AMD has launched Zen 3-based "Cezanne" Ryzen 5000 mobile APUs at 15W, 35W, 45W, and 45W+ (overclockable Ryzen 9 5980HX and Ryzen 9 5900HX) TDPs. Compared to the previous-generation Zen 2 "Renoir" APUs, Cezanne has 19% higher instructions per clock. L3 cache is doubled (to 16 MB) and accessible by any core. Integrated graphics performance ("Vega"-based) has not improved aside from slightly higher clock speeds. All of the Cezanne models announced so far have 6 or 8 cores, with 2 threads per core. In addition to Cezanne, there are three Zen 2-based "Lucienne" models (5300U, 5500U, and 5700U) that are refreshes/rebadges of 15W Renoir APUs.
Benchmarks have been spotted for a Cezanne desktop APU, the AMD Ryzen 7 5700G, which may not be limited to OEMs like Renoir APUs were.
AMD also launched two OEM-only Zen 3 desktop CPUs: the 8-core AMD Ryzen 7 5800 and 12-core Ryzen 9 5900. These are rated at 65 Watt TDPs, instead of the 105 Watt TDPs of their 'X' counterparts.
AMD's exclusivity deal with Lenovo for Threadripper Pro CPUs (Threadripper with 8 memory channels and 128 PCIe 4.0 lanes) has ended, and manufacturers are preparing new sWRX8 motherboards. 64-core AMD Ryzen Threadripper PRO 3995WX, 32-core 3975WX, and 16-core 3955WX should be available to consumers around March. There's no word yet on Zen 3-based Threadrippers, but AMD will be launching Zen 3-based "Milan" Epyc CPUs later in Q1.
See also: AMD CEO Dr. Lisa Su: Interview on 2021 Demand, Supply, Tariffs, Xilinx, and EPYC
AMD Demos 3rd Gen EPYC Milan 32 Core CPUs, Up To 68% Faster Than Comparable Intel Xeon Gold Platform
AMD's Ryzen Threadripper Pro CPUs & WRX80 Motherboards Coming To Consumer Segment in March 2021 – 64 Cores, 128 PCIe Lanes & 8-Channel Memory
Also at Phoronix and Wccftech.
Related Stories
AMD Announces Ryzen Threadripper Pro: Workstation Parts for OEMs Only
Last year we spotted that AMD was in the market to hire a new lead product manager for a 'workstation division'. This was a categorically different position to the lead PM for high-end desktop, and so we speculated what this actually means. Today, AMD is announcing its first set of workstation products, under the Ryzen Threadripper Pro branding. However, it should be noted that these processors will only be available as part of pre-built systems, and no corresponding consumer motherboards will be made available.
[...] Ryzen Threadripper Pro hardware will mirror single-socket EPYC in its features: eight memory channels up to DDR4-3200, 128 lanes of PCIe 4.0, support for RDIMMs and LRDIMMs, support for secure memory encryption, support for DASH manageability, and operating system image consistency as part of AMD's Pro Business Ready programme.
[...] There is also a small difference in DRAM support – TR Pro supports up to 2 TB, but EPYC supports 4 TB. All of the Ryzen Threadripper Pro processors are single socket only.
The top processor, the 3995WX, will offer all 64-cores. It goes above and beyond the traditional top EPYC 7742 (225 W, 2.25 GHz / 3.4 GHz) and even the 7H12 (280 W, 2.6 GHz / 3.3 GHz), by offering more base frequency at 2.7 GHz and a much higher turbo frequency at 4.2 GHz for 280W TDP. These processors might be taking advantage of the same manufacturing update as provided by the recent Ryzen 3000XT processors in order to drive these higher frequencies.
The other processors are the 32-core 3975WX, 16-core 3955WX, and 12-core 3945WX. Pricing is not available since these are OEM-only parts (being sold to Lenovo first).
Also at Wccftech.
