01:32 | Der8auer Nano-Structures Imaging
Colleague YouTuber Der8auer posted a three-part series that we wanted to draw attention to. Roman’s series uses a scanning electron microscope to view AMD Ryzen silicon nanostructures at more than 100,000 magnifications. Roman and the engineer on site examine the images in great detail and talk about the structures that make up the CPU: Among other aspects of silicon, Roman shows how small the composition of silicon is and zooms in on portions and interconnects within the CPU.
It is a very educational series and it is well produced. We recommend that you visit Der8auer on YouTube to see the series, which is available in English and German.
03:04 | Researchers exploring embedded liquid cooling
Even for all of its modern RGB LEDs, there is still some inherent inefficiency in liquid cooling a chip or component, namely in transferring the heat load into the cold plate and subsequently getting it into the liquid itself.
However, new research underway in Switzerland appears to show some promising progress in liquid cooling. Mainly in bringing liquid cooling directly to the chip itself, via cooling channels directly on top of the chip package. To demonstrate this on a semiconductor, the researchers used gallium nitride (GaN) on top of a silicon wafer to build a simple power conversion chip.
To build the built-in cooling system, the slits are cut through the GaN, reaching the silicon below. Then, these slits are etched at the silicon level to widen them into channels. Then, the cuts made through the GaN are then sealed with copper. Water flows through channels beneath the silicon via feeds and sinks.
The researchers then packaged this cooling system using a double-sided adhesive with laser-cut channels to accommodate the cooling of the chip, and the entire package was attached to a PCB with connections for both power and water supply. Overall, the researchers noted that the design is capable of handling thermal loads of up to 1
05:14 | AMD: Big Navi and Zen 3 in October
AMD finally made it clear when we could see Zen 3 and RDNA 2 (Big Navi) parts by announcing a future announcement – that’s the current trend. NVIDIA has done this for 21 consecutive days on RTX and ASUS has done it after, and now it’s AMD’s turn to pre-announce an announcement, with the announcement of RDNA2 landing suspiciously a day after its earnings call.
AMD announced that it will reveal the Zen 3-based Ryzen 4000 “Vermeer” CPUs and RDNA 2-based GPUs on October 8 and 28, respectively. These types of announcements typically precede product availability by at least a week or two, although AMD may be making availability at the time of the announcement, depending on how confident it is. If this is announced first and is available later, AMD is expected to hit the Black Friday purchase window to maximize impact, but it is also splitting silicon by console and CPU, as much of what AMD is currently doing is on the same wafer supply. AMD will have to figure out how to divide the limited offering between its products: consoles will have purchase orders that probably can’t be changed on the fly, especially with expected high volume, but they will provide some cash flow; CPUs, if using shared wafer supply, could potentially have a higher margin per wafer serving, which will be a hard balance if inventory is required on Ryzen CPUs. Dedicated GPUs seem the easiest to scale if provision is needed for a higher margin elsewhere, especially with larger dies for GPUs that will perform less than the smaller CPU silicon.
Previously, there was also a Easter egg in Fortnite suggesting that the new GPUs would be called the RX 6000 series. The Easter egg is much more interesting as a pre-announcement tease, and we think the product teasers are done well.
Of course, we’ll have to wait and see exactly how AMD positions its new GPUs, but there are plenty of precedents for last-minute price cuts between AMD and Nvidia. AMD took a similar stunt with its RX 5000 series as a countermeasure to try and deflate Nvidia’s RTX Super series. Likewise, Nvidia would end up lowering the price of its RTX 2060 in response to AMD’s aggressive pricing in the segment.
Radeon RX 6000 Series: https://twitter.com/Radeon/status/1303726639013036033
Zen 3 / Ryzen 4000: https://twitter.com/AMDRyzen/status/1303732899506774022
10:35 | Xbox Series X & S pricing, AMD SOC promising
After a leak that demanded a $ 299 price tag for the then unconfirmed Xbox Series S, Microsoft went out and confirmed both the console and the price. Microsoft also announced that the Xbox Series X will be priced at $ 499 and that both consoles will be available on November 10th. In addition to the new consoles, Microsoft is launching a renewed Game Pass Ultimate subscription that comes with an EA Play subscription at no additional cost, as well as financing options for purchasing the new consoles.
