|
So I just sold a couple of items on ebay to help pay for my sweet new Ryzen build and I accidentally swapped the shipping labels so both items went to the wrong buyer. FML.
|
# ? Feb 14, 2017 07:10 |
|
|
# ? May 30, 2024 14:18 |
|
Aesculus posted:How is the quad-core at 3.2ghz doing better than the octa-core at 4ghz per core Too much thread ripping, they get all jammed up when you put them together, like Three Stooges Syndrome? Or a big ol' cache miss, like this guy did: Bareback Werewolf posted:So I just sold a couple of items on ebay to help pay for my sweet new Ryzen build and I accidentally swapped the shipping labels so both items went to the wrong buyer. FML.
|
# ? Feb 14, 2017 07:25 |
|
Aesculus posted:How is the quad-core at 3.2ghz doing better than the octa-core at 4ghz per core The per core numbers seem to be just the total score divided by the # of cores so since the benchmark doesn't scale perfectly with additional cores any CPU with more cores will be disadvantaged on that chart.
|
# ? Feb 14, 2017 07:29 |
|
another wrinkle is that the 6-core seems to have the full 16mb of cache on the die to use -- which means it could be better than both the 4c and 8c in some contexts. part of the reason why the 5820k is in the absolute price-perf sweet spot is because of the relatively fatass cache they have on it.
|
# ? Feb 14, 2017 07:33 |
|
So Intel CPUs are still going to be better for gaming?
|
# ? Feb 14, 2017 09:26 |
|
spasticColon posted:So Intel CPUs are still going to be better for gaming? also, encoding 144fps or 4k video is now on the menu for people who have at most $1000-1200 to spend on a platform
|
# ? Feb 14, 2017 09:36 |
|
Mr.Radar posted:I've read that it's because those benchmarks heavily use AVX and Ryzen's AVX implementation is only "half width" (i.e. it processes AVX data in two 128-bit chunks whereas Intel's implementation can processes all 256-bits in one go) to save on transistors and power. Presumably that's something they'll address on the next generation Zen chips. AMD chips are traditionally 2 x 128 bits here, i.e. one 256-bit op per clock. Intel has 2 full 256 bit units (as of Haswell), so two 256-bit ops per clock It's not so much the dual half units (that's actually better for non-AVX code, no disadvantage vs Intel there), but that Intel's chip have doubled up AVX resources, including a cache subsystem that can sustain it. As pointed out, Intel's advantage is a bit reduced because their chips must downclock when the dual units are in use. But there's certainly linear algebra or even machine learning code that can make full use of the AVX units. That's also the kind of code that likes GPUs. So it's a bit of a wash - certainly a very reasonable trade-off from AMD.
|
# ? Feb 14, 2017 12:04 |
|
It’s a reasonable tradeoff on the whole, but it’s a little (but only a little) like the FDIV bug: Sure, it probably won’t affect you, but you never know when you’ll run some code where it does matter. It’s tempting to pay the Intel tax and guarantee that can’t happen. No one has ever been fired for choosing Intel.
|
# ? Feb 14, 2017 12:22 |
|
Platystemon posted:It’s tempting to pay the Intel tax and guarantee that can’t happen. https://www.reddit.com/r/networking/comments/5sbh7u/cisco_clock_issues_caused_by_faulty_intel_atom/
|
# ? Feb 14, 2017 12:30 |
|
Anime Schoolgirl posted:I'm sure Synology, Cisco, and a number of other network vendors will beg to differ. No one was fired for that outside of Intel because no one outside of Intel could possibly have known about the issue at the time of purchase.
|
# ? Feb 14, 2017 12:33 |
|
Platystemon posted:It’s a reasonable tradeoff on the whole, but it’s a little (but only a little) like the FDIV bug: Bullshit analogy. The chips don't produce wrong results. If you run code like this and it's important for you, you already know about it, because you're wondering why your Intel servers aren't running anywhere near the advertised clocks and/or TDP. poo poo will look like this: model name : Intel(R) Xeon(R) CPU E3-1240 v3 @ 3.40GHz stepping : 3 microcode : 0x1e cpu MHz : 2299.999 That's "Intel tax" for you. Now, it'll remains to be seen what the real life performance is, but there's a good chance that twice the amount of 3.4Ghz Ryzen cores leaves this 3.4Ghz-really-2.3Ghz Intel Xeon in the dust, despite the AVX disadvantage.
