|
PerrineClostermann posted:...Aren't most instructions broken down into common micro-ops anyway? Would you really save that much silicon by dropping old instruction sets? Depends on how they map. SSE turned into a huge savings, because you could map FP instructions onto an SSE instruction with next to no performance loss*, and SSE used up a fraction of the real estate a 'true' floating point unit did. I don't think there is much to be made trying to map SSE onto AVX, as they basically use the same circuit logic anyway. *Some FP instructions got much slower, but anything current that used them was well into being rebuilt to use SSE type instructions instead, which had huge performance gains, and anything older was mostly made up by pure increase in clock speed.
|
#
?
Mar 23, 2018 07:08
|
|
|
|
| # ? Jun 6, 2024 11:41 |
|
|
PerrineClostermann posted:...Aren't most instructions broken down into common micro-ops anyway? Would you really save that much silicon by dropping old instruction sets? Cool pic of older but still relevant CPU cores and how much space was taken up by the FP/SSE portion of them. https://imgur.com/a/QhaWu https://i.stack.imgur.com/QK4gm.jpg
|
|
#
?
Mar 23, 2018 07:18
|
|
|
Huh, I apologize. At some point I must have gotten it into my head that the use of older instruction sets were an issue, like moving to AVX2 would make everything better and in retrospect that seems kind of dumb.
|
|
#
?
Mar 23, 2018 07:31
|
|
|
CanardPC says they'll have a 2700x review in their print magazine tomorrow https://www.cpchardware.com/cpc-hardware-36-debarque-en-kiosque/
|
|
#
?
Mar 23, 2018 15:15
|
|
|
3peat posted:CanardPC says they'll have a 2700x review in their print magazine tomorrow https://www.cpchardware.com/cpc-hardware-36-debarque-en-kiosque/
|
|
#
?
Mar 23, 2018 15:42
|
|
|
3peat posted:CanardPC says they'll have a 2700x review in their print magazine tomorrow https://www.cpchardware.com/cpc-hardware-36-debarque-en-kiosque/ The “unpleasant surprise” is worrisome.
|
|
#
?
Mar 23, 2018 15:44
|
|
|
SourKraut posted:The “unpleasant surprise” is worrisome.
|
|
#
?
Mar 23, 2018 16:04
|
|
|
I'm betting increasing the clocks probably increased the power consumption a bit as well.
|
|
#
?
Mar 23, 2018 16:45
|
|
|
^ higher advertised TDP with a smaller die size = significantly increased power density, maybe that's it. 
|
|
#
?
Mar 23, 2018 17:08
|
|
|
Plus, they're moving from an efficiency-optimized process to a performance-optimized process. There's no free lunch, there's a reason they call it a "low-power process" instead of just "lovely-clocks process".
|
|
#
?
Mar 23, 2018 17:16
|
|
|
I kind of doubt that it's the power draw, people usually aren't too fussed if a desktop CPU pulls a bit more. My guess would be that they're running into the same clockspeed wall again except it's 300mhz higher.
|
|
#
?
Mar 23, 2018 19:02
|
|
|
I'll put 20% of my wager on "they're using paste/you have to de-lid" and 80% of my wager on "there is no OC room at all here, guys." The former sucks, the latter is fine because it means any half-decent X470/B450 motherboard should run these things with all the performance they can give and you don't have to worry about cooking VRMs out of spec.
|
|
#
?
Mar 23, 2018 19:08
|
|
|
3peat posted:CanardPC says they'll have a 2700x review in their print magazine tomorrow https://www.cpchardware.com/cpc-hardware-36-debarque-en-kiosque/ The mailman must've been a bit early cause I got the magazine in my hands right now. By skimming through the article, the "unpleasant surprise" is indeed the powerdraw: more than 140W for the 2700X (despite its official TDP of 105W) and almost 90W for the 2600 (official TDP of 65W). Basically the new 12nm process doesn't seem to bring any kind of power efficiency benefits.
|
|
#
?
