diff --git a/src/webgpu/api/operation/sampling/sampler_texture.spec.ts b/src/webgpu/api/operation/sampling/sampler_texture.spec.ts
new file mode 100644
index 000000000000..1304d2d81608
--- /dev/null
+++ b/src/webgpu/api/operation/sampling/sampler_texture.spec.ts
@@ -0,0 +1,270 @@
+export const description = `
+Tests samplers with textures.
+
+- test that you can use the maximum number of textures
+  with the maximum number of samplers.
+`;
+
+import { makeTestGroup } from '../../../../common/framework/test_group.js';
+import { assert, range } from '../../../../common/util/util.js';
+import { MaxLimitsTest, TextureTestMixin } from '../../../gpu_test.js';
+import { TexelView } from '../../../util/texture/texel_view.js';
+
+export const g = makeTestGroup(TextureTestMixin(MaxLimitsTest));
+
+g.test('sample_texture_combos')
+  .desc(
+    `
+Test that you can use the maximum number of textures with the maximum number of samplers.
+
+The test works by making the maximum number of texture+sampler combos.
+Each texture is [maxSamplersPerShaderStage, 1] in size where each texel is [textureId, samplerId]
+A function "useCombo<StageNum>(comboId)" is made that returns stage[stageNum].combo[comboId].texel[id, 0]
+or to put it another way, it returns the nth texel from the nth combo for that stage.
+
+These are read in both the vertex shader and fragment shader and written to a [maxSamplerPerShaderStage, 2]
+texture where the top row is the values from the vertex shader and the bottom row from the fragment shader.
+
+The result should be a texture that has a value in each texel unique to a particular combo
+`
+  )
+  .fn(t => {
+    const { device } = t;
+    const { maxSampledTexturesPerShaderStage, maxSamplersPerShaderStage, maxBindingsPerBindGroup } =
+      device.limits;
+
+    assert(maxSampledTexturesPerShaderStage < 0xfffe);
+    assert(maxSamplersPerShaderStage < 0xfffe);
+
+    const maxTestableCombosPerStage = t.isCompatibility
+      ? Math.min(maxSampledTexturesPerShaderStage, maxSamplersPerShaderStage)
+      : maxSampledTexturesPerShaderStage * maxSamplersPerShaderStage;
+
+    const textures: GPUTexture[] = [];
+    const declarationLines: string[] = [];
+    const groups: GPUBindGroupEntry[][] = [[]];
+    const layouts: GPUBindGroupLayoutEntry[][] = [[]];
+    const textureIds = new Set<string>();
+    const samplerIds = new Set<string>();
+    // per stage, per texel, each texel has 2 numbers, the texture id, and sampler id
+    const expected: number[][][] = [[], []];
+
+    function addResource(stage: number, resourceId: string, resource: GPUTextureView | GPUSampler) {
+      let bindGroupEntries = groups[groups.length - 1];
+      let bindGroupLayoutEntries = layouts[groups.length - 1];
+      if (bindGroupEntries.length === maxBindingsPerBindGroup) {
+        bindGroupEntries = [];
+        bindGroupLayoutEntries = [];
+        groups.push(bindGroupEntries);
+        layouts.push(bindGroupLayoutEntries);
+      }
+      const resourceType = resource instanceof GPUSampler ? 'sampler' : 'texture_2d<f32>';
+      const binding = bindGroupEntries.length;
+      declarationLines.push(
+        `    @group(${groups.length - 1}) @binding(${binding}) var ${resourceId}: ${resourceType};`
+      );
+      bindGroupEntries.push({
+        binding,
+        resource,
+      });
+      bindGroupLayoutEntries.push({
+        binding,
+        visibility: stage === 0 ? GPUShaderStage.VERTEX : GPUShaderStage.FRAGMENT,
+        ...(resource instanceof GPUSampler
+          ? {
+              sampler: {},
+            }
+          : {
+              texture: {},
+            }),
+      });
+    }
+
+    function addTexture(stage: number, textureNum: number) {
+      const textureId = `tex${stage}_${textureNum}`;
+      if (!textureIds.has(textureId)) {
+        textureIds.add(textureId);
+        const texture = t.createTextureTracked({
+          format: 'rgba8unorm',
+          size: [maxSamplersPerShaderStage, 1],
+          usage: GPUTextureUsage.TEXTURE_BINDING | GPUTextureUsage.COPY_DST,
+        });
+        textures.push(texture);
+        // Encode an rgba8unorm texture with rg16uint data where each texel is
+        // [(textureId + 1) | (stage << 15), {samplerId + 1}]
+        // The +1 is to avoid 0.
