That may well be true, but we Humans are also the only hope for all life on Earth, and the very planet itself.
Entropy, the inevitability of planetary cycles, along with cosmic dangers, Earth would merely be a victim of random events without our curiosity & intellect.
“Humans are a frightening oddity on this planet: not only do we lack any significant predators, we are a force which can extinguish both individual species as well as the ecosystem in its entirety.
Our very existence is a threat to the existence of life on Earth”
#Ecology
It appears that TSMC's attempt to expand in Arizona has been one giant shitshow. From US construction & trade Union causing problems to cost overruns, to site accidents due to the rushed pacing (or as one Taiwanese manger puts it, "US workers are retarded")..
I had a peek at the RTX 4060 reviewer guide.
NVIDIA tells reviewers to focus on DLSS + frame gen/hallucination, that's where their GPU shines & justifies its price tag.
Also NVIDIA: Refusing to pay studios to add RTX DLSS in PC ports. 👌
RTX consumer: It's AMD's fault! 🤡
@HardwareUnboxed
I think your commentary is spot on, particularly for mid-range GPUs. RTX is nice, in all of 3 games. Sadly, perf hit is still far too great. The solution isn't just DLSS 2.0 btw, it isn't perfect either, lots of visual artifacts, that reviewers tend to also gloss over.
@BallouxFrancois
You're analyzing the situation in China with the lens of heath & wellbeing of the population, which is not at all what the CCP objectives are.
This is why what they have been doing and are doing, make little sense.
View it another way, it makes perfect sense.
I am one of those people who bought a 3700X and cheap AM4 MB. At the time, the 9900K + Intel board price total was about 2x. The 9900K was ~10-20% faster @ 1080p. Tuned, it can get up to 25% faster at 720p.
But was it worth 2x? Nope.
The nice thing, I can slot in 5800X3D..
"The mystical future-proofing" they said...
When you decided to buy a Zen 2 3700X based on "realistic" 1440p benchmarks with a GTX 1080 Ti at launch you now might have some serious “doodoo face.” A tuned 9900K can be up to 20-30% faster using an RTX 3080 with DLSS at 1440p.
News from TW on TSMC 7N wafer customers. In Q4, AMD is set to use ~120K wafers for console SOCs, ~80K for PS5 + ~40K for SSX! This is ~80% of 7N wafers allocated to AMD in Q4. This console ramp is the main reason for lack of Zen 3 & Big Navi.
@Kanthan2030
X censorship & shadow/ghost banning algo is why I refuse to subscribe and give my $ to Elon to run this platform.
Commit to freedom of speech or not at all.
@HardwareUnboxed
What an Intel shill!
😅 j/k
AMD deserves all the humiliation. Launching a 6c CPU in 2022 for $299 is absurd on top of the expensive AM5 platform.
@IanCopeland5
There's literally no difference. NS = not significant.
Look at the Omicron data yourself, FFU dispersed similarly at ~2 x-axis, with 1 patient outlier to ~5 skewing the chart. Low N, outlier irrelevant.
You have a terrible debunking record. Try harder to understand the science.
AMD's new DX11 driver seems to be spreading the load across multi-cores better, in games that were primary thread bound. Interesting, I had thought they gave up on DX11.. 6 years late but better than never.
85% of consumers won't care that AMD's RX 6700XT is a better deal.
Because RTX ray tracing! RTX DLSS!!
They are happy to pay a premium for a worse performing GPU, with gimped vram, for RTX features.. then blame AMD why NV isn't spending more $ to push RTX features.
/facepalm
In case you want to get a 4060 tomorrow, I suggest you give this video a watch and check 6700 XT prices in your region, because it's simply a much better deal than 4060 at around the same price (50% more VRAM and much better performance overall):
After a few days of premature disappointment about Zen 4, and 31% vs 46% copium, it seems the hype train is back on track now.
5.85ghz ST!
5.5Ghz all-core PBO boost!!
6Ghz OC!!!
😁
None of this would be an issue if Jensen wasn't so fixated on planned obsolescence.
We've had mainstream GPUs with 8GB vram since 2016. Polaris, RX 480.
The 3070Ti is massively more powerful, yet still given 8GB vram.
Worse, are consumers so blinded, they defend NVIDIA.
I think pushing for more VRAM (within reason) is only a good thing. This isn't an AMD vs. Nvidia or 'GPU Wars' issue, it's about getting more for consumers and as product reviewers that's always what we're pushing for. Hopefully we can help make the next generation of mid-range
Great to see prices for Zen 4 Ryzen wasn't just temporary.