AMD Announces Radeon RX 6700 XT: RDNA2 For 1440p, Coming March 18th For $479
As part of AMD's latest Where Gaming Begins product presentation, the prolific processor designer announced the next member in its Radeon family of video cards, the Radeon RX 6700 XT. Following the tried and true scale-down release strategy that has come to define the GPU industry, the company is preparing its second RDNA2 GPU to further flesh out its lineup of video cards. Set to be released on March 18th, the Radeon RX 6700 XT will be AMD's anchor card for 1440p gaming, succeeding the last-generation RX 5700 XT and giving AMD's product lineup a more wallet-friendly option than their 4K-focused 6800/6900 series cards. The launch for the latest Radeon card will be an all-out affair, with both reference and partner custom cards launching the same day, with prices starting at $479.
[...] Finally, AMD has stated that they're also going to be priming more OEMs than usual with the RX 6700 XT as a third pillar of sorts to better ensure some kind of ongoing retail availability of the card. Despite AMD's best efforts, retail RX 6700 XT cards are virtually guaranteed to sell out within hours (if not sooner) due to how starved the video card market is. So the company wants to at least give customers another avenue towards getting a modern video card by supplying it as part of a complete system, especially as crypto miners seldom buy whole OEM systems.
The RX 6700 XT comes with 12 GB of GDDR6 VRAM and 96 MB of "Infinity Cache". Despite a smaller die, clock speeds are up significantly, which could allow it to perform at around 89% of an RX 6800.
Nvidia's RTX 3060 GPU is "available" at a $329 MSRP. Notably, that GPU also includes 12 GB of VRAM, more than the RTX 3080, 3070, and 3060 Ti.
AMD's Threadripper Pro workstation CPUs with 8 memory channels and 128 PCIe 4.0 lanes are now available, having previously been sold exclusively in Lenovo systems. The 64-core TR 3995WX has a suggested price of $5,490, compared to $3,990 for the 64-core TR 3990X.
Also at Tom's Hardware, The Verge, and Phoronix.
See also: AMD is bringing Smart Access Memory's frame rate boosts to Ryzen 3000 processors
AMD releasing new gaming card amid chip shortage
(Score: 2) by qzm on Saturday January 16 2021, @01:22AM (11 children)
Another round of more CPU and minor GPU tweaks..
Of course more loads are GPU limited than CPU limited at this level..
I can only imagine their reasoning is that improving the GPU significantly would cut into discrete GPU sales (not on laptops so much, but when these move to desktop..)
Still, it is a pity.
(Score: 4, Informative) by takyon on Saturday January 16 2021, @03:13AM
AMD's strategy is to pursue more CPU cores and performance with the high-end mobile chips, since OEMs often pair laptops with discrete mobile GPUs above $600 or so. Which could be either AMD's or Nvidia's:
European price of ASUS ROG Zephyrus G15 with an AMD Ryzen 9 5900HS, NVIDIA GeForce RTX 3080 Max-Q and a 165 Hz display revealed [notebookcheck.net]
AMD will include (presumably faster than Cezanne's 7-8 Vega CUs) RDNA 2 graphics with "Van Gogh" mobile APUs. Those will aim for lower TDPs than 15 Watts (e.g. ~7-10 Watts) and include only 4 CPU cores (Zen 2), instead of the 8 of Renoir/Cezanne. It's never intended to be paired with a discrete GPU. The announcement hasn't been made yet. Maybe they are waiting for DDR5 availability [phoronix.com].
This strategy is going to be disappointing for some, but it makes sense. RDNA 2/3 graphics will land in future successors to Cezanne (not sure which).
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by fakefuck39 on Saturday January 16 2021, @05:43AM (9 children)
Here's the thing though.. These mobile APUs are going into the dirt cheap netbooks - like a step below the Intel netbooks, and AMD mobile sucks on power usage compared to Intel already. Putting in a bigger GPU to drive that tiny screen for web browsing and watching movies is going to increase cost, and decrease battery life. Even to get some work done on a plane - you don't need a good GPU for making a little network diagram in visio or powerpoint deck.
And for the desktop - people are going to put in a separate GPU anywise, who cares what the integrated one does as long as it's enough for chrome and msoffice.