Unsurprisingly, the biggest changes come in the form of a smaller GPU, less GDDR6 memory, and less solid-state memory. The S-series uses a 7nm Zen 2 8C SoC, but is compensated for by a slightly lower CPU clock speed of 3.6 GHz. Meanwhile, the GPU runs at 1.56 GHz on 20 CUs. This is compared to the 52 CUs clocked at 1.82 GHz for the X series.
Additionally, the S series will come with 10GB of GDDR6, split as 8GB at 224GB / s and 2GB at 56GB / s. For storage, a 512GB PCIe 4.0 NVMe SSD and a 1TB Microsoft proprietary expansion card slot are available.
Interestingly, there are implications here with Xcloud: 20 CU versus 52 CU configurations should be widely varied in terms of capacity, but if Microsoft intends to effectively deliver a much more powerful Chromecast-style device, it could connect to Xcloud for gaming. more challenging. This is doubly true when you consider the new push for Game Pass subscriptions.
The Xbox Series S is aiming for 4 TFLOP processing, compared to the TFLOP score of 12.15 for the Xbox Series X. This is largely a useless number, but both Microsoft and Sony insist on pushing, so here we are. The Xbox Series S will primarily target 1440p / 120FPS gaming and appears to be positioned as the successor to the Xbox One S.
While all eyes are on Sony now, there has already been a rumor suggesting that its PS5 hardware was priced a little more exorbitantly than Microsoft’s. According to rumors, Sony is likely looking to match prices to Microsoft in response to the price of the Xbox Series X and Series S.
AMD’s console GPUs also look really promising for this generation, and we’re excited to test the new consoles and see how they do. Remember that, with console GPUs, there is always a big inherent advantage as developers only have one (or two) standard sets of hardware to build for. If everyone has that hardware, it’s theoretically easier to optimize it. That’s why you see consoles running games that look more impressive than you might find equivalent performance in a PC application.
15:21 | Western Digital’s “5400 RPM Class” designation
It seems that Western Digital wants to attract even more ire from consumers. This time around, it looks like Western Digital is muddying the water on spindle speed, as it refers to its WD Red line of HDD NAS.
According to Ars Technica, this problem actually surfaced about a year ago via the German hardware site Hardwareluxx.de, but only recently got it turned on in earnest thanks to a post on r / DataHoarder. The crux of the problem is that Wesern Digital identifies some WD Red hard drives as “5400 RPM class” despite evidence that the drives actually run at 7200 RPM. Furthermore, the 5400 RPM class designation does not seem to be limited only to the WD Reds, but also to some models of the My Elements and My Book lines. However, the 5400 RPM class issues only seem to affect 8TB models.
Again, this problem is not entirely new; another reddit user, u / Amaroko, had previously published his findings with the WD Elements WD80EMAZ and WD My Book WD80EZAZ models. Not only are HDDs advertised as 5400 RPM models, but the firmware also reports a spindle speed of 5400 RPM. User results contradict this.
“Last year I personally purchased some of these units (both WD80EMAZ-00WJTA0 and WD80EZAZ-11TDBA0). When someone on a German forum told me it was actually 7200 rpm, I didn’t even believe it at first. But then I took a microphone and started measuring. You see, 7200 rpm equals 120 rotations per second, which should manifest as a base frequency of 120 Hz. More perhaps integer multiples of that frequency (240 Hz, 360 Hz, 480 Hz, etc.), so called tones / harmonics. 5400 rpm instead corresponds to 90 Hz. Therefore, it should be easy to tell the difference between the two, “writes u / Amaroko.
Amoroko came to this conclusion after a quick frequency analysis on the hard drives, whereby the HDDs were placed on a cardboard box with a Yeti microphone placed on top of them, and recording the disks on. The spectral views were then captured with Adobe Audition, which showed a base frequency of 120Hz.
There have also been reports of affected units having received complaints regarding noise and temperature, which likely stem from units that are not actually 5400 RPM units. One of the main reasons some users and applications prefer 5400 RPM drives is the noise and lower operating temperatures.
