Separate names with a comma.
Discussion in 'AMD Processors' started by Kato1144, Jul 21, 2017.
Lol I was getting ready to post this.
The 7700k still looks good but that 7800X just plain sux compared to the 1600 considering the 50% cost of the AMD against it.
7800X or his motherboard, though?
Because motherboards that burn down socket are by all definitions shit.
I'm not quite getting what you are saying here, I'm not sure what "burn down socket" means as far as MOBD go, I have heard some of the X299 MOBD are said to be holding back performance of the CPU, to what degree i have no idea
Spoiler alert, here is what happens to his 7800X after all that:
wasn't talking about that, only the graphs in that vid, performance.
If motherboard does that to CPU i am not trusting that CPU performing properly in the first place.
Considering that they tested several GPU-limited games and some few that were frame-capped, it is not surprising that overclocking the 7800X by 34% only brings 3% extra performance compared to stock 7800X. Also we know that SKL-X was run with stock clocks on the mesh interconnect, whereas the RyZen chip was tested with IF interconnect overclocked by using 3200MHz RAM. Overclocking the mesh on SKL-X can improve gaming performance by double digit percents
lol, there is the excuse we are looking for. Got to say I am loving Skylakes recent releases, shoe on the other foot and all.
Looks like a bent socket pin made bad contact with the CPU. Not sure he can blame overclocking for that one...
oh wow, that is burnt
Mentioning that overclocking the 7800X by 34% only brings 3% increase performance because the CPU is being bottlenecked by GPU-limited games and frame-caped games is not an excuse, but the explanation of why the R5 managed to be closer in performance.
Mentioing that the R5 was tested with overclocked interconnect, whereas the i7 was tested with stock settings on the interconnect is not an excuse, but a further explanation of why the R5 managed to be even closer in performance.
Also I doubt any will be purchasing the X-series chip only for gaming. The target is people that does professional stuff and occasionally games. On professional tasks the six-core 7800X can beat the top 8-core RyZen and run circles around the R5.
We already saw that overclocking the mesh didn't really do much from Hardware Unboxed video but I guess you don't trust them.
There is a reason why 1600 is the best selling AMD cpu. Heck its selling more than 7600K (same price bracket) at mindfactory (one of the biggest e-tailer in eu area).
I did not notice if they synced the memory with it, you know (oh, they did run OCd memory but forgot to sync them...).
I have to admit syncing interconnect with memory is the one of few things Ryzen did right (well, SkyLake did it right in it's own way running it at 4Ghz by default...)
Seems like a strange comparison given the platform cost differences.
The 1600 Non X is a really fast processor, even at stock. It is even faster when paired with really fast ram. It is good to see AMD competing again, kicking butt and taking Intel names. Finally, for those Intel fans, something to be happy about because otherwise, you would have had a new 7900X at $2199 because Intel......
As mentioned in a former post, they used several GPU-limited games and some few that were frame-capped, overclocking the mesh is not going to do anything in such situation. Otherwise overclocking the mesh can increase fps up to 25%.
The 7600k is #3 in Amazon wordlwide sales. The 1600 is #5
And how long has each been out...
I tend to agree with guys who have been somewhat reliable in the past when doing hardware reviews and not just some random picture from god know where (4chan?). Then again, I do wanna see more tests with mesh OC but nobody hasn't really done them that much and we don't know if the mesh was overclocked on that 30 games test suite.
Oh well, I'm still gonna say that its not the mesh which is a problem with Skylake X. Its the new L3 cache hierarchy which just doesn't work well in gaming scenarios (a lot less L3 per core, victim cache change, cache not shared meaning node jumps through mesh when accessing L3 residing on other core) and the bigger L2 just doesn't make up for it.
I am too lazy to give better explanation.
Since when did Amazon.com become a worldwide seller? Last I checked, .com is for north america. They have completely different best seller lists on .de, .co.uk .fr, and .it.
Heck how are they even calculating those best sellers? At least with mindfactory you can see how many units each of the sku's has sold.
At what TDP though? While performance is better it's doing it at twice the TDP.
Nice, GO AMD.
He said the socket had no bent pins in the twitter thread though. He couldn't identify any sort of physical damage other than the arc marks on the CPU pad.
It couldn've been skin oil too from handling of course. Or dust. Fairly unlikely though.
Also fairly unlikely that the CPU was fried by a mild overclock It's a mystery for sure.
Rumor mill is saying the socket is just too much out of spec for the OC'd 6C/12T chips though. We'll have to wait and see if that is true.
ICCin_max is 190A by the spec. No way he would hit that with such overclock.
This is not about who is more reliable. If you are testing in a GPU-bound situation then it doesn't matter if you overclock the core by 30% (as they did) or ovcerclockf the uncore by 20% or just replace your CPU by one 10x faster. The fps will be about the same in all cases.
I know that Amazon.com ships products internationally to over 100 countries. I don't know when they start to sell worldwide. Local domains help customers cannot manage foreign languages and/or want to avoid customs. Also I consider that .com bestselling list provides a better estimation of overall sales than .de, .uk, or .it. for demographic and sociological reasons.
Marketing TDP? Real TDP? Effective TDP?
You tell me?
I really like this guy.
So similiar performance, less power, and half the cost. Damn fine job by the Ryzen 1600.