Mar 23, 2018 20:07
|
|
|
I'm curious about overclocking too, do they say anything there?
|
|
#
?
Mar 23, 2018 20:18
|
|
|
Paul MaudDib posted:I'm curious about overclocking too, do they say anything there? It's around 20W more draw under load at stock vs its predecessor which is what's expected.
|
|
#
?
Mar 23, 2018 20:34
|
|
|
If there’s a review embargo for Monday, does this mean an eventual paper launch? Interesting for the review(s) to drop with a high TDP on the same day as Nvidia probably announces Volta/Turing.
|
|
#
?
Mar 23, 2018 20:50
|
|
|
I suspect good cooling will be very important for these. The slides they've shown make it pretty clear that their two new technologies (XFR 2 and precision something) are essentially factory overclocks and will make use of what used to be OC headroom. One or two more die shrinks from now we'll probably end up where Pascal is at the moment: a frequency slider, a max TDP slider and progressive throttling as temperatures rise towards Tjmax. PS: maybe they're saving the best bins for Threadripper again?
|
|
#
?
Mar 23, 2018 21:43
|
|
|
ufarn posted:If there’s a review embargo for Monday, does this mean an eventual paper launch? There is a lot of skepticism going around about the CanardPC thing. It's either a hoax, or they aren't under embargo and are testing an ES or retail chip via a board partner leak or something. They apparently aren't getting announced until mid April, if rumor is to be believed.
|
|
#
?
Mar 23, 2018 22:41
|
|
|
Cygni posted:There is a lot of skepticism going around about the CanardPC thing. It's either a hoax, or they aren't under embargo and are testing an ES or retail chip via a board partner leak or something. They apparently aren't getting announced until mid April, if rumor is to be believed.
|
|
#
?
Mar 23, 2018 22:56
|
|
|
Cygni posted:There is a lot of skepticism going around about the CanardPC thing. It's either a hoax, or they aren't under embargo and are testing an ES or retail chip via a board partner leak or something. They apparently aren't getting announced until mid April, if rumor is to be believed. Because they are a magazine and not a website, most of their revenue come from sales (i.e. consumers) and not advertisers. Pulling a hoax like that would ruin their reputation and their sales, just as the French press is going through some serious turmoil. On the other hand, they're generally well informed, and tend not to bother too much with embargoes (their main responsability is towards their readers, not the companies marketing strategy)
|
|
#
?
Mar 23, 2018 22:57
|
|
|
Bourricot posted:Because they are a magazine and not a website, most of their revenue come from sales (i.e. consumers) and not advertisers. That’s not my understanding of how print industry economics work at all.
|
|
#
?
Mar 23, 2018 23:11
|
|
|
Subjunctive posted:That’s not my understanding of how print industry economics work at all. I can't remember the last time I read a print magazine that wasn't a glorified ad.
|
|
#
?
Mar 23, 2018 23:13
|
|
|
I personally don't think its a hoax, I just think they got access to an ES and tested it cause they have no embargo. But if you go to comment threads on other sites, everyone is screaming hoax cause apparently the QR code on the CPU on the magazine cover is a rick roll (which to me makes sense if you are masking the leaked sample you got)
|
|
#
?
Mar 23, 2018 23:14
|
|
|
Subjunctive posted:That’s not my understanding of how print industry economics work at all. They're an exception for sure: I have the magazine in my hands, and counted 9 pages of advertisement (out of a hundred): 8 for retailers, 1 for a videogame (compared to 30-40% for most magazines). I'm pretty sure most websites (Anandtech, Ars, THG,...) are a lot more dependent from ad funding and the goodwill of manufacturers (because they can't afford not to be invited to the latest tech reveal).
|
|
#
?
