Please reply in that thread about how to do a leak testThanks, don't know how I missed that thread. Don't recall if I saw it mentioned, but since it's your first custom loop, I'd do a leak test before plugging in the power to the motherboard.
OMG, people are complaining about gaming performance.
Some people only want to have one desktop computer in their house, so it makes sense to see if this CPU can fulfill the need for such folk.
Some people only want to have one desktop computer in their house, so it makes sense to see if this CPU can fulfill the need for such folk.
AMD themselves marketed the X series to gamers.If you own a system like it, chances are you have a daily driver or dedicated gaming box too (ex: i7 8700K or Ryzen 2700). Gaming is not one of it's main purposes by far
Yes agree, WX series is not for gamersAMD themselves marketed the X series to gamers.
If AMD marketing doesn't know any better, why expect this from consumers?
If you own a system like it, chances are you have a daily driver or dedicated gaming box too (ex: i7 8700K or Ryzen 2700). Gaming is not one of it's main purposes by far
AMD has made the right choices for now, but two years into the future when intel has icelake SP, EMIB, likely new protocols and topologies, AMD is going to need a better system.
(snip)
These are the instructions for a simple printf program. I'm sure you don't need this lesson as you seem informed. Now extrapolate this to a benchmark application.
Analysis of an execution flow :
https://en.wikipedia.org/wiki/Cycles_per_instruction
void InterleaveBytes(__m128i *in, __m128i *out, __m128i *scratch, long count, unsigned long passes, unsigned long offset)
{
__m128i *to, *nextTo;
__m128i *from = in;
if (passes-- & 1) { // odd
to = scratch;
nextTo = out;
} else {
to = out;
nextTo = scratch;
}
__m128i *end = &to[count];
do {
*to++ = _mm_unpacklo_epi8(from[0], from[offset]);
*to++ = _mm_unpackhi_epi8(from[0], from[offset]);
from++;
} while (to < end);
from = &to[-count];
to = nextTo;
do {
end = &to[count];
do {
*to++ = _mm_unpacklo_epi8(from[0], from[offset]);
*to++ = _mm_unpackhi_epi8(from[0], from[offset]);
from++;
} while (to < end);
end = &from[-(count>>1)];
from = &to[-count];
to = end;
} while (--passes > 0);
} /* 40+128=168 non loop operations
10m passes comes in at
0.48 sec so 10m/0.48 = 21m
4g / 168 = 24m
87% efficiency
*/ /* 24+64=88 non loop operations (sans byte bit swap)
10m passes comes in at 0.3 sec so
10m/0.3 = 33m
4g/88 = 49m
67% efficiency
*/Gaming's not its main purpose, sure. It's also not Skylake-X's main purpose, but that didn't stop people raking it over the coals when its gaming performance turned out to have dropped all the way back to Ivy Bridge levels in certain cases.OMG, people are complaining about gaming performance.
In my case, I wouldn't physically have the room for a dedicated gaming box. Meaning that if I were putting together a one-size-fits-all rig, my choices would boil down to the 2950X (good, cheap all-rounder, but not really the performance leader in anything), 7980XE (very solid, consistent performance, plus AVX-512 support, but ridiculously expensive and less PCIe connectivity) or 2990WX (crazy fast in multi-thread situations, but wildly inconsistent in single/low thread).If you own a system like it, chances are you have a daily driver or dedicated gaming box too (ex: i7 8700K or Ryzen 2700). Gaming is not one of it's main purposes by far
The 1950x and 2950x are the 'all things to all people's' perf/$ champ though...sure it is not the best at anything, except perhaps it's value...but it is best all round workhorse for your money, with no real weaknesses in anything.Gaming's not its main purpose, sure. It's also not Skylake-X's main purpose, but that didn't stop people raking it over the coals when its gaming performance turned out to have dropped all the way back to Ivy Bridge levels in certain cases.
In my case, I wouldn't physically have the room for a dedicated gaming box. Meaning that if I were putting together a one-size-fits-all rig, my choices would boil down to the 2950X (good, cheap all-rounder, but not really the performance leader in anything), 7980XE (very solid, consistent performance, plus AVX-512 support, but ridiculously expensive and less PCIe connectivity) or 2990WX (crazy fast in multi-thread situations, but wildly inconsistent in single/low thread).
There's not really an obvious "best" CPU out of that line-up, and I'm sure AMD would have loved to have marketed the 2990WX as the CPU that can be all things to all people - which, let's face it, is how Intel markets the 7980XE, despite it being by no means their best gaming chip - but clearly there's too many pitfalls as it is for them to do that.
You could also argue that having to build a second gaming box would eliminate its price advantage over Intel's line-up, but then again this thing's real competition will be the rumoured LGA3647 i9s, and right now we have absolutely no idea how they'll be priced or how well they'll do in games, so it'll still probably come out in front on that count.
Which is why I used the term "Some people" and not something like "Most people".
I think the other point still stands though: Once the buyer has built their 2990wx system they've spent so much money that a top-of-the-line graphics card suited for gaming and thus gaming at high resolution seems reasonable. And at that point aren't the video cards the limiting factor?
I think the other point still stands though: Once the buyer has built their 2990wx system they've spent so much money that a top-of-the-line graphics card suited for gaming and thus gaming at high resolution seems reasonable. And at that point aren't the video cards the limiting factor?
In addition,at those prices why not just get a gaming-dedicated cheaper computer or a console?
If money is no object buy an epyc or something, none of the draw backs, loads of everything.The 2990wx has a game mode which can bring it down to one die with 8c/16 where it performs about equal to a 2700. That should be fine for those who want to play the occasional game on a workstation system. . .
Epyc's clocked a lot lower, though. Besides, the overheads which are hurting performance in certain scenarios on the 2990WX might actually be worse on Epyc, since there's four memory pools for Windows to juggle instead of two.If money is no object buy an epyc or something, none of the draw backs, loads of everything.
Also, the linux performance is insanely good, I mean it crushs everything on all then benchmarks., and thats what I will be running after its all setup and OC'edEpyc's clocked a lot lower, though. Besides, the overheads which are hurting performance in certain scenarios on the 2990WX might actually be worse on Epyc, since there's four memory pools for Windows to juggle instead of two.
Seems like the issue lies with the way they configured the CCX cores to access the memory, only 2 out of the four can access the memory directly and this causing an issue. I wonder why then did they not make all CCX access the memory? Can someone with an expertise in CPU design answer what would be the drawback to allowing the other CCX cores to access the memory as well? Does it increase the die size, is there some other penalty?
Also, the linux performance is insanely good, I mean it crushs everything on all then benchmarks., and thats what I will be running after its all setup and OC'ed
Seems like the issue lies with the way they configured the CCX cores to access the memory, only 2 out of the four can access the memory directly and this causing an issue. I wonder why then did they not make all CCX access the memory? Can someone with an expertise in CPU design answer what would be the drawback to allowing the other CCX cores to access the memory as well? Does it increase the die size, is there some other penalty?