|
That video is pretty good. Nice to hear some details about the more arcane Zen specific settings though he doesn't dwell too much on any of them.
|
# ¿ Apr 28, 2017 05:16 |
|
|
# ¿ Apr 30, 2024 02:00 |
|
If you're actually doing something using all those threads then 4 channel memory is a must. For games and other stuff 16 threads or less I wouldn't expect to see much if any difference with one of those at all vs a similarly clocked 8C16T Ryzen. A 64 MB L3 could make a nice difference for somethings like Crystalwell did though. There has been no real word on what effect any of the minor fixes that were hinted at are going to have. I wouldn't expect much there. Probably just some bug fixes and little to nothing to do with performance. Platform wise if you really need lots of PCIe then you've got something to look for I guess. I'd care more about the integrated 10Gbe ports myself. The block diagram seems to show them integrated into the CPU itself which is a bit odd, should be pretty fast and power efficient as a result of that.
|
# ¿ May 11, 2017 12:16 |
|
Nah, a few care about it here since they actually want to do some semi serious work but ECC isn't seen as a must have at all really for the OC'ers/gamers. Bigger question would be if they have this nasty bug fixed so you can run VM's and legacy 16 bit code properly. http://www.os2museum.com/wp/vme-broken-on-amd-ryzen/ Supposedly its fixable through a microcode update since internally the CPU itself is just emulating all those old instructions anyways on the hardware. No word on a fix yet though. Seems to only have been known about publicly for a few days now as far as I can tell. I'd expect it to be fixed soon, their new server platform is launching in a few months after all...
|
# ¿ May 15, 2017 13:16 |
|
Munkeymon posted:Maybe their BIOS authors just suck? Maybe I should stop buying Asus? Only time will tell. Of course I also overclock too so it could easily be just that and none of it happens enough to be a real problem so I don't care much, just pointing out problems like these are everywhere and not much you can do but shrug about it. VealCutlet posted:you could also add more voltage to the RAM but I'm not 100% sure what is safe for 24/7 usage on DDR4. Go to 5:30 for where he talks about RAM overclocking in general. 6:00 for talk about RAM overvolting and "safe" ranges. https://www.youtube.com/watch?v=vZgpHTaQ10k tl&dr for OC'ing Ryzen: vCore 1.4v, vSoC 1.1v, vDRAM 1.4-5v = 4Ghz most every time. He talks about a few other things but that is the main take away. PC LOAD LETTER fucked around with this message at 01:46 on May 19, 2017 |
# ¿ May 19, 2017 01:36 |
|
It can help with both. Exactly how much? Dunno, probably varies per chip. Going by the video it doesn't seem an issue to set it to 1.1v though.
|
# ¿ May 19, 2017 02:27 |
|
SamDabbers posted:My kit's XMP profile says 15-16-16-35 and the BIOS on my board bumps it to 16-16-16. Maybe CL, TRCD, and TRP need to be set equal with this version of AGESA? Turdsdown Tom posted:Another RAM related question: is RAM timing in the BIOS accurate? My timing should be 15-17-17-35 but instead the MSI BIOS insists on displaying it at 15-15-15-36. Should I be concerned? Should I manually enter timing info? It should be accurate in the BIOS but if what is in the BIOS and on the RAM sticker itself don't match then I wouldn't be concerned unless you're not stable at the clocks you want. Then I'd manually set them to what the sticker says and see what happens. I don't think anyone actually knows what or how exactly AMD's current version of AGESA is setting the timings. About all you can seem to say is that it really likes to push tighter timings as much as possible. There is an update that is supposed to be coming soon that should allow lots more flexibility there but I don't think anyone knows exactly when its supposed to be out.
|
# ¿ May 20, 2017 02:49 |
|
Supposedly its a mish-mash of PCIe and their HT standard. They probably should've called it HT4.0. Its PCIe heritage is supposed to be why it scales up and down so easily but also is why its supposed to be relatively high latency too. That is all rumors at this point I believe, I don't think AMD has released many technical details about it so far.