Mar 23, 2018 23:22
|
|
|
Cygni posted:I personally don't think its a hoax, I just think they got access to an ES and tested it cause they have no embargo. But if you go to comment threads on other sites, everyone is screaming hoax cause apparently the QR code on the CPU on the magazine cover is a rick roll (which to me makes sense if you are masking the leaked sample you got) The rickroll is fully intentional, they like to hide secret messages in binary in the magazine. Edit: sorry for the double-post. The binary translates to "HiReddit!NGC=2023" Bourricot fucked around with this message at 23:34 on Mar 23, 2018 |
|
#
?
Mar 23, 2018 23:31
|
|
|
Bourricot posted:They're an exception for sure That almost literally the opposite of what you said before, but I understand what you mean.
|
|
#
?
Mar 24, 2018 00:06
|
|
|
Bourricot posted:The rickroll is fully intentional, they like to hide secret messages in binary in the magazine. ....NGC could be referring to any number of things, none of which are good.
|
|
#
?
Mar 24, 2018 03:55
|
|
|
SwissArmyDruid posted:....NGC could be referring to any number of things, none of which are good. I choose to believe they're making an obscure joke about nebulae instead of "you ain't seeing a GCN replacement until 2023, suckaaaaaaz"
|
|
#
?
Mar 24, 2018 07:05
|
|
|
Subjunctive posted:That’s not my understanding of how print industry economics work at all. It seems like most American magazines are glorified ads sure, something like C'T (German ) or Edge has much less ads, When I still read Wired every other page was an ad.
|
|
#
?
Mar 24, 2018 09:42
|
|
|
It’s power consumption. But to have an 8 core CPU running with similar consumption to my 4 core Ivy, I mean, I wouldn’t complain. I guess I can see why people who make redundant CPU upgrades would. The other interesting part is that the X470/B450 enhancements are mostly hoopla and you may be able to just go ahead and swing for a current board: quote:The magazine states that while these CPUs are tested on a A320 motherboard, which is the entry level chipset, an upgrade from X370 to X470 doesn’t deliver any substantial performance lift which shouldn’t affect performance too much.
|
|
#
?
Mar 26, 2018 22:17
|
|
|
ComputerBase sez they do indeed have a unit from the first two weeks of 2018, but they prefer a review with proper methodology and BIOS closer to release.
|
|
#
?
Mar 26, 2018 23:37
|
|
|
Guess I'm waiting on Gamersnexus, then, they seem to be the only publication taking specific care to watch Core Performance Boost settings, which could very well be the source of the increased power consumption.
|
|
#
?
Mar 26, 2018 23:55
|
|
|
SwissArmyDruid posted:Guess I'm waiting on Gamersnexus, then, they seem to be the only publication taking specific care to watch Core Performance Boost settings, which could very well be the source of the increased power consumption. This is why they used an A320 board, they specifically wanted to avoid auto-turbo "help" and those boards don't even put a scrap of metal over the VRMs. They're productivity/office boards.
|
|
#
?
Mar 27, 2018 04:17
|
|
|
I have some questions about the AMD Ryzen 5 2400G and because my use case is unusual I can't find much information anywhere else. Firstly, I want to build a ITX PC in this case: http://www.minicase.net/product_E-W150.html Am I correct in thinking that the AMD Ryzen 5 2400G has the fastest integrated graphics available and are any revisions expected to be coming in the next several months? Second, I will be connecting this to a monitor with an unusual resolution of 1360x768. My understanding as far as discrete GPU systems are concerned is that running games at lower resolutions like this is more likely to bottleneck the CPU rather than GPU. Does the same apply to APU builds? Thirdly and AFAIK APUs benefit from dual channel and high clock speed RAM. Is that still true?
|
|
#
?