|
# ? Feb 14, 2017 12:40 |
|
Platystemon posted:No one was fired for that outside of Intel because no one outside of Intel could possibly have known about the issue at the time of purchase.
|
# ? Feb 14, 2017 13:04 |
|
repiv posted:Firestrike doesn't use AVX so you wouldn't expect to see any benefit from Intel's beefier SIMD units there.
|
# ? Feb 14, 2017 13:05 |
|
Skuto posted:Bullshit analogy. The chips don't produce wrong results. it just causes your cooling solution and possibly both the internal (CPU) and external VRMs to go but yeah comparing this at all to a FDIV bug is just intel tramp stamping
|
# ? Feb 14, 2017 13:12 |
|
Skuto posted:Bullshit analogy. The chips don't produce wrong results. The FDIV bug affected like 0.00000001% of operand combinations, and the results were only subtly wrong. The negative perception was far in excess of the concern the problem warranted for most people.
|
# ? Feb 14, 2017 13:12 |
|
yeah sure it's no big deal when hundreds of millions of research equipment in late 90s dollars are on the line with that kind of bug half speed AVX? stop the presses
|
# ? Feb 14, 2017 13:17 |
|
It’s no big deal when hundreds of millions of dollars of research equipment are doing the work of half of hundreds of millions of dollars of research equipment. No one is choosing between half‐speed AVX or erroneous FDIV. They can both be bad things.
|
# ? Feb 14, 2017 13:27 |
|
Speaking of weird outliers:
|
# ? Feb 14, 2017 13:29 |
|
Platystemon posted:It’s no big deal when hundreds of millions of dollars of research equipment are doing the work of half of hundreds of millions of dollars of research equipment.
|
# ? Feb 14, 2017 13:30 |
|
Platystemon posted:It’s a reasonable tradeoff on the whole, but it’s a little (but only a little) like the FDIV bug: It’s a psychological point. Not a technical one. Next time I’ll stick with car analogies.
|
# ? Feb 14, 2017 13:35 |
|
that'd be prudent but the car analogy would be the car going at 120mph tops and the speedometer scale is maximum 150 and the owners manual and sales literature tells you as much, while the other car can go up to 240mph and the speedometer goes up to 300, but veers ever so slightly to the right "desert bus" style https://www.youtube.com/watch?v=nBr7EhL6Jpg Anime Schoolgirl fucked around with this message at 13:41 on Feb 14, 2017 |
# ? Feb 14, 2017 13:36 |
|
Anime Schoolgirl posted:that'd be prudent AMD’s car goes the same speed as Intel’s and costs 30% less. Except if your destination address is a divisible by 486, in which case AMD’s car is only half as fast. To be fair, the manual did mention this, and the guy at the Intel dealership wouldn’t shut up about it. Platystemon fucked around with this message at 13:45 on Feb 14, 2017 |
# ? Feb 14, 2017 13:39 |
|
"aw, this was in the loving brochure too! x86 hovercars were a mistake. i'm selling this and going to buy a POWER hovercar"
|
# ? Feb 14, 2017 13:47 |
|
But what does Kyle Bennet drive?
GRINDCORE MEGGIDO fucked around with this message at 14:22 on Feb 14, 2017 |
# ? Feb 14, 2017 14:18 |
|
GRINDCORE MEGGIDO posted:But what does Kyle Bennet drive? (this is a deep, deep reference)
|
# ? Feb 14, 2017 14:43 |
|
Aesculus posted:How is the quad-core at 3.2ghz doing better than the octa-core at 4ghz per core spasticColon posted:So Intel CPUs are still going to be better for gaming? They're random rear end leaked half benchmarks, just wait for real reviews.
|
# ? Feb 14, 2017 14:46 |
|
Anime Schoolgirl posted:haswell/broadwell-e work at full speed with avx though The numbers I quoted were a Haswell Xeon, which dips far below the rated clock speed. At least Haswell-EP downclocks to the base frequency (so less than 4-core Turbo) too. Depends on your definition of "full speed", I guess.