7600X $249
7700X $349
7900X $439
7950X $549
Are great perf & perf/$, combined with falling AM5 board prices, the line-up is now more than competitive vs Intel's offerings.
Great review. Damn shame that AMD's RTG actually over-sold the performance claims. It's quite an exaggeration to claim 50-70% faster than 6950XT, when its more like 35%.
I don't recall the last time RTG mislead by this much in their launch presentation.
@HardwareUnboxed
Clearly, their AI detected an anomaly when you decided to stop benchmarking for days in a row and suspiciously went outdoors for some fun.
@KimIversenShow
US has heaps of military bases all over the Middle East. Including airfields. So the argument that Israel is someone offering US special military advantage is flawed. It's actually a disadvantage, since it makes the entire region hate the US.
RFK Jr is controlled opposition.
So when is AMD going to drop the price on the 7600X and even the entire Ryzen 7000 lineup?
It's not at all competitive in perf/$ vs Intel 13th gen. Worse when factoring in more expensive AM5 boards.
Are they just going to keep head in sand & lalala away? 😅
@HardwareUnboxed
Imagine being anti-consumer for so long, and yet still have 88% of the dGPU marketshare! 😅
There's little incentive to be pro-consumer, when consumers willingly throwing money at NVIDIA.
H.U. is correct. I recall seeing some claim that H.U. was bias towards AMD in their reviews, because they had RDNA2 GPUs doing slightly better than other tech sites. The thing is, they test more games, and modern titles thus they are "forward looking" compared to others.
Quite a few of you have asked for my take on this data, or for us to re-test. The 'June 2021' data lines up with what we found in 'Dec 2020' and our margins have not changed. I assume this is down to the games used for testing, rather than any improvements on AMD's side.
2017, 8c/16 6900K, $1000.
2017, 8c/16 1800X, $499.
2019, 8c/16t 3800X, $399.
2020, 8c/16 5800X, $449.
The lack of a lower 8c SKU like 5700X at the current time, shouldn't be reason for gamers to cry foul that they are being ripped off or stepped on.
@ArmchairW
@ChowdahHill
Once the escorts run out of anti-air interceptors, the Carrier has to leave the area or risk being sunk. It's just US navy doctrine in action.
VLS DDs that are costly with limited interceptors vs mass drones & cheap missiles is not a good strategy.
My wife has been consumed by this Heard vs Depp trial past 2 weeks.
She said it's amazing how nasty & abusive a woman can be..
Me: This is news?
(What, are we supposed to believe all women, as if they are naturally saints?)
With FSR & XeSS pressure rising, gamers were hoping Jensen would finally relent allowing DLSS go open source.. but he throws crumbs for the peasant non-RTX masses instead, with another spatial upscaling FSR-alternative.
Seems like nerds on forums are up in arms over AMD sponsored PC ports with FSR without DLSS. They think it's nefarious AMD dictates to studios.
This is not how industry works.
PC ports are nearly always, minimal effort. Upscaling? Tick. Move on to bugs, widescreen, 144hz etc
News from Taiwan, recent Apple shifts to next-node has freed up some more 5 & 7N capacity. Bidding was fierce for Q2 allocation, with Mediatek & AMD big winners. Crypto ASIC makers got a small amount of wafers too. They wanted more & willing to pay, but TSMC prioritizes partners.
I don't buy the rumors of major hw bugs in RDNA3 N31.
Put the pieces together, it's a dense design. AMD tried to make smallest 5nm GCD die they could. High density = not going to clk high w/o +W!
They got shy, avoided 450W monster. Major uarch change, driver is sloppy.
With Lisa Su presenting EPYC & constantly comparing it against Xeons, I think we can put to rest this lack of NV comparison in the RDNA3 presentation as anything other than 7900 XTX being inferior to 4090.
It just does not look good in the comparison. Hence, $600 discount.
Denuvo = legit buyers of games are paying for malware. It wastes CPU cycles to confirm that you paid for the product.
When you've already paid for the product, with receipts to prove it.
Do not defend the indefensible.
PSA for everyone gaming on PC. When we tested the performance impact of Denuvo DRM we found that it measurably impacted CPU-limited performance by 7% on a Core i5 8400 (173 vs 184 FPS). That is an older CPU, newer ones would be even less affected.
If you are GPU-limited... (1/2)
I was wrong about this Elon Jet issue. I did not know about the private/anon jet transponder that Elon registered. Its akin to someone making their number or address private, so its not listed in public databases.
What this "kid" did is doxing. Nothing to do with free speech.
The argument used by EU & US politicians regarding China's "overcapacity" is that they are state funding, ie. subsidizing industries which gives them an unfair advantage.
However, the truth is that the US actually subsidizes vastly more, and still failing to compete.