Ars Technica contacted Western Digital and received the following statement:
“For select products, Western Digital has been publishing RPM speeds within a” class “or” performance class “for many years rather than publishing specific spindle speeds. In addition, we optimize the selected hard drive platforms and their characteristics. of HDDs to create different variants of those platforms to meet different market or application needs. In this way, we are able to leverage our economies of scale and pass those savings on to our customers. As with any Western Digital product, the details of our product, which includes power, acoustics and performance (data transfer rates), are tested to meet the specifications given in the product data sheet and marketing material. ”
18:46 | RTX 3090/3080 Waterblock surface
When Nvidia showed off its new RTX 3000 series cards, it boasted a new continuous-flow cooling design from Nvidia aimed at letting more air through the card and subsequently more hot air from it. To facilitate this, FE boards use a truncated custom PCB. Nvidia has a more traditional reference design that will go to AIB partners, and companies like EKWB are already working on water blocks based on this design.
However, it seems that Bykski has already produced waterblocks compatible with the PCB present in the FE boards, in particular that of the RTX 3090 and RTX 3080. The blocks are listed on ezmodding and are available in different finishes (acrylic, acetal, etc.) and are referred to as full coverage water blocks, which means the blocks should provide coverage not only for the GPU, but also for VRAM and VRM.
These blocks appear to be available for pre-order, and ezmodding also has placeholders for upcoming AIB models.
21:26 | Cryorig returns with the R5 CPU cooler and Crona RGB fans
We have long been talking about Cryorig’s potential return to the US market. Most recently, the company anticipated an upcoming product with the word “5OON”. And while we had a few laughs at the marketing, what Cyrorig was actually making fun of was the now official Cryorig R5 CPU cooler, which will serve as the latest addition to Cryorig’s R series of coolers.
The R5 cooler is a CPU tower cooler with two fans and has some interesting features and claims, such as a “quick” mounting system that supposedly allows for one-handed installation. The R5 uses 6 heat pipes and comes with a TDP of 200 W. Cyrorig didn’t mention many technical details on how the mounting system works. Cyrorig lists anterior fin density of 48 and anterior fin measurements of T = 0.4mm, spacing = 2mm.
Cyrorig has also announced his new Crona RGB fans. That is Crona, not to be confused with a certain demonetizing human malware epidemic that may or may not have a similar wording. In any case, the Crona RGB fans are 120mm fans that feature a circular RGB LED design and use Cryorig’s HPLN (High Precision Low Noise) bearing design.
Price and availability are TBA.
23:39 | Binning RTX 3090/3080
A new report from Igor’s Lab suggests that the quality of the upcoming RTX 3000 chips may be a bit scattered. If it wasn’t already clear, Igor has a clear boost on the GPU industry and market, as he recently nailed the launch of the RTX 3000. So, this information is likely highly reliable.
Going into specifics, both the upcoming RTX 3090 and 3080 will use the GA102 die, which in turn is based on a custom adaptation of Samsung’s 8nm process, dubbed 8N Nvidia. Regardless of node maturity, the GA102 is large and complex, and as always, it will take some time to fine-tune production to extract the best chips.
That said, Igor’s Lab reports some binning statistics for an early production of GA102-based chips for the RTX 3080. Sorted in bin 0 (medium), bin 1 (good) and bin 2 (really good), the early chips look like lean towards bin 1, representing an average of 60% chips. Another 30% make up container 0 and 10% of the GA102 chips are container 2.
Although Igor’s sources claim they have not received nearly the same number of chips for the RTX 3090 models, it has been reported that the overall containers will likely be similar. This data suggests that overall, Samsung’s custom 8nm process for Nvidia is very healthy, as most users have a good chance of getting a bin 1 chip. Igor is citing secret sources from vendors and council partners for these numbers, but again, Igor’s track record is very good.
Source: https://www.igorslab.de/en/chip-is-not-equal-to-chip-first-information-on-the-possibility-to-binning-and-dispersal-at-the-force- rtx-3080-andrtx-3090 /
Editorial: Eric Hamilton
Presenter, additional reports: Steve Burke
Video: Keegan Gallick