|
# ¿ May 25, 2017 05:26 |
|
GRINDCORE MEGGIDO posted:I'm not convinced an entirely new basic architecture is required all that often If you want big performance jumps than either a drastic redesign or new arch. is needed. The exception to that is if there were some major performance inhibiting bugs that get fixed but I don't think AMD's Zen has that issue. Progressive tweaks of an existing design tend to get you middling to high single digit performance increases with each revision. Sometimes not even that really. Kind've like how Intel has been doing since Sandybridge. Which no one is very impressed with, at least not for Intel's prices. AMD might be able to get away with minor progressive tweaks to Zen over the next few years if they a)keep prices noticeably lower than Intel's while b)still keeping overall performance fairly close to Intel's best and c)Intel doesn't suddenly produce a new arch. or process tweak that either gets them 20%+ more IPC or heaps (ie. 6Ghz+) more clockspeed at similar to current TDP's. The chances of c happening are pretty slim to none I think so AMD is probably safe for at least a couple years just tweaking Zen. Especially if Zen+ really does turn out to get 10-15% more IPC with similar or better clocks at the same TDP's. At that point it'll be on par or a bit faster per clock than Kabylake. They'll still have a process deficiency to try and make up for of course and that will hamper any chances of getting on par with Intel's clockspeeds either at stock or OC'd but competing on price will allow them to still sell heaps of chips at much higher ASP's than they've had during the Bulldozer years so its still a huge financial win and a great value buy for any buyers.
|
# ¿ May 27, 2017 03:06 |
|
Theris posted:Was Sandybridge a clean sheet redesign? I was under the impression that the current Cores are still an evolution of P6. Pentium M (Yonah, Banias, Merom?, others too mostly mobile oriented) was more closely related to the P6 than Sandybridge, but even then it was a very much a different architecture vs the original P6 from the mid 90's. GRINDCORE MEGGIDO posted:Yes. I think they can get by for a few years, by then ditching gloflo is a great opportunity. They should definitely be able to grow market share by just improving Zen in both the desktop and server markets though and that is definitely a big deal. Who knows, maybe in a couple of years they might start getting close to having 20% marketshare again. PC LOAD LETTER fucked around with this message at 15:37 on May 27, 2017 |
# ¿ May 27, 2017 15:30 |
|
Thanks for those links. I had no idea there was that much public information available on the K5. Its a pretty old CPU now but it seems to me the approach used was fairly out there (as I understand it the K5 was pretty much a 29000 RISC CPU with a x86 "front end") yet it worked fairly well still. Too bad they never got the clocks up much but I have a few fond memories of it still.
|
# ¿ May 29, 2017 14:01 |
|
Maybe in a few days. I'm hoping they'll offer the 16C/32T version for very close to $1K. Those are the magic mindfuck psychological numbers they'd probably need to hit to move lots of those chips fast. I dunno if I'd get one, really 8C/16T is more than enough for me right now but at that price I'd be tempted. It'll depend on what the platform costs too. That loving socket does not look cheap.
|
# ¿ May 30, 2017 15:24 |
|
Size seems typical for a high end 17" gaming laptop to me. The battery life is short on most of them yeah you do end up treating them like a 80's portable PC that you plug in everywhere.
|
# ¿ May 30, 2017 16:01 |
|
Geez if they can just put a "lousy" 1GB of HBM1 on their APU's it'd be a huge win for iGPU performance. It could probably compete fairly well with some of the mid-ish range dGPU's at 1080p or less resolution with that. Seems like that isn't gonna happen though. At least not for 2017.
|
# ¿ May 31, 2017 11:27 |
|
Arzachel posted:As much as I'd like to see a zen quad with a RX480-level GPU on die paired to a stack of HBM, but it doesn't make any sense for Raven Ridge when and the memory controller can push 30-40 GB/s on DDR4 at average clocks. The GPU part is absolutely tiny.
|
# ¿ May 31, 2017 16:14 |
|
Woooow they're gonna sell a bunch of them even if the mobo's cost $300+ a pop. I wonder what the actual server stuff is gonna cost if they're willing to drive prices this low on the desktop side. If production costs are that low and yields are really are super high they can sell for very high discount vs competing Intel products and still make more money than they've dreamed of in years.
