AMD Hype!
-
-
It's all just hype until people who aren't AMD employees get their hands on it.
I'm still waiting for when the RTX 3070 is released to reviewers for the same reason.
-
@loopback0 except unlike NVidia, we've got actual benchmark numbers instead of just "% better than previous".
Personally, I'm disappointed by the mid-range lineup. I hoped for an 8-core R5, instead we've got price bumps. There's $300 6-core, and $450 8-core, and nothing in between. Guess I'll be getting 3700X after all, assuming some good Black Friday deal.
-
except unlike NVidia, we've got actual benchmark numbers instead of just "% better than previous".
Sure, and if those are backed up by independent benchmarks then it's all good in the hood.
-
-
There's $300 6-core, and $450 8-core, and nothing in between.
A 7-core chip would be ridiculous.
-
There's $300 6-core, and $450 8-core, and nothing in between.
A 7-core chip would be ridiculous.
So are 1000-byte kilobytes, but yet here we are.
-
@dkf Not quite. Unorthodox, yes. There are 8 cores on one chiplet. I believe any one (and therefore any number) of them can be disabled during binning. There were 3-core AMD CPUs a long time ago.
-
@Applied-Mediocrity said in AMD Hype!:
There are 8 cores on one chiplet.
Given the way the geometry of these things works, it's more likely 4 cores per module and two of those on the chip. The access to the on-chip network works better that way.
-
There's $300 6-core, and $450 8-core, and nothing in between.
A 7-core chip would be ridiculous.
So are 1000-byte kilobytes, but yet here we are.
That one I actually am fine with. After all, the various prefixes (Mega, Giga, kilo, nano, ...) are all using the 10^n scheme where n is a negative or positive multiple of three. I.e. Base 10 when combined with all of the other units.
-
all using the 10^n scheme where n is a negative or positive multiple of three
What about “centi-” and “hect-”?
-
all using the 10^n scheme where n is a negative or positive multiple of three
What about “centi-” and “hect-”?
Those are only used by inferior people with lesser intellect
-
@dkf I remember reading that Zen 3 is to have 8 per module, to reduce latency. But this was before the actual announcement. Because I'm
to look into it now, I'll afford that you may be right.
-
@Applied-Mediocrity said in AMD Hype!:
I'll afford that you may be right.
It's just so much cheaper to do it that way: design one core that's about square, then reflect in X and Y axes to make four cores that you connect up to a common communications hub that talks to the on-chip network. The resulting square can then be stamped out as many times as you want, along with whatever common on-chip peripherals are desired (L2 caches, etc.) You see an equivalent pattern in lots of other SoC chips, and it's a very good way of keeping design costs manageable.
-
@Applied-Mediocrity said in AMD Hype!:
@dkf I remember reading that Zen 3 is to have 8 per module, to reduce latency. But this was before the actual announcement. Because I'm
to look into it now, I'll afford that you may be right.I thought it was 6 or 8 per chiplet too, but was only half-listening to a video that wasn't even the original AMD announcement.
EDIT:
The processors are still chiplet-based, with one chiplet having either six or eight cores. Ryzen 5 and Ryzen 7 will have one chiplet, while Ryzen 9 will have two chiplets – the easy way to identify this is through the amount of L3 cache each processor has.
-
@dkf Intel designs aren't exactly square either
(original: https://en.wikichip.org/wiki/File:coffee_lake_die_(octa_core)_(annotated).png )
-
@Applied-Mediocrity Interesting; they're just rotating the design. That's… a really boring and low-insight method of high-density layout, and won't scale well. The quad-based designs are how the super-high core count CPUs are developed.
-
AMD Is Said in Talks to Buy Rival Xilinx in $30 Billion Deal
(Bloomberg) -- Advanced Micro Devices Inc. is in advanced discussions to buy Xilinx Inc. in a takeover that could be valued at $30 billion, according to people familiar with the matter.The deal could come together as early as next week, though things remain in flux, the people said, asking not to...
-
@dcon It seems odd to me to call AMD and Xilinx rivals. AMD makes CPUs and GPUs. Xilinx makes FPGAs. Really different things for really different purposes and different markets.