And as I always note with these AMD announcements about beating Xeon. They don't. People who get Xeon don't want the lates fab - they want something proven for many years. That's why Xeon is on 14nm, not 10nm. People who run their production want proven reliability, and they couldn't care less if they get 800 cores of AMD or 1000 cores of Xeon, or if they save $50k on their million dollar VMware farm - risk is worth a lot more than that.
That comparison is bs too. They're comparing a 64 core AMD to a 28 core Xeon. The Xeon goes all the way down to 8 cores, for fastest single-thread, and 56 core Xeons for max vCPU performance. AMD also doesn't have AVX512. So interesting how to get that 68% number they compare 64 cores to 28, when a 56 core comparison would be more rational. Complete and utter lies and marketing BS.
Thankfully, in my 10+ years of presales, after my 10+ years of being a customer, I've never had a single account ask for AMD to be quoted for the datacenter. Hmm, I wonder why. AMD on desktop - awesome deal. On mobile or to run your company on? Gimme a freaking break.
(Score: 0) by Anonymous Coward on Saturday January 16 2021, @05:03PM (3 children)
It was said:
They're comparing a 64 core AMD to a 28 core Xeon. The Xeon goes all the way down to 8 cores, for fastest single-thread, and 56 core Xeons for max vCPU performance. AMD also doesn't have AVX512. So interesting how to get that 68% number they compare 64 cores to 28, when a 56 core comparison would be more rational.
I wasn't there, but it does appear that the Milan preview was based on a comparison of 56 Xeon cores (two Intel® Xeon® Gold 6258R processors) to 64 Milan cores (two AMD EPYC™ Milan 32 core processors) cores.
I don't have any expertise in this area, so please verify for yourself:
https://www.youtube.com/watch?t=2758&v=sxTjfX9v8M0 [youtube.com] (AMD at CES 2021 Milan Preview)
https://www.youtube.com/watch?t=2863&v=sxTjfX9v8M0 [youtube.com] (AMD at CES 2021 Weather Research and Forecast (WRF) Model server "specs" and comparison)
Or course, as Charlie Demerjian points out, AMD does need to do a much better job of disclosure.
From https://www.semiaccurate.com/2021/01/12/amds-ces-keynote-is-a-disclosure-own-goal/: [semiaccurate.com]
[snip] The numbers you will see when the full chips are disclosed will be rather shocking and this teaser would be enough but there is one problem, disclosure. There is literally none. AMD is making a claim and backing it up with nothing, literally no fine print. [snip]
(Score: 2) by fakefuck39 on Saturday January 16 2021, @05:26PM (2 children)
hold up - the link in the article about the 68%
https://wccftech.com/amd-demos-3rd-gen-epyc-milan-32-core-cpus-up-to-68-faster-than-intel-xeon/ [wccftech.com]
despite the title, the article itself says this:
So summing everything up for EPYC Milan, we are looking at the following main features:
Advanced 7nm Zen 3 cores (64 core / 128 thread)
Pin Compatible With SP3 Socket
120W-280W TDP SKUs
PCIe 4.0 Support
DDR4 Memory Support
Launch in Q1 2021
so they're not making that claim about the 32core version, they're making it about a 64 core version. now, I'll fully admit there's a chance I'm reading that wrong.
In addition, note they're comparing a Xeon that's one year old, to an AMD that was just released. They are not comparing the 7nm AMD to the new 10nm Xeons. And let's keep in mind that AMD's 7nm and Intel's 10nm aren't measuring the same thing, so no, AMD is not on a smaller fab.
(Score: 0) by Anonymous Coward on Saturday January 16 2021, @06:40PM (1 child)
I think that the Wccftech article may have switched channels by talking about the overall Milan family instead of the specific part used in the comparison performed by AMD as part of their CES 2021 presentation.
The "AMD's 3rd Gen EPYC Milan Server CPUs Demoed, Up To 68% Faster Than Competing Intel Xeon Gold Cascade Lake CPUs" section of the Wccftech article seems to correspond to the Youtube video of the AMD CES 2021 presentation previously mentioned.