+        const data = new Uint16Array(maxSamplersPerShaderStage * 2);
+        const rg = (textureNum + 1) | (stage << 15);
+        for (let x = 0; x < maxSamplersPerShaderStage; ++x) {
+          const offset = x * 2;
+          const samplerNum = x + 1;
+          data[offset + 0] = rg;
+          data[offset + 1] = samplerNum;
+        }
+        device.queue.writeTexture({ texture }, data, {}, [maxSamplersPerShaderStage]);
+        addResource(stage, textureId, texture.createView());
+      }
+      return textureId;
+    }
+
+    const kAddressModes = ['repeat', 'clamp-to-edge', 'mirror-repeat'] as const;
+    const getAddressMode = (hash: number, depth: number) => {
+      return kAddressModes[
+        ((hash / Math.pow(kAddressModes.length, depth)) | 0) % kAddressModes.length
+      ];
+    };
+
+    function addSampler(stage: number, samplerNum: number) {
+      const samplerId = `smp${stage}_${samplerNum}`;
+      if (!samplerIds.has(samplerId)) {
+        const samplerNum = samplerIds.size;
+        samplerIds.add(samplerId);
+        // try to make each sampler unique. This is because some backends
+        // coalesce samplers with the same settings.
+        const addressHash = samplerNum >> 3;
+        const sampler = device.createSampler({
+          minFilter: samplerNum & 1 ? 'linear' : 'nearest',
+          magFilter: samplerNum & 2 ? 'linear' : 'nearest',
+          mipmapFilter: samplerNum & 4 ? 'linear' : 'nearest',
+          addressModeU: getAddressMode(addressHash, 0),
+          addressModeV: getAddressMode(addressHash, 1),
+          addressModeW: getAddressMode(addressHash, 2),
+        });
+        addResource(stage, samplerId, sampler);
+      }
+      return samplerId;
+    }
+
+    // Note: We are storing textureId, samplerId in the texture. That suggests we could use rgba32uint
+    // texture but we can't do that because we want to be able to set the samplers to linear.
+    // Similarly we can't use rgba32float since they're not filterable by default.
+    // So, we encode via rgba8unorm where rg is a 16bit textureId and ba is a 16bit samplerId
+    const code = `
+    // maxTestableCombosPerStage: ${maxTestableCombosPerStage}
+
+    fn sample(t: texture_2d<f32>, s: sampler, validId: u32, currentId: u32, c: vec4f) -> vec4f {
+      let size = textureDimensions(t, 0);
+      let uv = vec2f((f32(currentId) + 0.5) / f32(size.x), 0.5);
+      let v = textureSampleLevel(t, s, uv, 0);
+      return select(c, v, currentId == validId);
+    }
+
+    ${range(
+      2,
+      stage => `
+      fn useCombos${stage}(id: u32) -> vec4f {
+        var c: vec4f;
+${range(maxTestableCombosPerStage, i => {
+  const texNum = (i / maxSamplersPerShaderStage) | 0;
+  const textureId = addTexture(stage, texNum);
+  const smpNum = i % maxSamplersPerShaderStage;
+  const samplerId = addSampler(stage, smpNum);
+  expected[stage].push([(texNum + 1) | (stage << 15), smpNum + 1]);
+  return `        c = sample(${textureId}, ${samplerId}, ${i}, id, c);`;
+}).join('\n')}
+        return c;
+      }
+    `
+    ).join('\n\n')}
+
+${declarationLines.join('\n')}
+
+    struct VOut {
+      @builtin(position) pos: vec4f,
+      @location(0) value: vec4f,
+    };
+
+    @vertex fn vs(@builtin(instance_index) iNdx: u32) -> VOut {
+      return VOut(
+        vec4f(0, 0, 0, 1),
+        useCombos0(iNdx),
+      );
+    }
+
+    @fragment fn fs(vin: VOut) -> @location(0) vec4u {
+      let ndx = u32(vin.pos.x);
+      let f = select(vin.value, useCombos1(ndx), vin.pos.y > 1.0);
+
+      // We're putting two u16 values in the source data but as rgba8unorm.
