*sigh* if you're going to get anal over specifics, that wasn't 100% of it's frequencies. Look, the point is you yourself said "the 7980XE is a lower frequency processor" and you just proved yourself right here and now that it is the same as the 1950X. Almost no desktop CPUs use the base frequency, especially when paired with 3rd party heatsinks (which the 7980XE requires). In other words, base frequencies are irrelevant to consider. In the event cooling/power isn't enough, the frequency range isn't so black and white; the frequencies can be in-between the base and max turbo. We all know that the turbo speeds are higher on fewer cores (which is specifically why I brought up the improvement in single-threaded performance) and the IPC of the two is very similar, where an additional 4 threads would make a substantial difference at the same frequency. Again, you have shown they both have an all-core freq of 3.4Ghz. So what are you not getting? It doesn't matter who does what better. HT isn't that much worse than SMT. HT is substantial enough to give these results. It's not like the 7980XE is lightyears ahead of the 1950X. The results seem right where they should, and are perfectly reasonable for what the CPU is. I don't see what there is to argue about. I never said the whole CPU goes to 4.2Ghz, so what are you ranting about? What about these results don't look correct? You said yourself the IPC is similar. The clocks are similar. The 7980XE has 2 more cores, and an appropriate performance increase as a result. What am I missing?
The boost clock is limited by the TDP, not the cooling performance. And I'm not sure where you get the idea that the base frequency is irrelevant - my 1950X almost always runs at or near the base clock of 3.4 GHz when it's computing (it shows roughly 180W at full load so can't really clock much higher). Considering that the 7980XE has a lower TDP (165W) and two more cores, I'm willing to bet that it will run nowhere near the all-core turbo when put through the same workload. The article mentioned that the CPU ran at 4.2 GHz during Cinebench. Considering that Cinebench utilizes all cores it is extremely unlikely that it turboed to that speed, even on one core (there is simply no room in the TDP). Like I said before, it's probably using enhanced turbo or was overclocked to 4.2 GHz.
All core turbo is pretty much implemented on every high-end motherboard and enabled by default. The only time intel CPUs really run at base clock is under very high temps or AVX loads(mainly 256/512 loads), which causes a ton of heat.
In my experience, Intel CPUs readily utilize the maximum boost clock per core configuration where possible. Meanwhile, AMD's TDP sensor tends to be inaccurate (in the case of your CPU, maybe your power profile has something to do with it?). As I said before, the clock speeds can be anywhere between the base and turbo freqs, and Intel's got a much farther range. In the event TDP became an issue to sustain the 7980XE's 3.4GHz all-core turbo, it likely could've dropped to around 3.0. An additional 2 cores would still give it a performance lead. I think it is safe to assume it wouldn't have exceeded wattage (or thermal) limits during the single-threaded tests. And sure, both of these CPUs are newer generations than the ones I am referring to and have MANY more cores, but ultimately the same point remains: the results aren't surprising. They're not better-than-expected, they're not worse. They're right where they should be. I feel we're all getting a little too carried away with speculations and suspicions. I think we can all agree to take these results with a grain of salt, and that the 1950X is obviously a more sensible product to buy.
I'm just trying to square the math here. Having two extra cores @ 3 GHz would just about cancel out the 1950X's higher clock speed. However, the Cinebench score was about 37% higher than the 1950X, which is not possible with just two additional cores. This is why I suspect that the CPU was running with enhanced turbo or was overclocked - to force 4.2 GHz on all cores. Doing some basic math: 3.4 x 16 = 54.5 4.2 x 18 = 75.6 (75.6 - 54.5) / 54.5 = 0.3871 = 38.7% This almost exactly matches the performance difference in Cinebench between the 1950X and the 7980XE, meaning that the latter was likely overclocked. Not that there's anything wrong with using overclocked results, just that it isn't representative of the stock performance (and I doubt many HEDT customers would overclock - I run my 1950X at stock since I run it full load at all times and need it to be stable).
That 'basic' math is only applicable to similar CPUs. 1950x architecture is not at all similar to intel
Nice Cinebench scores. That said, to state the obvious, those 7980 Cinebench scores are NOT at the base clock of 2.6 Ghz, not a chance.