I didn't take the time to view the Intel CES 2021 presentation, but https://www.crn.com/news/components-peripherals/intel-starts-shipping-ice-lake-xeon-cpus-promises-more-cores [crn.com] suggests that the 10nm Xeon parts just recently started shipping:
Intel has started shipping production versions of its third-generation Xeon Scalable processors, code-named Ice Lake, saying that the new chips will bring “significant increases in core count,” among other things.
The Santa Clara, Calif.-based company made the disclosure during its virtual CES 2021 keynote and said production of the processors — the first to use the chipmaker’s 10-nanometer process — are slated to ramp up over the course of the first quarter.
If these are the new 10nm Xeons, perhaps AMD didn't have access to a Xeon 10nm based system when the comparison was done,
I too am looking forward to an (I assume) Ice Lake vs. Milan comparison and I think that we can all agree that a "x"-nm vs. "y"-nm comparison is meaningless just as comparisons based solely on clock speed are worthless between different families of devices.
Even though I'm not involved with any of these server class processors, if you have a link to a vendor currently shipping systems based on the Xeon 10nm parts (or the Milan parts) that I could order today, I'd be interested in learning more about their capabilities. Thanks!
(Score: 2) by fakefuck39 on Saturday January 16 2021, @07:53PM
ice lake isn't available for purchase yet - it will be this quarter though. right now, I believe Oracle got the first batch, and their X9 offering also supposedly available this quarter uses ice lake.
I don't think anyone can do their own benchmarks - there's only what intel released.
https://www.hpcwire.com/2020/11/17/intel-teases-ice-lake-sp-shows-competitive-benchmarking/ [hpcwire.com]
vendors shipping stuff, like "R640 with ice lake" are always way behind the large cloud-type providers. I checked dell's DSA, and you can only order a 14nm xeon. But, you can't order Milan either. It's all going to go at the same time.
I honestly have no idea at this point if they compared 32 or 64 core to 28. My point was, it's still a last-gen vs current gen comparison. For example, Intel also claims the new 32 core ice lake beats a 64 core AMD gen2 by 30%.
(Score: 0) by Anonymous Coward on Saturday January 16 2021, @06:09PM (1 child)
Yeah, ok Windows-using Intel shill. You probably deal with other Windows-using Suited Whores. AMD is plenty reliable and their CPUs have always been more secure than Intel.
(Score: 1, Flamebait) by fakefuck39 on Saturday January 16 2021, @06:25PM
"plenty reliable" is not ok for a business. when you're storing a number in a database, and that computation being wrong can cost you a million dollars, you don't want "plenty reliable." you fully reliable. In fact, many banks and things like air traffic control run on mainframes where two cores do the same computation and compare results.
what this is telling me, is you've never had a real job doing something that's important, where an error means you die from surgery due to mismatched drug interactions, you deal with large amounts of money, or are in charge of systems where an error is not acceptable.
welcome to the world, mr pc gamer screwing around with a linux box in his room. here, the only thing that makes something proven reliable, is years of operation. this is why intel runs any new fab on desktop first, then transitions it to the datacenter after it's proven - by time. AMD can only compete with this last-gen intel fab by using the latest and greatest and unproven.
Intel is not less secure than AMD. Bugs were found, some of them on AMD as well, and were immediately patched before an exploit was in the wild. At a performance hit - which is irrelevant - you just get some more CPUs - that's dirt cheap.
You are comparing a performance hit from a patched security exploit to the risk of "5+5=8." This is why you've never had a real job doing anything important.
(Score: 1, Insightful) by Anonymous Coward on Saturday January 16 2021, @11:13PM (2 children)
1. AMD integrated graphics typically beat Intel integrated graphics by a good margin. A buyer that wants to do seriously graphics-intensive gaming like Doom Eternal or Call of Duty Whatever will need a discrete GPU, but an AMD integrated GPU will handle Minecraft, League of Legends, Starcraft 2, and similar casual games and Intel integrated GPUs may not.