|
# ¿ Jun 2, 2017 10:18 |
|
Intel has gigantic operating costs due the fabs they run and R&D costs to improve them that AMD doesn't have to deal with though. A price war might just end up hurting Intel more than AMD at this point. I think AMD is more than happy enough to "leave money on the table" with their current pricing since they're still getting more money, volume, and better ASP's than they've had in years and they probably really need the cash now while building some good will in the market. And this is without even having their new server chip out yet or mobile parts either and those should be much more profitable for them.
|
# ¿ Jun 7, 2017 17:40 |
|
RWT did a great article on GAA for anyone wanting read more on it: http://www.realworldtech.com/intel-10nm-qwfet/ Its a couple years old now but much of the information still holds up. Seems like DK was expecting it sooner too but maybe it just wasn't practical yet.
|
# ¿ Jun 10, 2017 01:21 |
|
wargames posted:What does 134,699 feet convert to real money? Watermelon Daiquiri posted:Gah good loving lord that 'quantum well' gate growing a layer of si in the middle of all that? https://www.researchgate.net/publication/260829824_Triple-Junction_Quantum-Well_Solar_Cells_In_Commercial_Production PC LOAD LETTER fucked around with this message at 15:32 on Jun 10, 2017 |
# ¿ Jun 10, 2017 15:27 |
|
Eh I don't read Hungarian though. drat all the PC enthusiasts must be stoked over in the EU then with AMD's prices. Seems like you guys normally get boned hard over there on cost. edit: so VAT still bones you some then eh? That is too bad.\/\/\/\/\/ PC LOAD LETTER fucked around with this message at 15:57 on Jun 10, 2017 |
# ¿ Jun 10, 2017 15:42 |
|
Man they are going after the profits hard with some of those top end chips but the lower end is surprisingly affordable for what it is. Clocks could stand to be higher but maybe they won't be terribly far behind the new Xeon's, if the leaks are correct. Getting the clock speeds up on high core count CPU's while also keeping TDP's non-stupid is tough for everyone. If those SPEC scores hold up they'll probably still sell a fair amount of those top end chips.
|
# ¿ Jun 15, 2017 16:15 |
|
underage at the vape shop posted:Are ryzen's reliable? If you mean "do they break a lot" then no, they seem to be holding up OK but there is no sort of high quality information you can go by publicly on Ryzen failure rates. If you mean "are there bugs that prevent them from running in a stable fashion" then generally no with a few exceptions that seem to be slowly getting fixed right now.
|
# ¿ Jun 16, 2017 08:52 |
|
I can't help but wonder if TR's rumored prices are correct if AMD is willing to sell 16C/32T Epycs for only $600-800. Even at just 2.2Ghz that is a whole lot of cores for that price. Even if the "cheap" TR mobo's end up costing around $300-400 a ~$500 12C/24T TR overclocked to ~4Ghz would be drat tough to beat if you wanted a HEDT system. Of course there is also supposed to be a 10C/20T version TR but there haven't even been rumors of its price yet, maybe it'll be ~$400? I think even Paul would have good things to say about that much performance for that money even at stock (3.1-3.7Ghz) speeds.
|
# ¿ Jun 16, 2017 16:44 |
|
Twerk from Home posted:AMD needs to price aggressively. If the rumored prices of the new top end Skylake Xeons are correct (around $12,000) then the top end Epyc's ($4,000) will more like priced around 70% less. That is pretty gigantic and the performance, overall, shouldn't be all that far behind either for many things that servers need to do. HPC will be a different story but I think AMD has as good as already admitted they don't expect to really compete much with Intel there.
|
# ¿ Jun 16, 2017 17:58 |
|
Paul MaudDib posted:Try resetting the CMOS before you write it off.
