Quantcast
[ 3 / biz / cgl / ck / diy / fa / g / ic / jp / lit / sci / tg / vr / vt ] [ index / top / reports / report a bug ] [ 4plebs / archived.moe / rbt ]

Due to resource constraints, /g/ and /tg/ will no longer be archived or available. Other archivers continue to archive these boards.Become a Patron!

/g/ - Technology

Search:


View post   

[ Toggle deleted replies ]
>> No.70847792 [View]
File: 995 KB, 1543x1615, 1544650877708.jpg [View same] [iqdb] [saucenao] [google] [report]
70847792

>>70847759
You know shills are desperate when they have to resort to namecalling.

>> No.70605536 [View]
File: 995 KB, 1543x1615, 1544650877708.jpg [View same] [iqdb] [saucenao] [google] [report]
70605536

>>70600550
OOOOOOOOOOOOOOOOOOOOOOF

>> No.69502698 [View]
File: 995 KB, 1543x1615, 1524598363512.jpg [View same] [iqdb] [saucenao] [google] [report]
69502698

>>69497159
>>69497953
>fell for the 970 meme

4690k
390x

>> No.69496986 [View]
File: 995 KB, 1543x1615, 1544650877708.jpg [View same] [iqdb] [saucenao] [google] [report]
69496986

>>69495598
Right but high end NAVI is set to outperform the rtx 2080ti. Navi en route this year is set to more or less match the RTX 2080 for under $400 pricetag for 2 reasons:
1.) Cheap 8-12 GB of GDDR6X vRAM
2.) Higher FP32 utilization

We already know that on paper the Vega 56 should in theory match the RTX 2080 because they have similar FP32 compute throughput but for some reason this is being throttled most likely to engineering mistake at the silicon level. All AMD has to do is fix GCN or create a new fresh GPU architecture where this poor FP32 utilization problem is solved at the silicon level.

Vega 56 MSRP is already $400 so a re-release on a better GCN/new gpu architecture with cheap GDDR6X vRAM allon 7nm lithography is absolutely possible for under $400.

>> No.69404359 [View]
File: 995 KB, 1543x1615, 1544650877708.jpg [View same] [iqdb] [saucenao] [google] [report]
69404359

Think of it this way you mongs: breaking down the 16GB/s offered by 16X PCI-Express 3.0 to GPUs into 1 millisecond transactions that's JUST ~16MB of max transfer rate. Once GPUs start using 10-12GB of vRAM for extremely detailed textures at 120Hz 4K res that 1ms transfer rate is going to be extremely important. It's going to be vital for stuff like VR.

Say you have 1GB of textures you want to load from RAM to vRAM

3.0: 64 milliseconds
5.0: 16 milliseconds

120 Hz frame times: 8ms

>> No.69350379 [View]
File: 995 KB, 1543x1615, the cute menace.jpg [View same] [iqdb] [saucenao] [google] [report]
69350379

>>69346285

>> No.69347698 [View]
File: 995 KB, 1543x1615, 1544650877708.jpg [View same] [iqdb] [saucenao] [google] [report]
69347698

>>69345003
J U S T
U
S
T

>> No.69202696 [View]
File: 995 KB, 1543x1615, 1544007939462.jpg [View same] [iqdb] [saucenao] [google] [report]
69202696

>>69202498

>> No.68951543 [View]
File: 995 KB, 1543x1615, 1544650877708.jpg [View same] [iqdb] [saucenao] [google] [report]
68951543

>>68951508
Let's not get too hyped up, now.

>> No.68900511 [View]
File: 995 KB, 1543x1615, 1544007939462.jpg [View same] [iqdb] [saucenao] [google] [report]
68900511

>>68900414

>> No.68830572 [View]
File: 995 KB, 1543x1615, the cute menace.jpg [View same] [iqdb] [saucenao] [google] [report]
68830572

>>68826307
They're starting to feel the heat; their industrial chillers must be acting up.

>> No.68796181 [DELETED]  [View]
File: 995 KB, 1543x1615, the cute menace.jpg [View same] [iqdb] [saucenao] [google] [report]
68796181

>>68796143
Double the cores for double the fun and even double the competition can't compete.



Navigation
View posts [+24] [+48] [+96]