The IPC differences between the CPUs is not that much different, so that "basic math" works just fine as a general idea, but not an absolute. There is, in general, a 1-5% IPC difference depending on which test you are doing. Which is close enough to get a "general" idea of what performance difference there would be at different frequencies and cores between current intel and current AMD, they just won't be "absolute" like many others are saying that others are just trying to dust off, these scores do not look like the 7980XE at base, or at all-core 3.4Ghz boost, but rather look like it's being overclocked. Why people can't see that is beyond me. Ultimately, we'll see if these scores hold true when guru3d makes a review about it, but i'd bet it does not, because these scores simply do not make sense with the information that intel, themselves, have given us. If it is correct, then intel literally pulled extra performance out of thin air. And, as others have stated before as well, their own pictures prove this: If the CPU is at 100% utilization, the max frequency for all-core boost is 3.4Ghz, so if anyone is trying to say "well maybe it's not overclocked and it really does perform this good" my question is: How do you explain the picture they provided, which clearly shows, 100% utilization @ 4.17Ghz, far beyond the max all-core boost of 3.4Ghz? Refuting this fact shown by the people who did the tests is quite frankly madness.
clock for clock, intel has a 10% advantage in single core performance. Comparing 7700k to 1800x in cinebench with both set to 3.5ghz Coffeelake will further that maybe another 5% give or take. As for the rest, don't know why you're directing that at me. It's clearly known that most highend motherboards have All core turbo boost enabled by default. That means under load it will put all cores to the max single core turbo boost. Nothing to do with Intel. Since 7980xe has a single turbo of 4.2ghz, it's safe to say that this setup is utilizing all core turbo setup.
If enhanced turbo is on by default and all cores clock to the max single-core turbo, then it will blow the TDP completely out of the water. Are you seriously saying that all high-end Intel motherboards are overclocked out of the box?
That is correct, and no it does not *blow* it out of the water. Not really overclocked as it's not *above* the rated max clock speeds. Totally dependent on mobo brand, but higher end boards allow a higher than default rated TDP based on cooling/temps. They only time TDP is *blown* out of the water is with increased voltage.
1-5% on average, 10% in extremes. This has been proven multiple times, i'm not entirely sure why you're trying to spread misinformation. In many cases the IPC difference has been in AMDs favor. Hence, AMDs and Intels IPC being close enough to do basic math for. I was not intending to direct it at you, it just so happened it was after what i quoted from you that i stated it. It is 100% overclocked because at that point you are using the processor beyond what is intended, which is overclocked. Intel very specifically have stated what their processor is rated to do if thermals are good, depending on what cores are being utilized. Anything beyond that, is overclocked. All-core boost of the 7980XE is 3.4Ghz, that is PER INTEL. Anything beyond that is an overclock. It does not matter that your motherboard is doing it automatically for you, it's still an overclock. Anything beyond these numbers are overclocked: This is fact. This is not opinion. This is quite literally, fact.
Do you even read what you type before hitting "post reply"? Every single thread about Intel is immediately flooded by AMD fanboys talking **** on Intel and anyone who buys their products instead of AMD, and not a word out of you about their trolling, fanboying, etc. Yet when one person fires back at AMD users... you cry foul. That is some of the most hypocritical nonsense i've seen in a while, and i've seen a lot. This site has gone from a great place to get information on PC hardware to a cesspool of **** talkers making it so you have to go through 2.5 pages of crap just to find one tidbit of information. I really wish HH would crack down on this nonsense. Once your website's reputation takes a dive, its 10x harder to recover from it. I could name quite a few great examples but i've distracted from the topic at hand long enough.