Fantastically revealing exchange with Janet Yellen.
The interviewer asks her: "[Regarding China's] overcapacity in advanced products, electric vehicles, batteries, all of those things. Is it possible we’re just being outcompeted by the Chinese in those fields?"
Yellen replies
Re: AMD upset, leaving TSMC for Samsung 3N. For those interested on this topic, I did some digging around, can safely say this rumor is not true. AMD will still be a major TSMC customer, both current & future nodes, for foreseeable future.
IHV rarely comment directly on benchmark results, and NVIDIA typically don't even consider AMD a competitor at all.
Thus, the fact NV went to disprove AMD's MI300X vs Hopper, but using NV-optimized libraries showing Hopper was faster running NV code. Then AMD debunked that..
300mm2 5nm GCD, yields ~82%. 151/180 good dies. $16K per wafer for AMD. $105 per GCD.
37mm2 MCD on older node + packaging..
Add GDDR6 & PCB.
I think there's heaps of profit margins at MSRP.
This is a very big deal. China will build a 180km long canal from the Mekong river to link Cambodia's capital Phnom Penh directly to a deep-water port in Cambodian waters in the Gulf of Thailand.
Previously Cambodia had to do its shipping exports via Vietnam, following the
The analysis is sound. RDNA3 has efficiency adv on perf/mm2 + perf/w. AMD could use this to create a larger GCD, ~450-500mm2, targeting 450W & destroy Ad102 on raster & compute perf (esp compute).
The reason for absence of bigger GCD, is prioritizing 4/5nm wafers for x86.
I think
@HardwareUnboxed
coverage of frame-gen tech has been more consistent, and accurate from the start.
They were always cautious to not over-sell "fake frames", and separate visual smoothness, vs actual smoothness (gameplay input lag).
Sadly, other reviewers.. 🤔
That rant vs NVIDIA from Linus (of Tech Tips, not Torvalds) is just epic! As a reminder, LTT was one of the few reviewers that called out the BS RTX 3090 "Titan class" marketing from NV on launch day. Far too many just regurgitate Jensen's claims.
More testing FSR3. The new upscaling algo (2.3x?) is very good, in-gameplay, in-motion, I can not tell apart vs DLSS.
The shimmering artifact on grass & water in the title screen is obvious, but in-game it's gone.
Has minor artifact on smoke particles & cloud, pixellation.
Today FSR 3 is making its way into two partner titles. Very proud of the team and what they've achieved. Expect to see a blog on
@GPUOpen
in a few hours that provides some insight on the tech such as optical flow, UI processing, frame pacing and what it could mean to frame times.
NVIDIA is wrecking AMD's GPU line-up with these Super refreshes.
4070S has great perf, *perf/w*, features (DLSS & RT) advantage.
7800XT, AMD's best Navi 32 SKU, can't really compete at above $449, preferably price @ $429. Meaning all the SKUS below it needs a price cut.
New upscaling algo in FSR 3.1, fixing temporal stability & reduced visual artifacts.
As we have said, you do not need "AI or ML" to achieve good quality upscaling results.
This is AMD's gift to everyone, game studios that are PC only, or console & cross-platform, & gamers.
Re: Weird RDNA3 clocks. This is what happens when it's ray tracing.
LDS hammered. VGPRs hammered. The designed data hierarchy causes severe bottlenecks. Despite "100%" load, SIMD downclocks because its essentially stalling.
48% of all new datacenter CPUs installed in March is AMD EPYC (according to Jefferies & Co. analyst).
Gaining momentum.
Zen 4 EPYC is going to extend the perf & perf/w lead by a huge margin.
This is the reception the 3090 should have received. It is even more overpriced, and despite the clever but misleading "Titan-class" marketing, it is nothing more than a gaming Geforce with extra vram.
Is there a logical reason why Pat @ Intel thinks they can win Apple back? From Apple's perspective, they've invested so much into their own custom SOCs, risking x86 issues, to basically solving it, and demonstrated excellence in CPU & GPU.. why go back to Intel?
If this leak on AMD's next-gen premium APU is accurate..
"Hell, it's about time!"
32MB $ is perfect for 40CUs, freed from the constraints of system memory, it would be a screamer at 1080p.
What a shocker, NVIDIA's Portal RTX runs horrible (slideshow) on their older RTX GPUs & AMD GPUs.
"Cinematic FPS" on their new $1600 (or is it $2000) flagship GPU though.
For AMD's 6700XT reveal, I am more interested on FX Super Resolution implementation. On weaker GPUs, I feel that a great upscaling technique matters more, as you can comfortably game at 1440p & higher, at internal 1080p reconstructed. If FxSR is a no show, hard to recommend.