Mar 28, 2018 10:29
|
|
|
First, and invest in a good CPU cooler under 65mm in height so as to get the most out of a 2400G. Unless you need all 8 threads, the 2200G is the better buy as it can basically reach stock performance of the 2400G when overclocked (which can be done on a wraith spire). It is currently the absolute fastest iGPU solution, since Intel discontinued anything with the Iris Pro 580. No major revisions are expected until the very end of 2018, a move to 12nm which will give marginally better CPU performance and maybe better GPU performance. For an APU, and especially this APU, 720p won't bottleneck anything, in fact a move to 1080p will cause the performance to jump off a cliff in many situations due to bandwidth limitations. Please use dual channel, you can cut the poor iGPU's performance in half with single channel. 3200CL14 provides the best performance, but 3000 CL14 is also nearly as good. Just don't choke it utterly with DDR4 2400 CL 16.
|
|
#
?
Mar 28, 2018 10:55
|
|
|
KingEup posted:I have some questions about the AMD Ryzen 5 2400G and because my use case is unusual I can't find much information anywhere else. Prebuilts with Intel's Hades Canyon (i7-8809/8805) are coming out soon-ish and will be a good deal more powerful and significantly more expensive, doesn't seem like the SoC itself will be available seperately any time soon either. I don't think 12nm/7nm AMD APUs have been announced. You're never going to be CPU limited even with a low resolution outside of some extreme corner cases. Dual channel is pretty much mandatory and memory clocks are also a huge deal, I'd aim for at least 3000 with decent timings.
|
|
#
?
Mar 28, 2018 11:16
|
|
|
KingEup posted:Am I correct in thinking that the AMD Ryzen 5 2400G has the fastest integrated graphics available and are any revisions expected to be coming in the next several months? Which one better fits you, well, it's going to depend on the kind of games you play, and/or your hunger for slightly higher detail settings. That said, the 2400G is not the bonkers value for money the 2200G is but may actually give you more as time goes by. And if you do eventually upgrade your display and get a dedicated GPU card, the 2400G has more threads and clock at stock so it won't be as irrelevant if you don't rely on it's graphics anymore. The one thing you need to keep in mind if you're buying very soon is a lot of stores are selling old stock that won't boot the APUs. I recommend the ASRock boards that Newegg is selling, they're marketed with a 2000-compatible BIOS already flashed (and I bought one and can confirm they do.) quote:Thirdly and AFAIK APUs benefit from dual channel and high clock speed RAM. Is that still true? Craptacular! fucked around with this message at 12:21 on Mar 28, 2018 |
|
#
?
Mar 28, 2018 12:19
|
|
|
Thanks for the all the replies people. I only play DOTA2 so I was hoping to play that on high settings without dropping below 60fps. No idea whether it'll make use of the additional threads on the 2400G. I might just wait until that Haydes Canyon nuc is released before I make a decision. It's also really quite important that I have a DP++ port on the MB so I'll have to chose one carefully. KingEup fucked around with this message at 12:46 on Mar 28, 2018 |
|
#
?
Mar 28, 2018 12:41
|
|
|
"thanks guys, i'm gonna wait until dota takes advantage of multiple threads"
|
|
#
?
Mar 28, 2018 12:44
|
|
|
|
| # ? Jun 6, 2024 11:41 |
|
|
KingEup posted:Thanks for the all the replies people. I only play DOTA2 so I was hoping to play that on high settings without dropping below 60fps. No idea whether it'll make use of the additional threads on the 2400G.  quote:It's also really quite important that I have a DP++ port on the MB so I'll have to chose one carefully. I don't know enough to know what DP++ is about, Gigabyte's GA-AB350N-Gaming WIFI is the only ITX motherboard for Ryzen I know of that has a DisplayPort. Some MSI GT 1030s have DisplayPort, depending on prices it may or may not be worth it for you to rethink this plan and consider a Pentium G4560 with a low profile 1030 instead. EDIT: 2200G is $99 and that board is the same price so it's $200, while a Pentium is like $50 and a basic ASRock board with wifi is $60. That leaves you with enough money to get the card you want even at these prices (MSI new at $80 on Newegg), and your FPS will be higher. The downside is you'll need to find a case that has a PCI slot. Craptacular! fucked around with this message at 13:33 on Mar 28, 2018 |
|
#
?
Mar 28, 2018 13:01
|
|