For those of you who aren't into hardware at the chip level, FPGAs are chips that are programmable at the hardware level — essentially huge arrays of AND and OR gates and switches between them that can be used to connect them together to perform any desired logic function. A big RAM holds the on/off state of all the connections between the logic that determines what the chip does. (Usually, especially in higher-end FPGAs, there is also some dedicated logic — embedded CPU core, things like USB, Ethernet and DRAM interfaces, maybe A/D and D/A, etc. — that can be connected to the programmable logic.) When you turn the power on, the RAM gets loaded with the interconnect switch data, the embedded CPU instruction store gets loaded with its programming, and the chip starts doing whatever you have programmed it to do (which, like any programming, may or may not be what you intended it to do).
FPGAs are used for prototyping and for low-volume applications where you don't have the capital or it's not worth investing the millions of dollars required to manufacture custom chips.
I can't address AMD's business case for acquiring Xilinx, but I'd call Xilinx's product line complementary to AMD's (if that), but certainly not a rival.
-
@HardwareGeek it's kinda sad that you feel the need to explain what FPGA is on this forum.
As for the Xilinx takeover - IMO it makes more sense than Nvidia buying ARM.
-
@HardwareGeek it's kinda sad that you feel the need to explain what FPGA is on this forum
Why? He feels the need to explain everything.
-
@boomzilla said in AMD Hype!:
@HardwareGeek it's kinda sad that you feel the need to explain what FPGA is on this forum
Why? He feels the need to explain everything.
EvenEspecially when someone else is wrong.
-
it's kinda sad that you feel the need to explain what FPGA is on this forum.
I didn't know what one was. I suspect I'm not the only one.
-
@loopback0 When some of y'all start talking about, say, async programming, promises, await, etc., or some arcane aspects of why some DB is good or bad for a particular kind of data, or metaprogramming, I have only a general, somewhat vague idea of what you're talking about. Um hm, I know some o' them words.
I don't think it's either wrong or insulting that some of you may have a similar lack of knowledge about hardware internals. Unless you are programming embedded systems or device drivers, you really don't need to know much about it, and I'd expect most programmers wouldn't necessarily have learned it (or remember it if they did learn it) unless they're working in a niche where they need it.
-
@loopback0 oh great. What next? Some millenial comes in and whines that he doesn't know what LPCTSTR stands for?
Remember how programmers used to know things?
-
What next? Some millenial comes in and whines that he doesn't know what LPCTSTR stands for?
I'll not whine about it.
Remember how programmers used to know things?
Remember how not everyone here is a programmer? Or how not everyone works in the same industry or area of IT?
Even if I was I could guarantee I'd have never encountered an FPGA or the need to know about them in my professional life.
-
What next? Some millenial comes in and whines that he doesn't know what LPCTSTR stands for?
I don't know what LPCTSTR stands for, and I'm hardly a millennial.
But (unlike a millennial, I guess) I didn't whine about it; I googled it. And now I know what it means. Until I forget, which will be quite soon, because I haven't done any Windows programming since, hmm, probably somewhere between 3.1 and 98, at the latest, so I don't need to remember Microsoft's weird abbreviations.
-
-
@sloosecannon LOL I can't even tell if it's a joke!
-
@sloosecannon said in AMD Hype!:
LPCTSTR
Is that some kind of Linux command?
I think it's how you undo the last change in Vim.
-
@loopback0 said in AMD Hype!:
it's kinda sad that you feel the need to explain what FPGA is on this forum.
I didn't know what one was. I suspect I'm not the only one.
Something to do with golf, I think.
-
@loopback0 said in AMD Hype!:
it's kinda sad that you feel the need to explain what FPGA is on this forum.
I didn't know what one was. I suspect I'm not the only one.
Something to do with
golfcricket, I think.Not sure if
or not.
-
-
@sloosecannon what's raw?
-
@loopback0 said in AMD Hype!:
@sloosecannon said in AMD Hype!:
LPCTSTR
Is that some kind of Linux command?
I think it's how you undo the last change in Vim.
Ah ok
-
@sloosecannon what's raw?
I think he's telling you to watch wrestling on TV
-
@boomzilla said in AMD Hype!:
@loopback0 said in AMD Hype!:
it's kinda sad that you feel the need to explain what FPGA is on this forum.
I didn't know what one was. I suspect I'm not the only one.
Something to do with
golfcricket, I think.Not sure if
or not.FPGA, I think was
he was going for.