I am against tech, in the sense that CPUs are now >200W.
GPUs are >400W.
Stop it.
This race to the last 5% perf destroys efficiency. Consider most users do not tweak hw, all of these parts in mass, inefficiency adds up to be a drain on our society needlessly.
True. But even if RDNA 3 was efficient, highly innovative and affordable, the majority of consumers will wait for NVIDIA to lower their prices and buy an RTX GPU anyway. 😅
#UE5
foundation is optimized for AMD due to the next-gen console focus (
#UE4
foundation is for Geforce). Give it time for NV's team of engineers to get down & dirty with the PC branch. Optimized shaders & async use on PC branch should help Ampere scale better.
Do not be confused about the 3090 being "Titan class". Because it's merely a gaming GPU, it lacks access to Titan's unique driver optimizations (Quadro branch) that is enabled only on Titan GPUs. This can make a huge difference in some workstation workloads.
AMD acknowledges 7900XT/X high idle power as driver bug:
"High idle power has situationally been observed when using select high resolution and high refresh rate displays."
What. PCI SIG warned NVIDIA about this very flawed design and they just shrugged and decided to ship it anyway?
2.5hrs isn't that long of a gaming session by gamer standards either.
NVIDIA consumer: AMD paid developers not to include NVIDIA proprietary tech!!
It is so incredibly stupid given the fact NV's revenue & profit margins allow them to play this "sponsor" game far better.
The question begs, why isn't NV paying more $ to get exclusive RTX PC ports??
@silkyjohns0n420
If it's a matter of paying off devs, do you realize NVIDIA has much more cashflow to play this game? A game they used to excel at, for those with short memories.
Exclusive & Benchmarks: AMD Quietly Funded A Drop-In CUDA Implementation Built On ROCm: It's Now Open-Source
This allows unmodified CUDA apps to run on Radeon GPUs and in some workloads CUDA on Radeon is faster than native OpenCL or HIP/ROCm backends!
16 threads of the fastest CPUs. Does the game engine come close to saturating all of that CPU resources? I doubt it. Why would you expect to see the effects of driver overhead, if there's ample CPU resources to spare?
It's pure Genius. Jensen waits until all the "8GB is gimped" commentary spreads among the tech community, then releases a 16GB 4060Ti variant for only $100 extra..
"Double the vram, only 20% more cost, the more you buy... the more you save!"
Blinken: "100% of the population in Gaza is at severe levels of acute food insecurity"
Meanwhile still arming Israel, enabling them to bomb more civilians, and maintain their blockade of foreign aid.
If 100% of the population is starving due to Israeli action, that is genocide.
Genuinely stunned that Blinken now admits this, he's basically saying a genocide IS happening:
"According to the most respected measure of these things, 100% of the population in Gaza is at severe levels of acute food insecurity. That's the first time an entire population has
AMD doesn't have to do much "catch up" for RT, gamedevs will optimize RT for RDNA2 in their game engines. :) That's the benefit of a common architecture for consoles + PC.
Was it a common sentiment back then that 2700X would age better than 8700K? How is that logical? 6c/12t with more IPC, more clocks, lower latency vs slower 8c/16t, assuming full utilization (rare), the 8700K will still be slightly ahead given its latency advantage in gaming.
What people still don't get, is if AMD had designed RDNA3 to be excellent both raster & RT & efficient, they would NOT be charging much lower prices than NV.
Thus, there will be little competition on perf/$, little incentive for Jensen to lower price or SKU naming.
You clearly misunderstood the point of the test. They noted that NV's DX12 overhead is real, it occurs in nearly every game they tested. But they noted in DX11 games they tested, NV runs faster. An AVG of all games tested, similar.
PCGH editors: "The fact that the two graphics cards work almost equally fast in the end is due to the variance of our benchmark suite and shows that the balance of power is fairly distributed."
Conclusion from pros. No "this specfic vendor has big problems" bla bla...
This is absurdly awful, power 240-290W on sustained load, 104C on one of the best coolers. A tiny improvement over last gen in performance, still quite behind Zen 3 5800X. Horrendous perf/w...
In advance of the official launch date, we obtained Intel's upcoming Core i7-11700K at retail, and
@IanCutress
tested it against its main competition. Does it Rocket to the Moon, or are we Rocked to our core? Come and read our review.
Tip
@Techmeme
AMD's ROCm maturing to a point where you can take a CUDA only project on github and it will run just fine with competitive performance on AMD MI-accelerators is a pleasant surprise.
The AI/ML industry really needs viable alternative to NVIDIA.
