While that may be true that particular benchmark is just a troll by WCCFTech which is known for posting half truth articles for page hits and riling up the comments section.
I think it’s best if we all all just wait and see what happens. Just based off the Skylake-SP results I don’t think it;s going to be that cut and dry.
This same story played out with Ryzen standard though, a supposed leak showed the chip performing quite poorly compared to Kaby Lake on Cinebench (despite having eight cores). The actual results showed said leak couldn’t possibly be further from the truth.
Power consumption numbers from the same review.
X2 Xeon 8176 - Idle 190 watts, POVRay 453 watts.
X2 EPYC 7601 - Idle 150 watts, POVRay 326 watts.
So more performance, more cores, more memory, less power at half the price. Pretty compelling package.
Fair observation, but considering that the high core counts have a higher max boost, it seems to be a limit that AMD have intentionally placed.
Intel compilers don’t cripple, they just don’t apply optimisations to AMD CPUs. Someone recently spoofed Ryzen CPUs to read as Intel and found performance boost while running certain benchmarks.
Strangely This time Intel i7 7820 X is arround 10% faster with 47% higher TDP and more than 100$ ryzen 1800x price.
INTEL here isn’t the best CPU.
How 7820x will perform against 1800x at same TDP ? 140/140 or 95/95.
More info from AMD’s Youtube channel:
1920X 12 core / 24 thread 3.5/4.0 Ghz = $799
1950X 16 core / 32 thread 3.4/4.0 Ghz = $999
Launch early August.
Cinebench R15 MT scores:
7900X = 2167
1920X = 2431 (+12%)
1950X = 3062 (+41%)
Highly impressed to be honest.
That is impressive, more so based on pricing…I will still wait and see. I am leaning toward AMD, but I never know what is in the cards…and re-reviews are what really count for me…reviews from owners after 6 months or so of usage…there is never any real reason for an end user to race out and make a rush purchase based on hype…I am being honest though when I say…I am glad AMD has finally come back to play hardball.
Indeed impressive results. that 12 core 24 thread is faster them my dual Xeon 2687w. (about 2150 points)
Power wise it was reported on few sites the TDP of the TR 1920\1950 to be in the 180w range. that is about 1/2 of my dual xeon setup (150w each)… impressive progress made in the past few years.
Now a bigger question. why would we need these highly threaded CPUs? wouldn’t a single GPU outperform this in Cycles?
Most people complain about GPU VRAM sizes, lack of support for some features and nearly locking the PC while rendering (I think this is being looked at).
And Cycles rendering is just a part of what you can do with a PC.
Alright, I’m getting tickets for the hype train. It’s 3.4Ghz on the 16-core btw.
Fixed.
Something, something 10 chars.
It’s much easier to program a CPU than it is a GPU. Not every demanding computation can be done with GPU or are done with a GPU. This is not CG, but for example compiling big programs takes quite a while, and is not done with a GPU.
Many physics libraries do not run on GPU. Many simulations, algorithms do not run on a GPU. I wonder how much Houdini contains GPU code. I would bet it’s mostly CPU.
4 Ghz boost is pretty good for a chip with 16 cores, and considering the reality of how core counts scale up, the speed boost is pretty decent as well.
EDIT: Found this image of benchmarks among the latest chips (from PC Perspective linked to on the Modo forums).
What’s amazing is how the 16 core is almost a linear 100 percent performance increase over the top Ryzen chip (which leads to the price making perfect sense as well). Looking at it, I think the chips really do deserve that name (they will rip through rendering tasks and they ripped the performance crown away from Intel and at a lower cost).
That poor Kaby Lake 7700K :spin:
Intel really got caught with their pants down this time. Good times indeed, and it reminds me of the good old days when AMD and Intel were competing head-on.
And don’t forget that the 1950X has 64 PCIe lanes to play with (which is crazy - I believe the top model $13,000 Xeon has 48?): 1TB RAM anyone?
It’s brilliant value for money. Game changer. Poor Intel…
I’m in agreement Herbert123.
Intel got to comfortable and didn’t learn from its own history.
However these systems are more for high end users, and not casual gamers. AMD still has some work in front of them.
Main thing though, competition is starting and Intel is showing that it is afraid (latest slides against AMD that it was just a glued approach and few other points).
If Intel adjusts their i9 series pricing then we will see some improvements.
I just hope AMD can get faster CPU’s out, not more cores. Either way, good job AMD.
The allegations that the i9 chips have major overheating problems (even with stock clocks) have been confirmed (with data to back it up).
This is a big deal for enthusiasts wanting a beastly machine for things like 3D rendering, because it shows that the overclocker who made the initial video was not just someone who didn’t know what he was doing.
What’s even more sad here is that a large chunk of it is apparently due to Intel using the TIM instead of actual solder so as to save a few pennies per chip (not to mention other noted issues by readers such as there being too much distance between some of the cores and the interface).
I did read on another website (the Modo community again) that the issues could potentially be resolved with the right motherboard, the downside though is that the good ones are very expensive.
omg i just thought out the long term future potential of such a system, i mean possible future upgrades with all those PCI lanes, and theres always a chance in the future the DDR4 RAM it uses might get really cheap and you could upgrade to a TB, or you could line up the PCI slots with 8 GPUs or something.
ok this is a list price for an AMD EPYC 7401p 24cores 48 threads, 3.0GHz 170watts, $1075. Its from Anandtech which has detailed benchmarks like SPEC view perf and all the server related tests. Well the EPYC cpus came out absolutely dominating the Xeons specifically at raytracing! upto 50% performance leads…
About the EPYC 7401p above, it is a Single-socket SKU. Competing wiht the 12 core Xeon 5118s AMD has 8 extra cores a less complex server board, much more PCIe bandwidth and a lower TDP. AMD should have a serious cost advantage on paper.
Another potential area of trouble for Intel, there’s an alleged report that Intel accidentally revealed the all-core clockspeed for the upcoming Intel 12 core i9 chip.
If this is true, then the 12 core chip will run at only 2.9 Ghz when under full load (which gives a major possibility of a 12 core threadripper actually beating it in heavily threaded tasks). I guess it might be surprising considering just how hot the current i9’s perform (and raises some serious questions of whether Intel can deliver a good quality 18 core chip without major design changes).
The EPYC CPUs actually performed poorly in several integer tests, like database. That’s much more important than Floating-Point or Raytraying for enterprise customers.
The 7401p has a base clock speed of 2Ghz, not 3Ghz. The Threadripper with 16 cores at 3.4Ghz is almost certainly going to outperform it everywhere. Don’t get one of these for a workstation.
About the EPYC 7401p above, it is a Single-socket SKU. Competing wiht the 12 core Xeon 5118s AMD has 8 extra cores a less complex server board, much more PCIe bandwidth and a lower TDP. AMD should have a serious cost advantage on paper.
EPYC is not actually competing with the 8xxx or even the 5xxx, those are 8-socket and 4-socket CPUs respectively and they’re priced much higher than Intel’s 2-socket variants. In the Anandtech tests the old 26xx CPUs are the closest to EPYC (2 socket) and there the price advantage isn’t that great anymore (30-40% in the better cases).
I’m quite surprised that Threadripper will run at 3.4Ghz base clock (unless AMD is doing misleading advertising here). Maybe the multi-chip design massively improves thermals.