Regardless, it's an enhanced turbo and pretty much all mainboard manufacturers utilize it as it doesn't affect temps or tdp much. You still can't directly compare intel and amd with that math. Cinebench is strictly IPC performance, with every other test you are adding other variables which will cause different results. That's why it doesn't make sense to use blanket statement math. Too many factors to account for. Just in games, 4 thread of 7700k vs 4 threads of ryzen can yield 40% difference in frame rates. But of course this attributes to other factors including ryzens poor memory performance(latency or whatnot), but that's another topic. I'm not spreading misinformation either, data is clearly out there. Heres a review showing 4cores disabled both at 4ghz http://www.zolkorn.com/en/amd-ryzen...i7-7700k-mhz-by-mhz-core-by-core-en/view-all/ "we have seen RYZEN is behind Kaby Lake about 10-15%"
For one, this website does not utilize Ryzen the best that it can be. It's already a pretty well known fact that Ryzen does rely on RAM to a pretty decent degree. I would say that not going above 2667Mhz for the sake of that's what all is supported could be logical, but it doesn't even do that, it runs in at 2400. Realistically, if both the Intel system can to 3200, and the AMD, it should be done at that. If you're going to gimp one system and not the other (since Intel does not rely so much in ram frequency) to state that the IPC is much different, that's pretty messed up. But regardless; From your own website you posted: "SuperPi 1M results. Percentage wise, the difference between the two CPUs is approximately 11%." Win for Intel IPC wise "Cinebench results. In single-threaded performance, the 1800X is around 6.8% slower compared to the i7-7700K" "Geekbench results. Here we see the i7-7700K beating the R7 1800X in both single- and multi-threading performance by around 12% and 10% respectively." Personally, this is a bit screwed up test if somehow the 1800x loses out to the 7700k in multi-threaded performance by that much, let alone at all. "RealBench 2.44 results. less than 1% difference favoring the i7-7700K." Very similar IPC "FHD Benchmark results. expected the i7-7700K to win here, but the Ryzen chip performed 3.2% better." Better IPC then Intel "Handbreak results. Kaby Lake architecture performs slightly better at around 5% to 6%" 3DMark Time Spy about 15% better for intel Fire Strike Ultra both of systems can work out the level is very similar in every part 3DMark firestrike physics score puts AMDs processor about 6% faster then intel 3D mark Skydiver Intel performs about 8% faster And i'm not going to get into the gaming aspect since that's not even what we're talking about here as well as has been shown that the memory speed greatly helps, which they did not do here, as well as it has been clearly shown that games that actually understand how to use a ryzen processor, get patched, etc. lowers the difference by a ton. Again, something they do not take into consideration. So that means: 3 tests performed 10% or better for intel 2 tests performed between 6-8% in intels favor 1 test performed between 1-5% in intels favor 2 tests performed about equally 2 tests performed in AMDs favor. Or in other words: 20% of the tests had AMDs performance beating Intels IPC 20% of tests had AMD and Intels performance at almost equal performance 50% of tests had AMDs performance at an IPC level of 5% or better when compared to Intel and 70% of tests had AMDs performance at an IPC of 8% or better Whereas only 30% of tests had AMDs performance at above 10%. And somehow 30% of tests being above 10% somehow makes AMDs overall IPC difference between 10-15%? By that logic i could say that due to the results showing 40% of tests show that AMD is either equal or better then Intels IPC, then that means AMDs overall IPC difference is 3% faster then Intel. I'm sorry, but there's no logic in that. Let alone the fact that these numbers are based off of gimped ryzen results due to using a low frequency memory speed. Either way, i'm done with this conversation since it clearly is about spreading misinformation for the sake of wanting something to be one way, rather then reality.
LOL. I use enhanced turbo myself on my 4790K and it has a significant impact on TDP (100 watts vs the default 88 watts). Note that this is only a 200 MHz difference from the all-core turbo across four cores (4.4 GHz vs 4.2 GHz). With the 7980XE, we're talking about a whopping 800 MHz difference across 18 cores (3.4 GHz to 4.2 GHz)! Like I said before, this would blow the TDP completely out of the water (there's no other way it could be). You have an Intel CPU as well, so you can test this out yourself (turn enhanced turbo off and on, and see what the difference in TDP is).
I didn't say it averaged out across every test because every test isn't strictly IPC performance. You're reading/believing what you want to believe. There are several factors in benchmark results. Other tests may be more dependent on L1/L2/L3 data size and latency. Cache can negatively impact performance if a data set is not able to fit inside. Ryzen has significantly more cache for L1/L2 and L3. 7700k has deficit on cache and certain benchmark that benefits from increased cache sizes will see Ryzen catching up. Cache is not directly related to IPC performance. IPC is correlated to pure integer performance which intel like i said before, is 10-15% faster. You're arguing about cinebench and how the score doesn't seem realistic and I return with my first post showing cinebench both CPUs at 4ghz is over 10% difference. Cinebench is strictly based on integer performance, that is why single core performance is much higher than ryzen in this test. That's why you saw bulldozer beating the 2600k/3770k in cinebench due to having more integer cores; it did not matter that it had half the FPU. In most real world tests, bulldozer wasnt even close. and yes intel does benefit on memory, depending on the application. No different at the same voltage for my setup. Depending on manufacturer auto voltage may scale up automatically based on core frequency
