a good idea of what to expect here.
The big expectation here is fresh news around AMD’s Zen 5 CPU core architecture, and the chips built around it. AMD’s most recent Zen 5 roadmap has it slated to deliver all three flavors of Zen 5 by the end of this year, and we’re coming up on the two-year anniversary of the Zen 4 architecture launch.
Along with client chips, AMD has been pushing their server CPUs hard, and they’ve previously told investors that the next-gen EPYC Turin CPU is “looking great”. So we’ll likely hear about both client and server Zen 5 product plans during this keynote.
On the GPU/accelerator side of matters, AMD is mid-cycle (at best) with their Instinct MI300 series accelerators. With the company’s sales repeatedly beating their own expectations, AMD doesn’t seem to need much help moving this premium silicon right now. But with AI being the operative buzzword of this year’s Computex (and indeed, the computing industry as a whole), it would be weird for AMD to not have something to say about their rapidly growing AI accelerator product line.


NVIDIA: What’s Next for the AI Ecosystem​

When: June 2nd, 7:00am ET (4:00am PT)
What: NVIDIA’s leather jacket-wielding CEO Jensen Huang will be presenting a talk “outlining what’s next for the AI ecosystem.”
What To Expect: Of the 4 Comuptex-aligned keynotes, NVIDIA’s is both the first and the most mysterious. The company has not telegraphed its intentions for its keynote – or even how long it will run. And with virtually all of NVIDIA’s products intersecting the AI market in some fashion, the “AI ecosystem” could conceivably cover anything from servers to smartphones.
With that said, NVIDIA is coming into this show as the vendor farthest from the start of their next product cycle. The company already presented a detailed roadmap for it’s all-important server products at their own GTC conference two months ago, leaving them with relatively little to announce at this spring’s expo. And NVIDIA’s consumer GPUs have traditionally followed a 2+ year cadence, meaning they’re not due for refresh until later this year at the earliest. Still, NVIDIA is solidly in the driver’s seat of its own destiny right now thanks to the wild success of its Grace Hopper-based accelerators, so the company can (and will) do whatever it pleases.
Notably, even NVIDIA’s keynote itself is a big rogue. While the AMD, Intel, and Qualcomm keynotes are all official Computex keynotes taking place at the TaiNEX 2 expo hall, NVIDIA’s keynote is off-site, taking place the night before the show officially starts at the National Taiwan University Sports Center.



Qualcomm: The PC Reborn​

When: June 3rd, 1:30am ET (10:30pm PT)
What: Qualcomm's CEO, Cristiano Amon, will be presenting a keynote entitled “The PC Reborn”
For now we'll pretend that they don't exist (until they establish themselves properly in the PC space).
While Snapdragon X is not Qualcomm’s first effort to ship an Arm-based SoC for Windows devices – there are 3 generations of 8cx Gen 3 platforms that everyone is happy never to mention again – the Snapdragon X is Qualcomm’s most serious effort yet.
 
Nvidia keynote live blog:

https://www.techradar.com/news/live/nvidia-computex-2024-keynote-liveblog

Summary:
OK, well, I can see why this wasn't an official Computex keynote. If you were hoping for consumer product news, you got a 10- to 15-second mention of a few RTX laptops that will have enhanced AI features, but given how quickly Jensen moved past all that to get back to the data center stuff really tells you all you need to know about where Nvidia is right now. They are no longer a graphics company, as Jensen reportedly told employees several months back, and every Nvidia keynote and live stream I've watched in the last year and a half really just reinforces that fact.

Nvidia is absolutely, 100% going to become an AI chip company, and whatever gaming appendage sticks around for a few years will become less and less of a focus for the company.
 
Last edited:
2024-06-02%2018_53_20.jpg



no, price with a july release ? +10% for Far Cry 6, it is an AMD sponsered game but still a bit of an odd choice considering the quite unimpressive upgrade., both AMD-Nvidia has not been that vocal in term of actual consumer product news, if I am not missing something, at least AMD had the above but that seem the most boring possible generation uplift unless you use AVX 512 that AES XTS overperformance probably due to that, but nothing for RDNA 4..... And Nvidia nothing for the 5090-5080...
 
Last edited:
AMD: We know naming is confusing! so here is a useful guide for naming conventions that shows that each number will be consistent and clear which model is newer and better! We will use for all our mobile processors moving forward 😊😊

AMD less than 18 months later: WOO RYZEN AI 9 HX 370 😱😱😱
 
Marees said:
AMD computex liveblog:

Zen 15% launching in July
no info of prices

https://www.anandtech.com/show/21423/the-amd-computex-2024-keynote-live-blog-630pm-pt0230-utc
Click to expand...

Summary:

The new Zen 5 CCDs are being fabbed on one of TSMC's 4nm processes (AMD hasn't confirmed which flavor), which is a modest shrink from the N5 process used for Zen 4 CCDs.

These enhancements collectively aim to deliver significant performance gains over the previous Zen 4 microarchitecture, with AMD touting an average (geomean) 16% IPC uplift over Zen 4 in desktop workloads. It should be noted, however, that the top result in this benchmark collection is in the GeekBench 5.3 AES XTS benchmark, which leverages the VAES512 and VAES256 extensions of the AVX-512 instruction set. So AMD's AVX-512 SIMD changes greatly impact that benchmark in particular (though not exclusively).


https://www.anandtech.com/show/2141...ter&utm_medium=social&utm_campaign=socialflow
 
Marees said:
Summary:

The new Zen 5 CCDs are being fabbed on one of TSMC's 4nm processes (AMD hasn't confirmed which flavor), which is a modest shrink from the N5 process used for Zen 4 CCDs.