AMD just got an extra 13K 7N wafer allocation at TSMC, from vacancy of another victim (Mediatek) of the Huawei sanctions. Journalist "sources" suggest AMD is devoting all these for PS5/SX production this quarter.
AMD's 7N is hammered. PS5, XSX already ate much of their supply for the past year, moving forward, I think they have underestimated the popularity of Steam Deck. I already know many colleagues wanting to buy one, not just for handheld gaming, but as tiny Linux & Windows PC.
Let me guess, you picked the one Ryzen 5800H laptop design where it throttles the CPU badly, either due to poor thermal or power design. Leave it to the independent tech press, because your propaganda efforts only tarnish the Intel name further.
In a data sharing mood again! Today’s focus is mobile high-performance gaming, and early test data that just hit from the lab is helping me cement what mobile gaming leadership really means. 10th Gen holding strong with 11th Gen H-series coming this year!
Who thought having the regulatory agency intended to keep big pharma honest being >90% funded by big pharma is a good idea?
If you were to ask common folk on the street, most would be able to identify the conflict of interesting instantly..
Bingo. All the undeclared conflicts of interest in one slide.
Remember that Kristine Macartney received $65m in government grants covering only two years. And then set out on a tour of Australia to propagandise to GPs about the COVID vaccine snake oil.
@TonyNikolic10
It's hard to argue the RX6500XT isn't one of the worse GPUs.. It's supposedly a leap from last-gen RX 5500, but ends up slower, massively with gen3 bus which most gamers are on. Overpriced, poor current perf, 4GB vram offers zero longevity. Nothing redeeming.
AMD's increased 7N wafer allocation throughout this year and the beginning of 5N in 2H 2021 (as previously mentioned) puts them at ~>50% of Apple's allocation, or the 2nd biggest customer at TSMC.
"The 3070 has RTX! AMD has nothing close to it!" I nearly facepalmed, but out of respect, I did not go into a rant. That's the power of NV marketing, even intelligent people fall for it.
"And Raja believes that loads of consumers just want something that's affordable and doesn't require a huge power supply and a ton of cooling"
Indeed. Raja is spot on. I would never buy a power hungry GPU just to game. Seems like a terrible waste.
Remember when AMD marketing made excuses for non-support of Zen3 on 300 & 400 chipsets? It was never hw limitation, just profit.
Now that Intel pressure with ADL, they are forced to officially support Z3 & especially Z3D on all AM4 chipsets.
@Petykemano
300 chipsets run Zen3 SKUs, not all MB, due to bios update variation among brands, but its not hw limitation. 400 chipsets better compatibility. Zen3D should work, if your MB brand updates bios.
Re: tech leakers. This episode once again shows us that;
1. AMD runs a very tight ship. They just don't leak.
2. NV is leakier, but only until AIBs get their info.
In future, be very sceptical about leaks regarding AMD tech. On NV, when its closer to the launch, OK.
There's a good reason why Intel never enforced or even suggest to motherboard vendors to stick with stock power settings.
They lose a huge chunk of perf running their CPUs within "stock spec" of the advertised power rating.
Data from
@HardwareUnboxed
shows it clearly.
Intel has long turned a blind eye to high-end motherboards ignoring their CPU power specs. But a reckoning is coming: in light of stability concerns, Intel has put out a request strongly encouraging vendors to actually stick to Intel's stock power settings
AMD's client revenue drop should really be a wake up call to their bean counters. Times have changed, we're now in a global recession. People unwilling or unable to pay premiums for hardware.
Price drop on AM5 & Ryzen 7000, ASAP.
RDNA3 must be priced reasonable.
This is a really bad move from AMD. Even GCN 1 is still very capable. Strong compute perf, flexible async compute scheduler, DX12 & Vulkan capable. No excuse not to support it. You can throw modern games at a 7970 and it will run it quite well @ 1080p med.
AMD this week is retiring driver support for a number of its older graphics products. As of the new 12.6.1 drivers, GCN 1, 2, and 3 hardware - both GPUs and APUs - are no longer supported. Windows 7 support is also getting discontinued
Very detailed micro-benchmarks to analyze architecture changes by
@ChipsandCheese9
RDNA3 massively improved LDS (shared mem) latency, which explains most the ray tracing perf gains.
Also, dual-issue FP32 can shine or flop, due to compiler (driver) VOPD as expected.
You could say the same for 12900K/S.
The majority DIY will buy 5600, 5700X, 12400, 12600K etc. Even these CPUs are overkill for gaming when the avg GPU is 3060.
Flagship CPU & GPUs are not high volume. It's high margins & bragging rights, or in corporate speak, mindshare.