2. On server market share, AMD EPYC parts seems to have made progress because in July 2020 announced that they had over 10% of new sales for the first half of 2020. That's the first time AMD has had a market share for servers that large in decades.
3. I suspect, but am too lazy to check, that the comparison was based on pricing. Intel has been milking their advantages for a long time, and even if their parts are flat out better - not guaranteed - if you can get a 90% as good part for 60% of the price, you'll buy it.
4. AVX512 isn't applicable in all algorithms. There are some workloads in which its presence is a crucial factor, and Intel wins at any price. But you can't declare all servers everywhere must have it.
Your post sounds like marketing from an Intel employee. I have three Intel CPUs operating right now in the room I'm sitting, and just one AMD CPU. Intel has made great stuff and had the clear performance edge for more than ten years. But AMD is competitive now, and it's dishonest to pretend we're still in 2015.
(Score: 2) by takyon on Saturday January 16 2021, @11:44PM (1 child)
Now that Intel has doubled performance or whatever it is with Xe graphics in Tiger Lake, you can expect AMD and Intel to trade the top spot for a while. Except that AMD's Cezanne made essentially no improvement in integrated graphics (explained in another comment), and AMD will put better graphics in the 5-10 Watt range with Van Gogh (chips competing with Tiger Lake-Y and Amber Lake). AMD will switch from Vega to RDNA 2 in Rembrandt [wccftech.com] (successor to Renoir and Cezanne), so maybe that will outperform Intel's iGPU. But I bet it won't have many compute units (graphics cores) and Intel will leapfrog that too. Competition is a good thing.
AVX-512 is overhyped, but AMD is expected to include support for it by around Zen 4 (maybe only in Epyc).
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 0) by Anonymous Coward on Tuesday January 19 2021, @03:55AM
If you really need to do zillions of AVX stuff that aren't latency sensitive, you should be doing it on GPUs unless someone is paying you enough to do it on Intel.
(Score: 0) by Anonymous Coward on Saturday January 16 2021, @01:48AM (1 child)
My laptop is core-2-duo Lanovo, which, surprisingly, remains quite competent. Guess single-thread performance still rules.
Convince me to buy a new one - I am quite partial to AMD chips.
(Score: 4, Informative) by takyon on Saturday January 16 2021, @03:20AM
Are you talking about a Penryn mobile CPU from around 2008 to 2009, like these [wikipedia.org]? What's the exact CPU model?
If you are, pretty much everything on the market will crush it in single-thread performance, like a Celeron N4000 or AMD A6-9220C found in ~$100 laptops. And they'll do it while using 20% the power, with better GPU performance, support for hardware acceleration of newer video codecs, etc.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 3, Interesting) by fakefuck39 on Saturday January 16 2021, @04:31AM (9 children)
These CPU caches are getting huge, but we're not seeing big growth in the clock speed. From what I remember, in the early days of desktop CPUs, newer models decreased CPU cache size because a larger cache actually slowed down the CPU. I'm guessing this somehow isn't an issue anymore anymore, maybe because it's an L3 cache, not L1 or L2? If there's a EE on here, clue me in please.
(Score: 3, Interesting) by takyon on Saturday January 16 2021, @05:27AM (1 child)
We're moving to 8+ cores, but single-thread workloads are still more important. In the Zen 3 unified core complex design, 1 of the 8 cores can use up all of the cache if needed. Physically larger L3 cache can increase latency, but it can be worth it if DRAM doesn't need to be accessed as often. There's some details about how the caches in Zen 3 were tweaked here:
https://www.anandtech.com/show/16214/amd-zen-3-ryzen-deep-dive-review-5950x-5900x-5800x-and-5700x-tested/4 [anandtech.com]
L3 cache dominates the die space now. An individual Zen 2/3 core is tiny. Newer process nodes don't allow SRAM to be scaled down much. For example, TSMC's N3 node [techradar.com] will have 1.7x the logic density but only 1.2x the SRAM density of N5.
