|
FaustianQ posted:I literally don't get how you think AMD can't make something competitive with Haswell. Just because Construction cores suck doesn't mean Zen will, the logic doesn't follow, and just like Pentium D-> Conroe, the uarchs are totally unrelated. Because they have not really managed to catch up with their own like 2010 products, so frankly I don't trust them to catch up with a few years old Intel stuff. And even if they can in raw performance I'd expect it to still be not even close for power draw, which will be a killer. I can already hear the attempts to downplay the 400 watt draw of their new chip that manages to tie with Haswell.
|
# ? Aug 15, 2015 05:01 |
|
|
# ? May 21, 2024 12:25 |
|
lDDQD posted:If Zen ends up being only 5% slower than Broadwell, I might just buy it out of pity. If it's within 10% I'll buy it; especially for 8 cores.
|
# ? Aug 15, 2015 05:04 |
|
Nintendo Kid posted:Because they have not really managed to catch up with their own like 2010 products, so frankly I don't trust them to catch up with a few years old Intel stuff. See, this isn't a fair comparison since one of the biggest reasons construction cores have lovely IPC is their design, like Zen is entirely something else and actually more or less mimics Intel design except likely in a few key areas. It's pretty much a design limitation on construction cores. I doubt power draw will be on par for Zen, Zen+ might be much closer, but it's shouldn't be anywhere near what Piledriver or Kaveri is pulling. Like, you're either shilling, hyperbolic or just so cynical that you're not making a coherent argument. I mean this is like looking at a P4D and going "well, there's no way Intel is going to catch up to AMD on single threaded IPC, and even if they do it's going to be something dumb like a 150W, lol". Except that Netburst and Core aren't comparable at all and trying to gauge Cores performance based on Netburst is asinine.
|
# ? Aug 15, 2015 05:56 |
|
Yeah, but I can trust Intel to do something like that, because they literally have all the money in the market. It's like playing poker against Intel and AMD's down to their last couple thousand in chips. They're dying because they only have enough chips to buy in each hand, but even if they fold every bad hand, they're still on a downward spiral. Now, there's always the chance that AMD hits a home run, but I'm sure we all agree that this is AMD's final four to eight quarters before bankruptcy. AMD CPU and Platform Discussion: It's The Final Countdown
|
# ? Aug 15, 2015 06:52 |
|
AMD CPU and Platform Discussion: New Zen platform might actually be good!
|
# ? Aug 15, 2015 06:59 |
|
SwissArmyDruid posted:Yeah, but I can trust Intel to do something like that, because they literally have all the money in the market. It's like playing poker against Intel and AMD's down to their last couple thousand in chips. They're dying because they only have enough chips to buy in each hand, but even if they fold every bad hand, they're still on a downward spiral. I'm just saying it's disingenuous to compare Construction cores v. Zen cores, just as much as it is to compare Netburst and Core, as a basis for extrapolating the performance latters performance. I mean, do Zen cores even share similarity to whatever AMD has worked on in the past? That at least would be a good starting point, but the best we have for comparison right now is to Intel stock.
|
# ? Aug 15, 2015 07:14 |
|
Much more interested in K12 being a killer chip these days - at least they might stand a chance long term in the ARM market.
|
# ? Aug 15, 2015 13:34 |
|
wipeout posted:Much more interested in K12 being a killer chip these days - at least they might stand a chance long term in the ARM market.