|
# ¿ Jun 22, 2017 09:41 |
|
FaustianQ posted:I'm still wondering if Raven Ridge can be MCM'ed since both Ryzen and Vega support IF, would a mere MCM allow Vega dies to communicate as one, or would they just automatically XCF themselves? Also probably not feasible due to the cost. At least right now anyways. If MCM's/interposers get cheap enough you'll see a change of course. I remember some leaks on TR/Epyc saying the manufacturing cost of dies and the packaging (ie. MCM substrate, IHS, testing, etc.) was around $120-140 which is actually pretty low if true. But still too high for a APU which are probably going to be mostly low cost/low end products destined for business machines with some mid range priced versions meant for entry level "gaming" systems. At $400 or so for entry level TR/Epyc AMD should still be able to make a nice profit, even after R&D costs are accounted for, on those so it makes sense on higher end products.
|
# ¿ Jul 7, 2017 16:31 |
|
NewFatMike posted:Low end pricing has to be something AMD have figured out of Navi is going to be fully modular like Zen, though. If AMD can pull off making 2-4 GPU's on a MCM/interposer work well enough together to fake being a single die big GPU it'd be a big win I think. They'd be able to cover all the product lines with only 1 die from the foundry effectively and still keep production costs down which would be a slick hat trick to pull off. If they manage to pull it off well enough with Epyc/TR I really don't know why they wouldn't be able to with Navi.
|
# ¿ Jul 7, 2017 21:46 |
|
NewFatMike posted:Don't you think they might target APU GPU dies and move up from that like the Zen CCX? Cygni posted:GPUs are theoretically more suited to multiple die packages too due to the parallel workload. Cygni posted:Epyc's latency for going between the dies is fairly atrocious, and AMDs way of addressing that was essentially working around that reality (just like Intel did with the Pentium D). My understanding with a ccNUMA set up like AMD appears to be using is that so long as latency on each inter core/die/package/whatever bus hovers around what you'd expect from main system RAM then you'd be OK. Then it just comes down to bandwidth and the number of hops necessary between dies and it seems like they've done a OK job on bandwidth and minimzing hops on Epyc with the sheer number of buses going on there. Proof will be in the pudding so to speak but nothing about it looks actually bad so far. Cygni posted:I mean heck, the concept you're describing for Navi is basically VSA-100. Cygni posted:The problem Navi faces is the same one AMD has faced for years: Nvidia isn't just sitting around waiting for them to catch up. Volta is already at the reticle limit of a 12nm process, so I imagine Nvidia is as fully aware of the manufacturing constraints and future performance needs as AMD is.
|
# ¿ Jul 8, 2017 04:32 |
|
GRINDCORE MEGGIDO posted:If RTG don't sort their poo poo out and nVidia launch an MCM GPU before Navi, it'll just be cosmically sad. FaustianQ posted:not enough money for marketing and engineering combined. I really don't know why AMD keeps having these issues, and execution has been a problem for years and years there, but whenever it comes down to do or die they usually seem able to pull something off. To me that is strongly suggestive of a management issue and not a engineering or marketing one. Upper management has seen a whole lot of turnover there so maybe we won't see a repeat of past behaviors. Have to wait n' see....
|
# ¿ Jul 8, 2017 05:42 |
|
FaustianQ posted:but I was thinking about marketing signing checks the engineers can't cash. Stuff like, 1.5Ghz 4096SP Vega 10XTX @ 225W, vs current reality, or the over promises on Barcelona and Bulldozer (performance, release date, etc.). I don't work at AMD so I have no firsthand knowledge of how they do things there but every place I've ever worked before, or ever heard of, the management had firm control over marketing stooges so that is why I'm looking at things that way. FaustianQ posted:But yea I can definitely see how past management was a problem, esp. w/r/t Bulldozer for instance.
|
# ¿ Jul 8, 2017 06:54 |
|
Combat Pretzel posted:So about MCM latencies of TR/EPYC, I suppose it would have been better to have a memory controller as separate entity on the IF and have the CCX groups be autonomous? I'm not a CPU designer but I think the approach you're talking about makes more sense if there are more hops between dies OR if there are scaling issues with high(er) core counts with their current approach (I have no idea if there are).