These enhancements collectively aim to deliver significant performance gains over the previous Zen 4 microarchitecture, with AMD touting an average (geomean) 16% IPC uplift over Zen 4 in desktop workloads. It should be noted, however, that the top result in this benchmark collection is in the GeekBench 5.3 AES XTS benchmark, which leverages the VAES512 and VAES256 extensions of the AVX-512 instruction set. So AMD's AVX-512 SIMD changes greatly impact that benchmark in particular (though not exclusively).


https://www.anandtech.com/show/2141...ter&utm_medium=social&utm_campaign=socialflow
Click to expand...

That AVX512 benchmark is impressive seeing as how Zen 4 was already an AVX beast.
 
LukeTbk said:
View attachment 657517


no, price with a july release ? +10% for Far Cry 6, it is an AMD sponsered game but still a bit of an odd choice considering the quite unimpressive upgrade., both AMD-Nvidia has not been that vocal in term of actual consumer product news, if I am not missing something, at least AMD had the above but that seem the most boring possible generation uplift unless you use AVX 512 that AES XTS overperformance probably due to that, but nothing for RDNA 4..... And Nvidia nothing for the 5090-5080...
Click to expand...
They have a Geekbench score listed. Looking at it, there's quite a bit of synthetic benchmarks listed.
tenor.gif
 
Last edited:
Gonna nab a 9950x. Just hope the memory controller has improved, as I'm eyeing 4x48gb since my system is a 3d workstation. Would be nice to avoid the 4 stick frequency problem.
 
LukeTbk said:
View attachment 657517


no, price with a july release ? +10% for Far Cry 6, it is an AMD sponsered game but still a bit of an odd choice considering the quite unimpressive upgrade., both AMD-Nvidia has not been that vocal in term of actual consumer product news, if I am not missing something, at least AMD had the above but that seem the most boring possible generation uplift unless you use AVX 512 that AES XTS overperformance probably due to that, but nothing for RDNA 4..... And Nvidia nothing for the 5090-5080...
Click to expand...

Heh, 21% increase in League of Legends. Yippie.
 
Marees said:
Microsoft's CoPilot+ will be supported on Qualcomm's APUs. (Seems that models yet to be trained on AMD/Intel APUs)
Click to expand...

Ryzen AI 300 chips are copilot+ ready, not sure what it mean
https://www.xda-developers.com/amd-microsoft-copilot-plus-partnership-ai-pc/
Lisa Su, announced that AMD was partnering with Microsoft to support Copilot+ in the next generation of Windows 11 AI PCs.

I am not sure if models are trained for specific hardware, they tend to run on everything from a CPU, GPU, NPU, Windows for the copilot stuff will tend to detect the best hardware to run on, Recall for example do work on any system it is just slow and bad without good TOPS performance.
 
Ranulfo said:
Heh, 21% increase in League of Legends. Yippie.
Click to expand...
Well maybe they will have a good frequency uplift, IPC alone just tell us half of the story (but if they boost much higher, would have shown the total performance not just the IPC)
 
erek said:
Bot eeror

https://www.tomshardware.com/tech-industry/what-to-expect-computex-2024
Click to expand...
Update the top post with that link.
Marees said:
Nvidia keynote live blog:

https://www.techradar.com/news/live/nvidia-computex-2024-keynote-liveblog

Summary:
OK, well, I can see why this wasn't an official Computex keynote. If you were hoping for consumer product news, you got a 10- to 15-second mention of a few RTX laptops that will have enhanced AI features, but given how quickly Jensen moved past all that to get back to the data center stuff really tells you all you need to know about where Nvidia is right now. They are no longer a graphics company, as Jensen reportedly told employees several months back, and every Nvidia keynote and live stream I've watched in the last year and a half really just reinforces that fact.

Nvidia is absolutely, 100% going to become an AI chip company, and whatever gaming appendage sticks around for a few years will become less and less of a focus for the company.
Click to expand...
The person doing that live blog is either an idiot or purposely posting FUD. NVIDIA almost never announces consumer video cards during or around big events like this. They usually do it during GTC, which is owned and run by NVIDIA themselves, or during an exclusive event. The one exception in recent memory was Turing, which was announced during SIGGRAPH (but not a part of SIGGRAPH). If I'm wrong, please correct me and let us know when they announced a new gaming video card at Computex.
 
erek said:
https://twitter.com/GamersNexus/status/1797235060917281023
View attachment 657576
Click to expand...
They're rolling in cash right now and just want to pump out more AI accelerators. Not surprised there's no mention of consumer GPUs, I doubt they even want to allocate any fab space to them currently. Wouldn't shock me if the 5090 weren't until next year. I feel like at some point orders will slow down and Nvidia will come back down a bit, but it seems like we still have a little way to go.
 
Decko87 said:
They're rolling in cash right now and just want to pump out more AI accelerators. Not surprised there's no mention of consumer GPUs, I doubt they even want to allocate any fab space to them currently. Wouldn't shock me if the 5090 weren't until next year. I feel like at some point orders will slow down and Nvidia will come back down a bit, but it seems like we still have a little way to go.
Click to expand...
Last three generations were announced in September, with availability in October, 2 years apart. Consumer Blackwell will more than likely be announced in September.
 
Armenius said:
Last three generations were announced in September, with availability in October, 2 years apart. Consumer Blackwell will more than likely be announced in September.
Click to expand...
Hum in that context, not sure why we thought it could be this computex that announce it, rumours were always more maybe pushed to Q1-2025 than a faster than 2 years cadence with a surprise summer launch.
 
Last edited:
You must log in or register to reply here.
Back
Top