|
# ? Aug 15, 2015 14:44 |
|
FaustianQ posted:See, this isn't a fair comparison since one of the biggest reasons construction cores have lovely IPC is their design, like Zen is entirely something else and actually more or less mimics Intel design except likely in a few key areas. It's pretty much a design limitation on construction cores. I doubt power draw will be on par for Zen, Zen+ might be much closer, but it's shouldn't be anywhere near what Piledriver or Kaveri is pulling. Like, you're either shilling, hyperbolic or just so cynical that you're not making a coherent argument. But it is a fair comparison. What we know from AMD is for the last x years they've been selling tire fires, and it's reasonable to not accept "it'll totally be good this time" from them until they can prove it. And they've had a worse record on power usage for longer than they've been doing the construction cores. No, the latter example is dumb because Intel still had scads of money at hand during the Pentium 4 fiasco and was already realizing that reverting to P3 styled chips (starting with Pentium M, in laptops) was the way to go. And it's still funny, because 220 watt AMD chips is a thing that actually happened that we were joking about when the first wave of construction chips came out. FaustianQ posted:I'm just saying it's disingenuous to compare Construction cores v. Zen cores, just as much as it is to compare Netburst and Core, as a basis for extrapolating the performance latters performance. I mean, do Zen cores even share similarity to whatever AMD has worked on in the past? That at least would be a good starting point, but the best we have for comparison right now is to Intel stock. Again, comparing Netburst to Core ignores that the Pentium M chips were a thing Intel already had in full production, starting in early 2003. Early Pentium M chips were already stomping many Pentium 4 chips when they came out. AMD didn't have a similar alternate architecture available within 3 years of the Bulldozer tire fires, dude.
|
# ? Aug 15, 2015 15:34 |
|
Intel had poo poo loads of money in the P4 days, a good design waiting in the wings, and hadn't been bleeding talent non stop for 4+ years. It's possible that AMD hits a home run but you can't compare the two situations and say it's as easy as not designing a bad CPU
|
# ? Aug 15, 2015 16:32 |
|
Nintendo Kid posted:But it is a fair comparison. What we know from AMD is for the last x years they've been selling tire fires, and it's reasonable to not accept "it'll totally be good this time" from them until they can prove it. And they've had a worse record on power usage for longer than they've been doing the construction cores. Sorry. you can't compare different uarchs to guess performance, Construction cores are non-determinant of Zen performance, just like you can't use Netburst to guess Core performance. You can however guess based on similar uarchs, which is why Pentium M is a good gauge of early Core, and it's why you can hazard at guessing Zen performance based on Core. This is pretty different than from saying AMD might screw up designing a Core style uarch.
|
# ? Aug 15, 2015 16:36 |
|
It's almost like we won't know how good it is until it is actually released and we see some benchmarks. I don't really think there is much of a point of in saying how good it is or how lovely it is until we actually see it.
|
# ? Aug 15, 2015 16:41 |
|
FaustianQ posted:Sorry. you can't compare different uarchs to guess performance, Construction cores are non-determinant of Zen performance, just like you can't use Netburst to guess Core performance. You can however guess based on similar uarchs, which is why Pentium M is a good gauge of early Core, and it's why you can hazard at guessing Zen performance based on Core. This is pretty different than from saying AMD might screw up designing a Core style uarch. Again, dude, Intel didn't go straight from Netburst to Core, they had Pentium M up and going in under 3 years from the Netburst launch and already showing significant improvements at launch. And then over the next 3 years until it really became Core it kept getting improved and improved. Meanwhile what does AMD have? Well they'll supposedly have another architecture "soon", instead of say already having it available middle of last year, which would be on par with Intel being able to change course and bring out the Pentium M. And that new architecture is probably just going to be able to stomp over existing construction cores, not be very competitive with anything recent from Intel when it first launches. We really can compare different uarchs to guess performance. Guess what? AMD has a legacy of power hunger and an engineering team that's more restricted/underfunded then ever. It's the rational thing to expect Zen to be power hungry as hell compared to similar performance Intel chips, and to also not get anywhere close to recent ones by the time it's out.