Going forward, 3D stacking will be used to increase performance. Zen 4 will probably include L4 cache stacked on the I/O die (for desktop chips). I assume there would not be a huge latency penalty since a stacked cache could be bypassed or turned off entirely. There's a possibility of using 3D SRAM [anandtech.com] to increase L3 cache sizes further without increasing the 2D die area and latency.
As for clock speeds, they might be able to go up if gigabytes of memory can be put directly onto/into the CPU die, shortening the distance between main memory and cores to almost nothing. Maybe put Intel Optane or something like that in the DIMM slots instead of DRAM.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by fakefuck39 on Saturday January 16 2021, @05:49AM
So, I've actually quoted out quite a few Optane HCI farms recently, and no one ever goes with it. Now, instead of NVME for storage cache - everyone buys it. But putting in one of those RAM+Flash Optanes kills your slots. Very little actual RAM on those to use it instead of memory, and no one buys them. This is mostly because the OS does a better job of moving between Optane and DRAM vs what Optane is able to do internally. Which makes sense, because the OS knows what to page in before it's used much better than the Card, which isn't aware of threads and processes like the kernel.
Most Optane is actually going into storage arrays. It's not what Intel intended it for, but it's where vendors are using it.
(Score: 2) by shortscreen on Saturday January 16 2021, @08:09PM (6 children)
There are two time periods that come to mind when caches got smaller. The first was around 2000 when the L2 moved from the motherboard or cartridge (remember slot 1?) to the CPU die. Most socket 7 motherboards had 512KB of L2, but socket 370 boards had none because by that point SDRAM had the same bandwidth and only somewhat worse latency. So Celerons and Coppermine Pentium IIIs were including 128KB/256KB of L2 on the die instead which ran at CPU speed instead of FSB speed.
On the cartridge CPUs which had L2 as seperate memory chips outside of the CPU die (the first Pentium IIIs and a bunch of Athlons), the L2 couldn't run at the full speed of the CPU and they divided the clock down. 550MHz P3 ran the L2 cache at 275MHz for instance.
The other thing that happened was when the Core 2 lineup was replaced by Nehalem and instead of 6MB shared L2 there was only 256KB for each core. This did of course come with a latency improvement, as just because something is on the die doesn't mean it is instantaneous. Latency for the L2 had crept up from 10 cycles on a Pentium M to 15 on Core 2 CPUs (with no contention between cores). Nehalem got it back down to 11 cycles.
It's unlikely that cache speed would hold back the CPU frequency, since that would mean that the engineers goofed and made the timing too aggressive.
(Score: 2) by fakefuck39 on Saturday January 16 2021, @11:03PM (5 children)
I'm talking like back in 8086 times. I'm old. I think what was going on is to get higher clock out of the 8086 they had to reduce the CPU cache.
(Score: 2) by shortscreen on Sunday January 17 2021, @07:39AM (4 children)
8086 didn't have any cache. The x86 line didn't have on-chip cache until the 486, so it's hard to know what you might be referring to.
(Score: 2) by fakefuck39 on Sunday January 17 2021, @10:56AM (3 children)
"on-chip" is something you put in all on your own buddy. IBM had CPU in the 70s, the 68k had cache in the 80s, the 386 had cache too I believe, just not on the CPU.
the larger the cache, the more clock cycles it takes to retrieve the data and the higher its latency. there was in fact some CPU from *around* the time of the 8086 that I remember where the next version of the CPU made by the manufacturer decreased cache size to achieve a higher clock. if you are confused about what I am talking about, google is your friend - very common question. https://www.quora.com/Can-having-too-large-of-a-CPU-cache-reduce-performance [quora.com]
(Score: 2) by shortscreen on Sunday January 17 2021, @05:57PM (2 children)
Stuff that is outside the CPU is not part of it. I guarantee you that no CPU vendor ever released a new version of a CPU just so that whoever was making boards could install less external cache. But if you've decided that you already know everything then there is no need for further discussion.