Prime95 works as a power virus for Intel chips, but doesn't work as power virus for RyZen. RyZen consumes about the same power on Prime95 than in ordinary workloads as Excel or Luxmark. Also Broadwell-E and SKL-E chips have 256bit and 512bit FMAC units, whereas RyZen only has 128bit and that is reflected in the power consumption. That extra power consumption in Prime95 (which has AVX512 support) is accompanied by a much higher performance by crunching wider vectors. For instance the i9 is consuming about 80% more power than the R7-1700X due to the use of 512bit vectors, but the performance that brings AVX512 to the i9 is about 3--4x bigger than for the R7
On ordinary workloads RyZen is less efficient, because the '95W' is a marketing TDP. Here you have power consumption values for Excel and x264
I am sure by the amount of posts that dont show and quotes that dont show, Juanrga must be in here stomping his feet and showing the same 3 charts he has. Plus you have that nice IPC reduction with the Intels with mesh architecture which sure didn't help in benchmarks. Also I am sure that somehow Prime95 is a power virus again and it's broken on AMD Ryzen chips. Not broken man it's called AVX 512 and unless you limit it in the bios then it's working the way Intel wants it to, live with the power useage. Some just need to adjust to the fact that this new release and platform is kind of a giant bummer for Intel as the performance just is not there and the power usage is high. Cant wait to see the higher core counts from Intel and the massive disappointment they will be running at such low speeds.
Skylake-X has some real issues with gaming. Although many of the games were cpu limited, there were plenty that were not.
The 7800x had a 35% o/c while the 1600 only had a 25% o/c yet the 1600 had larger gains on all non-cpu limited setups save Resident Evil.
There must be alot of Broadwell boys chuckling right now.
On this random chart, it helped Tomb Raider by 13% or so. It helped Total war only 7% vs XMP.
Curious that it was tested at only 4.0 Ghz. Why not 4.7 ghz AND a mesh O/c?? I would love to know the amount of heat and power for that setup!
Pretty sure the guy thay did the mesh o/c mentioned the heat went up so you can give up on that garbage already.
Juangra, you know it is a garbage post when not even Shintai likes it. TR goes from 140+ fps to 150+ fps. Yippie! Civ 6 shows lower performance over xmp in the left chart but higher in the right....
It's not a "power virus" it's dependent on workload. As you have pointed out Intel chips have hardware for AVX 512. The problem with this logic is that you can't sit up here and say OMG the Intel chips are great because they can do X and then call their power usage a virus when the power details come out about it. The problem with using Excel to determine power is that the chances that you are you going to run an endless loop on Excel is zero, or even an endless loop integer like that is slim. Hell there is little reason to do so. Endless loops in floating point? Well those are everywhere and way more common.
This whole marketing TDP kick you are on is retarded. TDP's Intel or AMD are exceeded ALL OF THE TIME. It's been that way since the dawn of time. I've been working with computers all of my life and even server configurations if given the right workload can and often do exceed the supposed TDP.
AMD Ryzen 5 1600 vs Intel Core i7-7800X: 30 Game Battle! Hexa-Core Head-to-Head
Make sure that you read the conclusion.
If you care at all about value, the Ryzen 5 1600 is clearly the way to go. This is why we recently named it the best value performance desktop CPU. It was unlikely that the Core i7-7800X was going to change that, but we hoped the performance would at least be a compelling reason to buy Intel's new six-core processor.
Ryzen will hit 4GHz with the box cooler but it will be a more mild experience with a $20 aftermarket cooler like the Cooler Master 212, so keep that in mind. The 7800X on the other hand cannot be overclocked to 4.7GHz using a 240mm AIO closed loop solution. Instead, it required a $380 custom loop setup to achieve that result.
It is all simpler than that. One cannot take power measurements on Prime95 and combine them with performance measurements on Cinebench or similar benches to obtain efficiency, because Prime95 uses 512bit, whereas Cinebench doesn't. To measure efficiency one has to measure both power and performance in the same workload, was it Cinebench, Excel, Blender, x264,...
The same reviews that found that AMD did cheat the TDP for RyZen (e.g. 95W marketing label correspond to real 128W TDP), found that SKL-X real TDP satisfies the official value.
If you insist on repeating the same mistake again and again, then be prepared to receive the same correction.
No one is blaming Prime95 for the high power consumption when using AVX512. That wasn't the point. We know that 512bit unnits are power hungry because they are four-times wider than the units on Zen and two-times wider than the units on Broadwell. The point is that some people is doing incorrect claims about efficiency because measures the power consumed by 512bit units, but doesn't measure the performance that they generate.
and you totally miss the point as well. It speaks to ability to maintain clocks, which this far skylake can not do in the HEDT platform. Obfuscate all you want, come Aug15 or so we may see a new performance leader in that bracket.
That's weird because they blow past their TDP on other workloads too. Feel free to look up the "marketing TDP" for those.
When you get done with that here's Blender:
As I said before. This happens ALL OF THE TIME and it greatly depends on workload.
Sure they did, or are you talking about the one that used Excel (WTF?) to determine power consumption?
Values as 245W and 253W are total platform consumption, not the power consumed by the CPU alone. Thus official TDP ratings aren't being violated. Toms measured 250W on the CPU when overclocked it to 4.5GHz and with AVX512 offset disabled. Correcting for clocks we obtain for stock settings
250W (3.3/4.5)^2 = 134W
which agrees with the TDP. In fact, AT measured 149W for the same CPU on stock settings, and the 9W difference being easily explained by measurement errors (including loses from circuitry). However, the difference between 129W measured for the 1800X on x264 and the official 95W is not the result of small uncertainties in the measurement apparatus or small loses from 95% efficiency circuits. This 36% gap between measured TDP and marketing TDP is a result of the chip violating the marketing value.