|
# ? Aug 15, 2015 16:54 |
|
Nintendo Kid posted:It's the rational thing to expect Zen to be power hungry as hell compared to similar performance Intel chips, and to also not get anywhere close to recent ones by the time it's out. At this point it wouldn't be difficult to get within a decent range of Skylake performance since Intel has stagnated on the performance front in order to optimize power efficiency. So while yes, it very well could be less efficient power-wise, I doubt many are going to care if AMD's Zen is 110W or whatever and within a decent performance % of Skylake at 95W for desktop usage if it means some competitiveness back in the market, and especially if Zen is at least overclockable to help make up that % since we know Skylake really isn't. You're also "mis-remembering" the entire Pentium M situation, since the Pentium M was just a continued development of the Pentium III (Tualatin) at the time, so there was no magical "3-year" period where Intel magically had Pentium M "up and running" - it always was "up and running", because the Pentium III already existed. Sure, it took a lot of R&D to get it to where it ended up being, but it's disingenuous to represent it as some magical new architecture. But I'll chalk it all up to "Fishmechism".
|
# ? Aug 15, 2015 17:06 |
|
SourKraut posted:Except that we don't know how long AMD has been working on this architecture - they've known for awhile that the Construction cores weren't their long-term future, so for all you know it's been in development for 4-5 years already. They worked long as hell on the construction cores and they still sucked, I don't see how working long on this new thing would for sure make it not suck. It would be pretty difficult for them to get int that range since it would require a massive leap in performance from anything they've done before. And power draw really does matter if they ever hope to be serious competitors in laptops and to a lesser extent servers again. If Pentium M is "just" Pentium III then Core is just Pentium M as well, and so on and so forth dude. Meaning that we're basically just running multicore Pentium IIIs today. It's a useless thing to claim that III and M are the same but M and Core ain't.
|
# ? Aug 15, 2015 17:11 |
|
Nintendo Kid posted:They worked long as hell on the construction cores and they still sucked, I don't see how working long on this new thing would for sure make it not suck.
|
# ? Aug 15, 2015 18:26 |
|
wipeout posted:Much more interested in K12 being a killer chip these days - at least they might stand a chance long term in the ARM market.
|
# ? Aug 15, 2015 19:30 |
|
It's a market that only exists in the minds of the most vividly dreaming AMD fanatics.
|
# ? Aug 15, 2015 19:36 |
|
Professor Science posted:AMD has no mobile presence. This has a slim chance of happening, of course, but it's pretty much their only clear shot at getting as much as they wish for out of Zen. Anime Schoolgirl fucked around with this message at 20:07 on Aug 15, 2015 |
# ? Aug 15, 2015 20:01 |
|
Anime Schoolgirl posted:This is pretty much the most important part of the equation. They have to at least put something out that squishes the low-TDP i3/i5 mobile chips in terms of features and performance at the same power envelope, and it's something the laptop market is in need of if we want actual build quality in the 600-700 dollar price range. But didn't they actually do that already with Carrizo? (An actual laptop with the second highest Carrizo in). It has better gaming performance than the stuff Intel has at the same price, and it even has hardware x265 decoding which can push out 3840×2160 60 FPS video. But almost nobody is building laptops with them. There's probably no chance you'll see them in a decent laptop, either, because Intel has the higher end brand image. vv But it is in the same power envelope, and definitely will be cheaper, and has better graphics performance. CPU, of course not, but it's 'good enough'. I actually think Carrizo could do well if it had a chance. HalloKitty fucked around with this message at 20:41 on Aug 15, 2015 |
# ? Aug 15, 2015 20:12 |
|
HalloKitty posted:But didn't they actually do that already with Carrizo? (An actual laptop with the second highest Carrizo in). Part of the problem is with the OEMs themselves, as well. The ultrabook craze has made everyone throw build quality out the window in favor of using putty-plastic that bends at will for the sake of losing a quarter of an inch in thickness. Plus everyone wants an i3 or better and i3 mobile chips cost twice as much as i3 desktop chips and thus you have keyboards and trackpads that stop being operational and responsive at anything beyond the slightest touch at the SKU price people actually buy, aka 600-700 dollars. Something is wrong when Chromebooks are sturdier than those. Tangential, but I'd like to be able to build ECC-equipped PCs again without paying the $150-200 Intel Xeon chipset and CPU premium. At least I finally built a $1500 computer without being ashamed of it. Anime Schoolgirl fucked around with this message at 20:37 on Aug 15, 2015 |
# ? Aug 15, 2015 20:35 |
|
So what your saying is that for the same price, carrizzo laptops could have superior build quality? Here's a dumb question, can AMD leverage its experience with x86 to enhance K12 to run x86 programs? Or has Intel effectively covered this with the Transmeta deal?