|
# ¿ Jul 9, 2017 22:31 |
|
Its a latency issue not a bandwidth issue with inter CCX data transfers.
|
# ¿ Jul 10, 2017 16:21 |
|
FaustianQ posted:Sounds like Intel hosed up Skylake X L3 design and AMD might come away with the better overall scalable solution even without considering MCM. Unimpressive gaming performance by SkylakeX definitely gives TR and 8C16T Ryzen in general an edge they wouldn't have had otherwise but probably not a decisive one in of itself. I think its when you combine that issue with some of the other X299 platform SNAFU's + higher pricing that Intel is demanding for Skylake X that TR and 8C16T Ryzen starts to look like the way to go. I'm sure some will still buy Intel no matter what though. After all even Netburst was a solid seller for Intel! FaustianQ posted:if Intel continues to push the mesh style uarchs which seem to regress performance
|
# ¿ Jul 19, 2017 13:33 |
|
eames posted:AMD is in an excellent position if they can squeeze more frequency/IPC out of Zen+. At that point it'll come down to how many chips GF, or pooooooooooooooossssibly Samsung I guess too, can fab up for them. AMD was always fab constrained in the past when they had a competitive architecture, I think for the longest time (even going back to the K6 era) the best they could supply was around 30% of the world's x86 demand (the Dresden fab was supposed to allow them to supply around 50%+ x86 demand initially but then multicore started to take off which blew up die sizes and that was that...), which guaranteed Intel to be able to still sell a heap of chips no matter what. If after going fabless, and after all the poo poo they've been through, they end up being able to really meet a majority of x86 CPU demand for once in their history I'm gonna laugh and laugh and laugh.
|
# ¿ Jul 19, 2017 14:06 |
|
Zen+ and Zen2 are the same thing. Its a branding mix up for a unreleased product so its irritating but doesn't actually matter since almost no one but us and a few other industry types even looks at AMD presentation slides.
|
# ¿ Jul 20, 2017 08:37 |
|
eames posted:Alpha PAL8045 with Delta fans. http://imgur.com/a/MHcwY Alpha P7125's were primo for their time though.
|
# ¿ Jul 20, 2017 14:04 |
|
Otakufag posted:Hi somebody from the future told me Zen 2 will be AMD's Sandybridge, is it true?? IMO right now seems reminds me of the K6 or K6-2 vs PII era with Intel having a (minor) performance advantage but AMD having a (large) value advantage.
|
# ¿ Jul 22, 2017 06:41 |
|
Palladium posted:Nah Zen is the 2000-2001 era Durons: outstanding perf/price as long as you are willing to tolerate a slightly flakier platform. If you want to bring chipsets and platforms into it everything gets more confused and difficult to compare. For instance the current AMD chipsets are waaay more stable than those old VIA KT133's on up the Durons had to use. They're current ones are just somewhat slower than Intel's current chipsets but not enough to matter much in typical practical use. Even getting to DDR4 3200 speeds is becoming more routine now, though still not a sure thing yet. Palladium posted:K6s were rubbish gaming CPUs running on rubbish chipsets Palladium posted:got massacred six ways to sunday by the integrated L2 cache Celerons. My old Celeron 433A
|
# ¿ Jul 22, 2017 14:34 |
|
Ihmemies posted:No normal/workstation mobos yet again I guess? If you can't wait for that then just buy TR and turn off all the LED's on your mobo of choice. HEDT on either AMD and Intel is at least as much about doubling down on RGB blinkenlights now as it is about having moar corezzz.
|
# ¿ Jul 26, 2017 02:21 |
|
If its cheap enough its worth it even if it is power hungry. But yeah probably won't be cheap enough at launch I'm guessing.
|
# ¿ Jul 26, 2017 11:24 |
|
|
# ¿ Apr 30, 2024 02:00 |
|
GRINDCORE MEGGIDO posted:Yea. I really hope Ryzen results in the 6c coffeelake chips costing less. 6c needs to be a mainstream thing really for desktops, IMHO.
|
# ¿ Jul 27, 2017 04:44 |