+      // Convert them back to u32 then split them back into two u16s
+      let bytes = pack4x8unorm(f);
+      return vec4u(bytes & 0xffff, bytes >> 16, 0, 0);
+    }
+    `;
+
+    const module = device.createShaderModule({ code });
+    const bindGroupLayouts = layouts.map(entries => device.createBindGroupLayout({ entries }));
+
+    const pipeline = device.createRenderPipeline({
+      layout: device.createPipelineLayout({ bindGroupLayouts }),
+      vertex: {
+        module,
+      },
+      fragment: {
+        module,
+        targets: [{ format: 'rg16uint' }],
+      },
+      primitive: { topology: 'point-list' },
+    });
+
+    const bindGroups = groups.map((entries, i) =>
+      device.createBindGroup({
+        layout: pipeline.getBindGroupLayout(i),
+        entries,
+      })
+    );
+
+    const renderTarget = t.createTextureTracked({
+      format: 'rg16uint',
+      size: [maxTestableCombosPerStage, 2],
+      usage: GPUTextureUsage.RENDER_ATTACHMENT | GPUTextureUsage.COPY_SRC,
+    });
+    textures.push(renderTarget);
+
+    const encoder = device.createCommandEncoder();
+    const pass = encoder.beginRenderPass({
+      colorAttachments: [
+        {
+          view: renderTarget.createView(),
+          loadOp: 'clear',
+          storeOp: 'store',
+        },
+      ],
+    });
+    pass.setPipeline(pipeline);
+    bindGroups.forEach((bindGroup, i) => pass.setBindGroup(i, bindGroup));
+    for (let y = 0; y < 2; ++y) {
+      for (let x = 0; x < maxTestableCombosPerStage; ++x) {
+        pass.setViewport(x, y, 1, 1, 0, 1);
+        pass.draw(1, 1, 0, x);
+      }
+    }
+    pass.end();
+
+    device.queue.submit([encoder.finish()]);
+
+    const expectedData = new Uint16Array(maxTestableCombosPerStage * 2 * 2);
+    for (let stage = 0; stage < 2; ++stage) {
+      expected[stage].forEach(([tid, sid], i) => {
+        const offset = (maxTestableCombosPerStage * stage + i) * 2;
+        expectedData[offset + 0] = tid;
+        expectedData[offset + 1] = sid;
+      });
+    }
+
+    const expTexelView = TexelView.fromTextureDataByReference(
+      'rg16uint',
+      new Uint8Array(expectedData.buffer),
+      {
+        bytesPerRow: maxTestableCombosPerStage * 4,
+        rowsPerImage: 2,
+        subrectOrigin: [0, 0, 0],
+        subrectSize: [maxTestableCombosPerStage, 2],
+      }
+    );
+
+    const size = [maxSamplersPerShaderStage, 2];
+    t.expectTexelViewComparisonIsOkInTexture({ texture: renderTarget }, expTexelView, size);
+
+    textures.forEach(texture => texture.destroy());
+  });