|
# ? Aug 15, 2015 21:13 |
|
FaustianQ posted:So what your saying is that for the same price, carrizzo laptops could have superior build quality? Unfortunately they'd rather put them on 400 dollar SKUs with one channel RAM which effectively halves the graphics power on those things.
|
# ? Aug 15, 2015 22:10 |
|
FaustianQ posted:Here's a dumb question, can AMD leverage its experience with x86 to enhance K12 to run x86 programs? Or has Intel effectively covered this with the Transmeta deal? They can, but, that would probably be mighty pokey, and I have no idea what sort of OS setup could effectively use it.
|
# ? Aug 15, 2015 22:24 |
|
Anime Schoolgirl posted:That is correct. 100-150 dollars less could have them use actual sturdy lightweight plastic, an actual screen instead of 1366x768, and still maintain a profit margin. 1366x768 seems to hurt my eyes and I have no idea why, but hey at 15W, a Carizzo Chromebook sounds amazing (that will never happen ) Nintendo Kid posted:They can, but, that would probably be mighty pokey, and I have no idea what sort of OS setup could effectively use it. Just thinking out loud about AMD moving away from x86 or at least being able to bridge the gap.
|
# ? Aug 15, 2015 23:27 |
|
FaustianQ posted:Just thinking out loud about AMD moving away from x86 or at least being able to bridge the gap. Out of x86 and into what? Wouldn't going into ARM get them trounced by more experienced actors?
|
# ? Aug 15, 2015 23:37 |
|
They'd be up against Samsung and Qualcomm at the high end with mediatek and a hundred other Chinese shops at the low/mid-range. Not much room to maneuver there.
|
# ? Aug 16, 2015 00:13 |
|
Boiled Water posted:Out of x86 and into what? Wouldn't going into ARM get them trounced by more experienced actors? Isn't one of the biggest issue with ARM, besides performance obviously, is it's backwards compatibility with x86? cisco privilege posted:They'd be up against Samsung and Qualcomm at the high end with mediatek and a hundred other Chinese shops at the low/mid-range. Not much room to maneuver there. I thought Qualcomm and Intel were in talks? Also, AMD is apparently partnering with Mediatek.
|
# ? Aug 16, 2015 00:22 |
|
FaustianQ posted:Isn't one of the biggest issue with ARM, besides performance obviously, is it's backwards compatibility with x86? In transitioning minor applications over sure, but for servers people will port poo poo to anything for business line applications, including Itanium, so long as the chip works well for a server. And having a slow x86 core on top of a decent ARM core doesn't really help for anything, you'd need a radically modified OS to handle some of the stuff going on on x86 and some of the things going on on ARM. There are some mainframe OSes which can handle splitting tasks over diverse microarcitectures, but it's useless for say a Windows application.
|
# ? Aug 16, 2015 00:36 |
|
Anime Schoolgirl posted:That is correct. 100-150 dollars less could have them use actual sturdy lightweight plastic, an actual screen instead of 1366x768, and still maintain a profit margin. I'll never get why anyone would use single channel RAM over dual or better. I could maybe see it being about cost, but that stuff has been around for ages & can't be that pricey. I'm still hoping AMD can pull a surprise like back in the K7 & above days so we can get some good competition/pricing again.