-
@HardwareGeek said in AMD Hype!:
@boomzilla said in AMD Hype!:
@loopback0 said in AMD Hype!:
it's kinda sad that you feel the need to explain what FPGA is on this forum.
I didn't know what one was. I suspect I'm not the only one.
Something to do with
golfcricket, I think.Not sure if
or not.FPGA, I think was
he was going for.
-
@HardwareGeek said in AMD Hype!:
@boomzilla said in AMD Hype!:
@loopback0 said in AMD Hype!:
it's kinda sad that you feel the need to explain what FPGA is on this forum.
I didn't know what one was. I suspect I'm not the only one.
Something to do with
golfcricket, I think.Not sure if
or not.FPGA, I think was
he was going for.Here he goes again,
mannerdsplaining things.
-
@sloosecannon said in AMD Hype!:
LPCTSTR
Is that some kind of Linux command?
You’re confusing it with
lpctstr.
-
@Gąska I so hope that milkenials don't know.
And that the rest of us is able to forget.
-
And that the rest of us is able to forget.
The worst part was the L. Although having to deal with the T garbage is also pretty annoying.
Does Windows 10 default to UTF-8 yet?
-
Does Windows 10 default to UTF-8 yet?
Nope, the winapi convention hasn't changed, and internally it is all still ucs2, afaik.
Or do you mean in the console? Also nope
-
And that the rest of us is able to forget.
The worst part was the L. Although having to deal with the T garbage is also pretty annoying.
Does Windows 10 default to UTF-8 yet?Win32 has wide support for UTF-8, as long as you only use code points 00-7F.
-
As for the Xilinx takeover - IMO it makes more sense than Nvidia buying ARM.
Actually, the Nvidia purchase of ARM makes some sense (at a rarefied CEO level, up where there's not enough oxygen to think properly) as the obvious next step for CPU evolution is to put features from GPUs on them. (That's happened before with other accelerators.) Doing that, provided heating can be controlled, gets quite a good performance boost and is valuable. That in turn would leave Nvidia extremely exposed; they need to get into the CPU game in order to stay relevant over the next few decades. And Intel wasn't really available, whereas Softbank was indicating that ARM was (for the right price). What's more, ARM have a lot of experience with controlling power use, an area that Nvidia have historically struggled with; their energy profligacy was starting to eat into the bottom line.
Unfortunately, there are other aspects where things really aren't a good fit. For example, the value in ARM is virtually all in selling the designs to others to integrate into larger SoCs, and that's very much not the Nvidia model (they sell the chips and cards based on the chips). It's obvious that it's not going to be a marriage made in heaven, and it is instead very likely to go down in history as a CEO wasting a lot of money. “Incompatible business models” is a very big red flag, perhaps the biggest single one, when it comes to mergers…
By contrast, AMD and Xilinx at least have more compatible business models.
-
And that the rest of us is able to forget.
The worst part was the L. Although having to deal with the T garbage is also pretty annoying.
Does Windows 10 default to UTF-8 yet?Some of us started programming with Win32 when that
Lmeant something. (Thank $deity 16bit is dead.)
-
(Thank $deity 16bit is dead.)
It's pretty dead for us, too. I can't remember the last time I worked on any 16-bit hardware. 8-bit, yes, for inexpensive, low-power devices. 32-bit, yes; most embedded peripherals on SoCs seem to be 32-bit, even on chips with 64-bit CPUs. But 16-bit, no.
-
And that the rest of us is able to forget.
The worst part was the L. Although having to deal with the T garbage is also pretty annoying.
Does Windows 10 default to UTF-8 yet?Some of us started programming with Win32 when that
Lmeant something. (Thank $deity 16bit is dead.)The first class I took that involved machine language included segment addressing. Thankfully I've managed to forget most of it. Not so much with indexed indirect and indirect indexed addressing over on the 6502.
-
@HardwareGeek it's kinda sad that you feel the need to explain what FPGA is on this forum.
We're software developers not hardware engineers.
-
@DogsB there was a time when, as part of driving test, you had to change a wheel.
temporarily borrowing a belt onion from my mom, who lived in Soviet Poland
AMD Ryzen 5000 and Zen 3 on Nov 5th: +19% IPC, Claims Best Gaming CPU
_(annotated).png){kind=link}