(Score: 2) by fakefuck39 on Sunday January 17 2021, @06:30PM
>Stuff that is outside the CPU is not part of it
what's your point? the 386 had a cache for the CPU on the motherboard. Yes, it was not a part of the CPU. congratulations, you've said the sky is blue.
>guarantee you that no CPU vendor ever released a new version of a CPU just so that whoever was making boards could install less external cache
that's cool. not sure what that has to do with what I was talking about. issues with reading comprehension? i may not know everything, but I at least can read. hint: the text in front of you is what people are saying. not a voice you hear in your head.
what I said was: the CPU vendor released a new version of the CPU with less cache, because it increased speed, due to large cache adding cycles and latency to accessing that cache.
the only discussion here that we're having is you talking to some voices in your head that are making stuff up for you to argue with, and me laughing at the guy with a mental disability. hahhahahahah!
(Score: 2) by fakefuck39 on Sunday January 17 2021, @07:15PM
you know what, I just feel like making you feel even dumber, so I googled a few CPUs for you, and I'm gonna be that guy today who replies twice to the same comment. think of it as you shitting yourself and people laughing at you, and then you slipping and falling and getting covered in your own shit, and people laughing at you again.
POWER4 1.9GHz: L1 cache 64+32 kB/core
POWER5 2.3GHz: L1 cache 32+32 KB/core
IBM stepped down the cache to increase clock speed. And oh, you dummy, both of those had their L3 cache off-chip, on the motherboard. You dummy.
this is pretty recent. and this has happened several times on several CPU lines, including some old one I remember from the 1980s.
have I mentioned you're a dummy, dummy? well, you're a dummy.
(Score: 2) by inertnet on Saturday January 16 2021, @12:48PM (4 children)
I'm still waiting for the 5900X, which was released on November 5, but there's still no information when it will actually be available. The same goes for the new GeForce 30 series. Is there a good explanation for this, or is it normal for AMD to release stuff that you can't buy for half a year?
(Score: 3, Informative) by takyon on Saturday January 16 2021, @01:28PM (3 children)
There is unprecedented demand, a pandemic, and scalper bots buying tech products.
https://www.tomshardware.com/news/where-and-how-to-buy-ryzen-5-5600x-7-5800x-9-5900x-9-5950x [tomshardware.com]
5900X (1 unit) is in stock in Duluth, GA
5950X is in stock in Marietta, GA and Overland Park, KS
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by inertnet on Saturday January 16 2021, @02:20PM (2 children)
I guess they're probably available in the USA, but over here they're still too expensive and very hard to come by.
Current 9500X availability in the Netherlands from several vendors:
604 euro ($731) most likely an opportunist claim to get people to order, but has nothing in stock.
812 euro ($983) claims their supplier has stock but I don't believe that. Or at least not as many as are really needed.
999 euro ($1209) claims to have them in stock. For that price it's more plausible that it's grey import. You'd have to really want one for double the USA price.
One trustworthy vendor just says that they took them off their sales list because they can't promise when they'll have them again. For 583 euro ($705) but at the moment they don't show prices. Just before Christmas they had a backlog of probably over a 1000 for various new Ryzen processors. I'm sure that list of people waiting for their hardware is still very long, although some will have cancelled their order, or changed it to a cheaper, readily available AM4 processor for the moment.
(Score: 2) by takyon on Saturday January 16 2021, @11:47PM (1 child)
U.S. tends to get more supply and cheaper tech products in any normal year. This year? You'll have endless problems. Ryzen 5000 clearly exists in Europe, though. [wccftech.com]
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by inertnet on Sunday January 17 2021, @12:25PM
That article is mostly about the Zen 2 processors. I just checked German shops because I might simply order a 5900X there, but the situation looks the same. Some shops list them for over 800 euro, some even claim to have them in stock but most don't.
I checked some French shops as well, it's the same there. They're listed for 700+ euro, but they show "Date de livraison inconnue" (delivery date unknown) or "rupture de stock" (sold out).
So I'll just have to wait, my 10 year old system around a Core i7 2600K is still working fine. My complaint really is that a product is put on the market but it's impossible to buy it for the next 3 to 6 months.