|
# ? Aug 16, 2015 00:52 |
|
Nintendo Kid posted:In transitioning minor applications over sure, but for servers people will port poo poo to anything for business line applications, including Itanium, so long as the chip works well for a server. The real problem, of course, is that ARM isn't easier to make a chip for than x86 so who even cares if you're building ARM or x86 anyway. Larrabee/the Knights line disproved the whole "instruction decode on x86 will completely dominate the power for a small chip" thing (also Edison).
|
# ? Aug 16, 2015 01:55 |
|
Can this be the general CPU thread please? It'd make it more interesting and active compared to the slow wait for whatever AMD is releasing, and a rundown on different uarchs would be a good primer to discussion.
|
# ? Aug 16, 2015 02:06 |
|
Professor Science posted:nah, you'd just do binary translation like Denver or x86 Android. Houdini on x86 runs ARM code on x86, after all. The problem is that the perf hit sucks in a close market. Er that's the thing. You're talking about software emulation essentially. His thing would be multiple dissimilar cores, like 1+ ARM and 1+ x86. People already had some pretty annoying issues with chipsets of two different kinds of ARM activated at different times in an attempt to save power, some cores ARM and some cores x86 would be even more difficult.
|
# ? Aug 16, 2015 02:20 |
|
Professor Science posted:nah, you'd just do binary translation like Denver or x86 Android. There have been no hints of them doing anything like this though. Most likely they'd end up doing binary translation like everyone else does since that is much simpler over all to do anyways.
|
# ? Aug 16, 2015 03:33 |
|
Ozz81 posted:I'm still hoping AMD can pull a surprise like back in the K7 & above days so we can get some good competition/pricing again. http://www.anandtech.com/show/6129/apple-a4a5-designer-k8-lead-architect-jim-keller-returns-to-amd
|
# ? Aug 16, 2015 06:09 |
|
Boiled Water posted:It's a market that only exists in the minds of the most vividly dreaming AMD fanatics. Ultimately, I think the only future AMD has is pumping out ARM chips. K12 performing would give me hope they could do that, even if its just because it gives them a load of architectural modules for future ARM chips, mobile ones if they have sense / more value for acquisition. Hopefully Zen will be awesome but even if it is, the x86 war of attrition is going to kerb stomp them eventually and its just not interesting watching that happen. The only time I was fanatical at all about AMD was when they were actually better then Intel chips, and people still didn't use them. Makes u think. GRINDCORE MEGGIDO fucked around with this message at 14:09 on Aug 16, 2015 |
# ? Aug 16, 2015 13:52 |
|
wipeout posted:Ultimately, I think the only future AMD has is pumping out ARM chips. K12 performing would give me hope they could do that, even if its just because it gives them a load of architectural modules for future ARM chips, mobile ones if they have sense / more value for acquisition. The only times AMD held the performance crown was versus Katmai P3s, Willamette and Prescott P4s, the latter two where Intel pruposely dived head first into executing that awful Netburst uarch. Rest of the time they are only competitively priced at the low end. For portables which has steadily grown to 2/3 share of the entire PC market, they basically handed the entire segment to Intel on a silver platter since god knows when.
|
# ? Aug 17, 2015 04:11 |
|
japtor posted:Well they did get one guy from those days back a few years ago: Please PLEASE let this guy do something awesome for AMD again...video cards and APUs are great, but I'd love to have another Opteron/Athlon/Phenom-level PC again that's at least somewhat comparable to Intel's chips.
|
# ? Aug 19, 2015 17:50 |
|
Semiaccurate predicts financial changes for AMD. I know, Semiaccurate, yadda yadda, grain of salt, but what will be spun off/sold to Samsung?
|
# ? Aug 20, 2015 06:17 |
|
|
# ? May 21, 2024 12:25 |
|
I doubt Samsung wants any part of AMD. I could see Intel or NV snapping up the GPU portion and maybe some patents but that is about it.
|
# ? Aug 20, 2015 09:31 |