|
# ? Feb 14, 2017 14:46 |
|
Daily dose of Ryzen leaks! Passmark benchmarks of the R1700X http://www.pcgamer.com/new-amd-ryzen-details-and-pricing-leaks/ Better single core IPC than Kaby Lake — to be taken with clockspeeds: https://notebookspec.com/pc-cpu/AMD-Ryzen-3-1100/254 https://notebookspec.com/pc-cpu/AMD-Ryzen-3-1200X/255 https://notebookspec.com/pc-cpu/AMD-Ryzen-5-1300/256 https://notebookspec.com/pc-cpu/AMD-Ryzen-5-1400X/257 https://notebookspec.com/pc-cpu/AMD-Ryzen-5-1500/258 https://notebookspec.com/pc-cpu/AMD-Ryzen-5-1600X/259 https://notebookspec.com/pc-cpu/AMD-Ryzen-7-1700/260 https://notebookspec.com/pc-cpu/AMD-Ryzen-7-1700X/261 https://notebookspec.com/pc-cpu/AMD-Ryzen-7-1800X/262 eames fucked around with this message at 16:01 on Feb 14, 2017 |
# ? Feb 14, 2017 15:57 |
|
eames posted:Daily dose of Ryzen leaks! Not pictured: The mountain of salt needed.
|
# ? Feb 14, 2017 16:09 |
|
What's the Intel offering at ~$150? Those low end CPUs are looking a little anemic clockspeed wise.
|
# ? Feb 14, 2017 16:19 |
NewFatMike posted:
There is the i3-7300 2C/4T at 4GHz for $150, but you are probably better off with a Pentium G4560 at 3.5GHz for $65 since the Pentiums now have hyperthreading just like the i3s. There is the i5-6400 4C at 3.3GHz boost for $180 too which seems like a better deal than the i3 even with the lower clock speeds.
|
|
# ? Feb 14, 2017 16:44 |
|
Josh Lyman posted:Toyota MR2 I remember this reference but it's been so long I forgot the origin
|
# ? Feb 14, 2017 16:52 |
|
Arzachel posted:Speaking of weird outliers:
|
# ? Feb 14, 2017 17:10 |
|
https://twitter.com/derrickgott007/status/831509055424360451
|
# ? Feb 14, 2017 18:22 |
|
I can't wait for Easter AMD tweets.
|
# ? Feb 14, 2017 18:30 |
|
My Z68 motherboard died. If I can't source a local replacement, a refurb from HK would get here at around the same time as the Ryzen preorders are supposed to start. Tempted to double down and sprinkle some tax return love on AMD. I don't regret my Fury yet, I'll play ginuea pig. I really want to see what this will do with a nice AIO like the H110X or w/e So I guess this but unironically maybe?
|
# ? Feb 15, 2017 06:35 |
|
AMD CPU and Platfrom Discussion: Threadripper? I Hardly Know Her!
|
# ? Feb 15, 2017 06:38 |
|
RyuHimora posted:Ubisoft games are well known for performing well with high clock speeds vs IPC. Far Cry 4, for example, scales almost perfectly with clock speed, not IPC. They do this specifically because its less work to make games perform well cross-platform. Then, of course, there are outliers like AssCreed 4. Ubisoft is not very consistent. (as shown in your post by 5960X still getting the highest performance) That's not the case with Watchdogs 2. It scales so well with extra threads that a 2600k and 8350 beat a 6600k. Aesculus posted:How is the quad-core at 3.2ghz doing better than the octa-core at 4ghz per core The prime number and physics benchmarks mostly scale off the memory latency. A guy on the anandtech forums posted his scores with various memory setting and going from 18ns to 9ns got him a >75% increase in those two. Basically: Gwaihir posted:They're random rear end leaked half benchmarks, just wait for real reviews.
|
# ? Feb 15, 2017 08:11 |
|
PCPer + Kanter + Zen Discussion: https://www.youtube.com/watch?v=plXoBXFodHI
|
# ? Feb 15, 2017 11:15 |
|
|
# ? May 30, 2024 14:18 |
|
I really only care about increasing my FPS in Overwatch. Currently I have an i5 4570, 8GB DDR3 1600, and a GTX 1060 6GB. My modest setup averages roughly 120 FPS with the "Ultra" graphic preset. Would a 4C/8T or 6C/12T Ryzen with whatever budget mainboard and DDR4 RAM get me considerably more Overwatch FPS?
|
# ? Feb 15, 2017 17:06 |